arXiv Daily Digest - 2026-07-02

Software diagrams are difficult to edit through human-friendly interfaces because edits expressed in natural language must still preserve visual layout, editable structure, and semantic relationships. As a step forward, we present SAGE, a browser-based tool for prompt-guided editing of Draw.io and Mermaid-style engineering diagrams. The tool maps diagrams into an editable graph representation, translates natural language requests into structured edit intents, analyzes those intents into graph-oriented operation steps, validates and repairs common Draw.io XML issues, and stores successful results as recoverable versioned artifacts. This design separates structured state management from model-driven interpretation, while acknowledging that some prompt-guided XML edits remain model-assisted. The tool also supports direct canvas editing and a secondary mask-based image-editing workflow. We evaluate the system using unit tests and a Kubernetes architecture case study, measuring structural validity, edit success, preservation of unrelated elements, and failure causes.

Task-based chatbots are nowadays widely adopted software systems, usually integrated into real-world applications and communication channels, designed to assist users in completing tasks through conversational interfaces. Like any other software, even chatbots are prone to bugs. Despite their increasing pervasiveness in everyday activities, existing techniques for assessing their quality still exhibit several limitations, such as the simplicity of generated test scenarios and oracle weaknesses. In this paper, we present LexTester, an automated model-based testing technique for Amazon Lex chatbots. The technique explores the conversational space of the chatbot under test to generate a Dialog Graph of all possible interactions, from which an executable test suite is generated according to different coverage strategies. LexTester was evaluated against the state-of-the-practice chatbot testing tool Botium on five Amazon Lex chatbots, consistently outperforming it in all subjects, generating more tests with nearly double complexity, achieving overall 83-95% coverage of conversational elements, and improving fault detection effectiveness by up to four times at comparable time costs.

Active learning reduces labeling cost by querying the most informative unlabeled samples, but standard coreset methods ignore known data symmetries and can waste budget on transformed versions of the same instance. We propose GRINCO, a group-invariant coreset framework that performs acquisition in the quotient space induced by a transformation group, so that selection operates on orbits rather than raw samples. The method uses either canonical representatives or learned orbit-separating invariant embeddings to define practical quotient metrics, and combines quotient-space k-center selection with invariant training through an orbit-averaged loss. We further derive a generalization bound that relates excess orbit-averaged risk to quotient-space coverage, label uncertainty, and intra-orbit variability. Experiments on synthetic scale-invariant data and image benchmarks with rotation-induced redundancy show that GRINCO improves orbit coverage and achieves stronger label efficiency than conventional coreset baselines, especially when group-induced redundancy is substantial.

Robotics foundation models (RFMs) are making general-purpose robots increasingly practical for factory deployments. While RFM serving systems are central to this vision, existing systems are largely shaped by a single-robot, single-model assumption: inference is treated as an edge-computing problem handled by an on-robot or dedicated nearby GPU, and the serving objective is to minimize the latency of a single action model. In this paper, we propose ROSA, an RFM serving system for robot factories designed around three key principles. First, ROSA adopts shared GPU-pool serving, allowing a fleet of robots to access powerful server-class GPUs over the network in order to improve inference performance, battery duration, and GPU utilization. Second, ROSA provides a robotics-aware programming abstraction and system design that supports multi-model pipelines, per-task performance requirements, and failure handling. Third, ROSA uses factory-objective-driven scheduling to maximize SLO-qualified factory productivity rather than minimizing individual request latency. We implement ROSA on top of Ray Serve for distributed orchestration, with vLLM, PyTorch, and JAX as model-serving backends, and evaluate it on both real robots and synthetic large-scale workloads. The results show that ROSA improves factory productivity by up to 12.06x over conventional dedicated serving systems.

Generative AI is shifting software engineering from a practice organized around scarce implementation effort toward one organized around abundant, low-cost code production. This shift changes the central engineering problem: not whether AI can generate useful code, but how engineers organize architectures, tools, evidence, and feedback loops so that AI-mediated development remains inspectable, correctable, and maintainable. We study this problem through a first-person case study: a 12-week development effort in which a single expert software engineer used frontier AI coding agents to build a document accessibility remediation system. The empirical record comprises 88 contemporaneous field notes, 420 KLOC of production code, and 1.16 MLOC of tests, lints, supporting documentation, and agent tooling. From this record, we develop a candidate middle-range theory of governance conversion, expressed as a process model explaining how high-velocity agentic implementation becomes governable. The model explains how agentic implementation velocity surfaces recurring structural failure classes, and how engineering judgment sustains velocity by converting those failures into durable governance mechanisms. In contrast to existing governance models that derive controls from known obligations, governance conversion explains how controls are discovered from failures that become visible only during agentic work. We use our model to make testable predictions and to describe implications for software engineering research and practice.

The evaluation of long-term video quality understanding remains an open challenge for large vision-language models (LVLMs). Existing video quality benchmarks predominantly focus on short clips and isolated distortions, overlooking the temporal continuity, cumulative degradation, and reasoning complexity inherent in long-duration content. To address these limitations, we present LongVQUBench, a comprehensive benchmark for long-term video quality understanding. LongVQUBench contains over 1200 diverse videos spanning movies, documentaries, surveillance footage, egocentric recordings, and animated content, accompanied by 1500 multiple-choice and open-ended questions for validation and testing. To assess perceptual reasoning across different temporal scopes, we introduce three progressively complex evaluation levels: (i) local event quality understanding (LQU) for analyzing localized distortions; (ii) cross-event quality reasoning (CQR) for integrating multiple degraded events; and (iii) global quality understanding (GQU) for holistic perceptual evaluation over extended durations. Furthermore, a needle distortion question-answering (NDQA) paradigm is embedded across all three levels, where spatial or temporal artifacts are sparsely inserted to probe fine-grained detection and reasoning capabilities. Extensive experiments on 14 state-of-the-art LVLMs reveal significant performance degradation with increasing video length and reasoning depth, highlighting their limited capacity for long-range temporal integration and perceptual attribution. We envision LongVQUBench as a foundational step toward the systematic, hierarchical, and explainable evaluation of LVLMs' long-term video quality understanding.

While Large Language Model (LLM) agents demonstrate proficiency in static benchmarks, their deployment in real-world scenarios is hindered by the dynamic nature of user queries, tool sets, and interaction dynamics. To address this generalization gap, we formalize OpenAgent (Tool-Use Agent in Open-World), a problem setting characterized by distributional shifts across query, action, observation, and domain dimensions. To systematically diagnose its impact, we construct a controlled sandbox environment where we define fine-grained environmental shifts across a four-tier hierarchy, Perception, Interaction, Reasoning, and Internalization, and conduct a comprehensive series of experiments. Our analysis yields a series of key insights, demonstrating that agents trained via both Supervised Fine-Tuning(SFT) and Reinforcement Learning suffer from varying degrees of performance degradation when confronting open environmental shifts. Building on these insights, we propose Perturbation-Augmented Fine-Tuning, a disturbance-based intervention strategy for SFT that lays the foundation for enhancing agent robustness and utility in realistic environments. Our code will be released at: https://github. com/LAMDA-NeSy/OpenAgent.

High-throughput RLHF systems often decouple rollout generation from policy optimization, leading to the use of stale rollouts during learner updates. In this work, we study the effect of such staleness in asynchronous GRPO. We make the behavior policy explicit in the GRPO surrogate objective and distinguish between the surrogate-gradient mapping used by the learner and the true total derivative of a distribution-dependent population objective. Under assumptions of local boundedness, distributional smoothness, and behavior-policy smoothness, we show that stale rollouts introduce a per-step surrogate-gradient bias of order O(S * eta), where S denotes the maximum rollout lag and eta denotes the learning rate. We further derive a conditional collapse-time scaling law: when within-cycle drift remains below a batch-level clipping radius, collapse is governed primarily by cumulative learner drift T * eta; when the stale-rollout constraint is active, stability instead depends explicitly on S * eta. This yields a two-constraint stability condition eta << min{R_batch / (S * G_upd), R_crit / (T * G_upd)}, explaining why the maximum stable learning rate may appear weakly dependent on staleness in the horizon-limited regime.

Spatio-temporal point-process models must often generalise across space when local event histories are sparse. We study whether exogenous spatial context can compensate in such regimes. Using a fixed log-Gaussian Cox process backbone, we compare an event-only model with the same model augmented by AlphaEarth embeddings as linear spatial context. We evaluate spatial transfer on emergency medical services (EMS) forecasting across eight held-out regions, fixed forecast anchors, and a sweep over history length $w$, using only AlphaEarth (AE) embeddings available strictly before each anchor. AE improves out-of-region predictive performance across all history regimes, with the largest gains under scarce histories: approximately $2$--$6\times$ multiplicative improvements at $1-2$ weeks, tapering to roughly $10$--$20\%$ at $w=20$--$104$ weeks. These results show that contextual information can substantially stabilise spatially transferred point-process forecasts when event history is limited.

We investigate Gaussian process (GP) bandit optimization with quantum kernels, assuming the mean reward function lies in the reproducing kernel Hilbert space (RKHS) induced by the quantum kernel. This setting is motivated by NISQ-era tasks such as quantum control, state preparation and variational quantum algorithms. While quantum kernels can offer a `quantum advantage' via domain-specific inductive biases, naïvely using full, high-dimensional kernels increases model complexity and information gain, leading to higher cumulative regret and poor learnability. To address this, we propose projected quantum kernels and classical kernel approximation techniques that reduce feature dimensionality while preserving key quantum properties. Using these approximate kernels, we develop misspecified GP bandit algorithms and derive regret bounds that characterize the trade-off between approximation error and information gain. The regret bounds provide principled guidance for selecting the optimal model complexity. Empirically, our methods outperform full quantum kernels in sample efficiency, while substantially reducing computational overhead, enabling scalable GP optimization for quantum-native applications.

While inference-time scaling has improved the reasoning abilities of large language models (LLMs), the need to generate long chains-of-thought (CoTs) is a computational bottleneck. Thus, in contrast to sequential scaling methods like CoT, recent parallel scaling techniques instead use fork and join (FJ) primitives to divide work across multiple LLM threads. However, in the fork-join paradigm, threads are typically transient and do not communicate pointwise with one another which limits scalability. To tackle this, we introduce Message Passing Language Models (MPLMs), a framework for LLM reasoning in which threads communicate directly via lightweight send and receive primitives. MPLMs enable efficient scaling through two key mechanisms: (1) reduced communication costs, achieved by avoiding redundant context sharing, and (2) preemption, which allows threads to terminate early based on partial information from their peers. We demonstrate the promise of MPLMs on 3 classes of tasks. First, on Sudoku puzzles, we show that MPLMs require an asymptotically smaller context than both serial CoT and parallel FJ. We then fine-tune a single model to solve 25 x 25 puzzles that remain challenging for standard CoT and FJ approaches, as well as frontier reasoning models without tools. Second, on 3-SAT puzzles, the capability of preemption allows termination of unpromising branches, which results in improved efficiency. Finally, we show that appropriately prompted large pre-trained models follow the MPLM protocol, achieving competitive results on long-context question answering relative to popular fork-join approaches.

Memory has emerged as a cornerstone of modern LLM-based agents, supporting their evolution from single-turn assistants to long-term collaborators. However, memory is not always beneficial: retrieved memories often induce a critical issue of sycophancy, causing agents to over-align with the user at the cost of factual accuracy or objective reasoning. Despite this emerging risk, existing memory benchmarks primarily evaluate whether memories are correctly stored, retrieved, or updated, while overlooking how retrieved memories influence downstream reasoning and decision-making. To bridge this gap, we propose MemSyco-Bench, a comprehensive benchmark for evaluating memory-induced sycophancy in agent systems. MemSyco-Bench measures when memory should influence a decision and how valid memory should be used. Specifically, it covers five tasks that assess whether agents can reject memory as factual evidence, respect its applicable scope, resolve conflicts between memory and objective evidence, track memory updates, and use valid memory for personalization. All related resources are collected for the community at https://github.com/XMUDeepLIT/MemSyco-Bench.

The deployment of Large Language Models (LLMs) with extended context windows is increasingly constrained by the linear growth of Key-Value (KV) cache memory. Vector Quantization (VQ), particularly Residual Quantization (RQ), is a promising approach for pushing KV cache storage toward the sub-1-bit regime by progressively encoding residuals with small codebooks. However, most VQ methods still rely on standard $\ell_2$ $K$-means as the core codebook-learning primitive. We identify a subtle high-dimensional issue of this primitive: Euclidean centroid averaging can induce centroid shrinkage, which weakens the angular alignment term in the $\ell_2$ distortion and makes directional preservation harder. To address this issue, we propose Gain-Shape $K$-means (GSKM), a drop-in replacement for $K$-means that improves directional fidelity while matching, and in some regimes improving, $\ell_2$ distortion. We then build Gain-Shape Residual Quantization (GSRQ) by incorporating a weighted extension of GSKM into an RQ pipeline. On LLaMA-3-8B, GSRQ substantially improves over strong KV cache quantization baselines across bit rates. At 1-bit, it improves the average accuracy across LongBench tasks from 11.34 to 33.54, a gain of 22.20 percentage points over VQLLM.

Large input spaces and complex inter-operation dependencies make black-box REST API testing challenging. AutoRestTest combines a Semantic Property Dependency Graph, multi-agent reinforcement learning, and large language models to intelligently explore large API input spaces. In the SBFT 2026 REST League, AutoRestTest ranked first in all three evaluation categories -- fault detection, overall efficiency, and overall effectiveness -- on 11 APIs (317 operations, approximately 29 per API), averaging 67.09 unique server errors and 17.27 successfully processed operations per API under a one-hour testing budget.

Computer-assisted synthesis planning breaks target molecules into accessible precursors using large libraries of reaction rules that assign each transformation a deterministic, interpretable label. But chemistry is long-tailed, making manual encoding intractable, and existing tools rely on fixed rulesets that cannot adapt to new chemistries. Here we present a fully automated pipeline in which a multi-agent framework of large language models (LLMs) classifies reactions and writes the rules themselves across 665,901 US patent reactions, generating each rule under a verification loop that tests it against the corpus. It expands a standard taxonomy from 68 to 14,073 classes without human curation. With a lightweight fingerprint classifier, it classifies 97.7\% of unseen reactions, matching a leading proprietary classifier while resolving chemistry more finely and extending on demand to chemistry outside its training distribution. The result is a living reactivity database and a general route to turning generative models into reliable, self-expanding symbolic systems.

We study a broad class of graphical models whose independencies correspond to vertex separation in mixed graphs with directed, undirected, and bidirected edges, that are capable of encoding independence structures arising from feedback, latent and selection mechanisms. In particular, we introduce separable graphs, in which each missing edge implies the existence of a separating set for its endpoints, and essentially separable graphs, those graphs separation equivalent to a separable graph. We show that these models include many existing graph families used to define graphical models an provide several characterizations of separable graphs and essentially separable graphs. We also provide multiple characterizations of separation equivalence for separable graphs. One is a graphical characterization in terms of ordinary graph properties, extending earlier results for specific subfamilies Another is a separational characterization depending only on graph separation properties. Finally, we provide a canonical representation for the equivalence classes of essentially separable graphs and develop an algorithm that, under suitable assumptions, identifies the equivalence class of any essentially separable graph.

Complexity and interpretability rarely coincide: systems rich enough for complex behaviours to emerge are usually too opaque to question, while transparent ones are too simple for anything complex to emerge. A single large language model (LLM) is a static artefact, hardly exhibiting any of the emergent properties we associate with life. This changes through interaction: populations of LLMs display emergent dynamics absent from isolated models. Furthermore, LLMs can be endowed with persistent memory, tools and shared skills, and the capacity to initiate actions unprompted, i.e., turning LLMs agentic. In this paper, we argue that such collectives of agents can serve as a computational substrate for Artificial Life (ALife) research. Critically, since the agents communicate in natural language, their collective behaviour can be directly interrogated by examining textual traces and asking the agents themselves. We outline the notion of interpretability in language-model research and extend it for collectives of agents. Lastly, we survey recent examples of agentic LLM collectives that already instantiate the idea of agentic substrates, from controlled experiments to deployments in the wild.

Memory-based discrete vision-language navigation (VLN) agents must act under partial observability, yet even strong frozen backbones remain vulnerable at test time. Two common failure modes are stale historical evidence at memory readout and inefficient local backtracking during action selection. We present DART-VLN, a training-free test-time control framework for discrete VLN. DART-VLN combines Test-Time Memory Decay, a read-side memory reweighting rule that suppresses stale and redundant evidence without rewriting stored content, with Anti-Loop Regularization, a lightweight next-hop penalty that discourages immediate reversals during action selection. The framework introduces no new learnable parameters and leaves the learned backbone unchanged. Experiments on R2R and REVERIE show a consistent pattern: decay-only provides stable read-side gains, while decay+anti-loop achieves the best overall quality-efficiency trade-off, yielding shorter trajectories, lower runtime, and improved navigation performance in key settings. Behavioral analysis further confirms that anti-loop regularization reduces local backtracking and improves path efficiency under frozen backbones. Overall, the results show that modest test-time control can make memory-based discrete VLN more reliable and efficient without retraining.

Program comprehension is a central research topic in software engineering, focusing on how developers understand a program's structure, behavior, and intent. Eye-tracking studies have traditionally relied on display-based measurements, where gaze positions are represented as screen coordinates. However, syntax-based analyses have recently emerged. Prior work proposed methods to convert eye movements into transitions between nodes in an abstract syntax tree, but the relationship between task correctness and eye-movement features for specific syntactic elements remains unclear. This study converts eye-tracking data into transitions between syntactic nodes and analyzes fixation proportions and gaze transition patterns. We investigate the relationship between these patterns and task correctness, comparing correct and incorrect groups. Our results reveal distinct differences in gaze transition patterns between the two groups. In particular, successful participants exhibit more systematic transitions across syntactic elements, suggesting the use of structured reading strategies.

Therapy-induced cardiotoxicity is the leading non-oncological cause of treatment interruption in breast cancer patients, yet early, automated risk stratification from routine cardiac imaging remains an unsolved problem. We present EchoRisk, the first curated, multicentre, longitudinal echocardiography dataset with explicit cardiotoxicity labels, released as the primary technical reference for the EchoRisk-MICCAI 2026 challenge. The dataset comprises 422 patients enrolled in the EU-funded CARDIOCARE prospective study across five European sites, yielding 2,159 echocardiography videos across 1,123 clinical exams acquired at up to five longitudinal timepoints, alongside a dedicated cohort of 280 patients with baseline imaging for early cardiotoxicity prediction. Three clinically grounded tasks are defined: automated estimation of left ventricular ejection fraction from cine video (Task 1), classification of LV dysfunction from longitudinal imaging (Task 2), and early prediction of therapy-induced cardiotoxicity from pre-therapy baseline echocardiography alone (Task 3). For each task we specify the evaluation protocol, primary and secondary metrics, and ranking procedure. We establish baseline performance using an R(2+1)D video backbone with LSTM aggregation trained from Kinetics-400 pretrained weights, demonstrating strong discriminative performance for cardiac functional assessment and LV dysfunction classification, while early cardiotoxicity prediction from a single pre-therapy video remains a significant open problem for the community. The dataset, evaluation code, and baseline implementations are publicly available to serve as a benchmark for further collaboration, comparison, and the creation of task-specific architectures in cardio-oncology.

Large language model (LLM)-based conversational agents (CAs) are now ubiquitous, creating new opportunities for AI-mediated behavior change. Their capacity to project nuanced personalities and adopt diverse metaphorical roles raises a design question: how should an agent's persona and personality be calibrated to the moment? Recent evidence suggests that (i) moderate personality expression outperforms low or high extremes on trust, enjoyment, and intention to adopt in goal-oriented tasks, and (ii) context-appropriate metaphors outperform static one-note assistants on user experience and uptake. Yet most CAs still fix both persona and style, risking misalignment when dynamics, urgency, and formality vary, for example in medical information seeking, fitness coaching, and reflective learning. We propose a Fluid Personality Framework that jointly adapts (1) the agent's metaphorical persona, such as coach, tutor, librarian, or tool, and (2) its personality expression intensity, low, medium, or high, as a function of task context, user goals and traits, and situational urgency. We sketch the framework and its core design dimensions.

Model organisms (MOs) - language models trained to exhibit undesired or unnatural behaviours - are frequently used as testbeds for evaluating white-box interpretability techniques. Current MOs are typically constructed via post-hoc supervised fine-tuning (SFT) on behavioural transcripts or synthetic documents. Prior research has shown that interpretability methods can easily identify hidden behaviours in these MOs. However, recent work suggests that such post-hoc training methods may make interpretability unrealistically easy. We investigate this claim by constructing a suite of 54 $\verb|OLMo2-1B|$- and $\verb|gemma-3-1b-it|$-based MOs trained with seven different techniques, including standard post-hoc SFT, post-hoc DPO, and more realistic integration of MO data into the OLMo post-training DPO phase. We use these MO variants to benchmark activation oracles, activation steering, logit lens, and sparse autoencoders. Our findings show that (i) MO interpretability depends strongly on training objective, target behaviour, model architecture, and training data generation pipeline; (ii) substantial variance remains even after controlling for differences in the strength of target behaviour expression; and (iii) our more realistic $\textit{integrated training}$ often yields less interpretable MOs than standard post-hoc methods. Our results cast substantial doubt on the validity of current MOs as interpretability proxies.

Radio-frequency (RF) sensing is a central modality for counter-unmanned-aerial-system (counter-UAS) defence because it exploits the control, telemetry, and video links between a drone and its operator. Reported accuracies for RF-based drone detection and identification are often very high, but many are obtained using cross-validation that splits a small number of continuous recordings into short segments. This can place near-duplicate slices of the same recording in both training and test partitions, creating data leakage. We study this leakage problem through theory and measurement. We formalise the optimism of segment-level cross-validation and show, using Cover's function-counting theorem, that a classifier can exactly memorise the recording-to-label map when the number of independent recordings, R, is small relative to the feature dimension, d. In particular, this can occur when 2R is less than or approximately equal to d. Under these conditions, naive accuracy approaches 1, and the inflation gap approaches 1 - ACC*, where ACC* is the Bayes accuracy. The inflation eases only once R grows beyond this separability threshold. A controlled synthetic experiment with 10 seeds confirms the predicted curves: naive balanced accuracy rises from the Bayes level toward 1.0 as recording-specific nuisance variation grows, while honest recording-grouped evaluation declines to chance, with a gap reaching about 0.5. On the public DroneRF dataset, pooled leave-one-recording-out cross-validation shows drone type identification, AR versus Bebop, collapsing from a naive macro-F1 of 0.74 to 0.46, the two-class chance level. A leakage-pathway ablation attributes essentially all of the inflation to segment-level leakage.

Financial markets evolve in response to real-world events reported in news, yet these drivers often remain implicit in text. To better explain market dynamics, event-market relations must be explicitly modeled through factual, company-centric, and environment-aware knowledge graphs. We present FinKG-News, a framework that automatically constructs such graphs by extracting news events as anchors linked to companies. Using FinKG-News as grounded evidence that integrates events, news, and company data, we develop an in-context learning architecture for credit risk report generation across three core financial dimensions. Automatic and human evaluations show that automated hallucination detection and quality assessment remain unreliable, making expert judgment indispensable. Our approach consistently outperforms baselines, improving quality by 19%-34% while reducing hallucinations. The source code and project resources are publicly available at: https://github.com/ichise-laboratory/FINKG-news.

Spatiotemporal point processes (STPPs) model event data in continuous time and space, with applications in mobility, epidemiology, and public safety. Recent neural STPPs span expressive intensity models, conditional density models, continuous-time latent dynamics, normalizing-flow spatial decoders, and score-based generative mechanisms. Yet comparison remains fragile because implementations differ in preprocessing, coordinate normalization, splits, likelihood conventions, and evaluation protocols. We present SEAHORSE, a unified framework for reproducible STPP experimentation. SEAHORSE formalizes neural STPPs through a common encode-evolve-decode interface and trains, tunes, and evaluates every model family under a single executable benchmark protocol with raw-coordinate likelihood reporting. This enables fair comparisons but, more importantly, controlled diagnostic studies. We pair SEAHORSE with HawkesNest, a synthetic stress-test suite, and show that increasing event-pattern complexity exposes each family's inductive bias, degrading some models sharply and leaving others stable. Code: https://github.com/YahyaAalaila/seahorse.

Large-scale crowd management requires pedestrian simulations that are both computationally efficient and compatible with feedback-based control. However, most open-source tools are either microscopic or not designed for network-scale closed-loop evaluation. This paper presents PedNStream (Pedestrian Network Flow Simulation), an open-source, Python-native simulator for macroscopic pedestrian network loading based on the Link Transmission Model (LTM). The framework extends LTM-based pedestrian models by incorporating stochastic link dynamics that capture diffusion and activity-induced variability, and replaces dynamic user equilibrium route choice with a utility-based formulation suited to uncertain, intervention-driven settings. PedNStream is implemented as a modular framework with built-in controller interfaces for interventions such as gating, flow separation, and route guidance. We evaluate the framework in a staged manner. Synthetic scenarios verify key mechanisms, including queue formation, spillback, congestion dissipation, and adaptive rerouting. Real-network experiments assess large-scale behavior and consistency with observed pedestrian counts. A closed-loop case study demonstrates controller integration, and a runtime analysis quantifies scalability. These results establish PedNStream as an efficient and practical testbed for large-scale pedestrian network simulation and control.

Reading order inference remains a critical bottleneck in the digitization of complex historical manuscripts, where pages contain multiple spatially interleaved reading streams, the canonical example being the Glossa Ordinaria layout, in which a central text is surrounded by commentaries that wrap around it in non-rectangular, non-convex regions. We present a training-free, graph-based framework: each OCR text line becomes a node in a directed candidate-transition graph, edges are scored by a weighted additive ensemble of two lightweight language-model signals (causal language model conditional likelihood and BERT next-sentence prediction, NSP; a third sentence-embedding signal was evaluated but did not improve reading order), and the global reading order is recovered as a degree-constrained directed path cover. To avoid the cascading "edge-theft" failures of greedy edge selection, we propose a max-regret inference rule that prioritizes commitments with high opportunity cost. We evaluate on synthetic Glossa Ordinaria grid layouts, on 23 ALTO page geometries (10 historical source pages plus mirrored and flipped variants), and on a 140-page multi-column English subset of OmniDocBench, comparing our method against the canonical recursive XY-cut (PaddleOCR PP-StructureV3) and two LayoutReader variants (layout-only and text+layout) on identical inputs. On wrap-around Glossa layouts our method recovers 95% of ground-truth successor edges on average vs. XY-cut's 50%; on the OmniDocBench multi-column subset it reaches 88% macro edge accuracy versus XY-cut's 75% and LayoutReader's 25%. The LayoutReader baselines transfer poorly due to a word-level vs. line-level granularity mismatch. We additionally verify mirror-invariance under horizontal and vertical page reflections: Our method changes by less than 1 percentage point, classical XY-cut by 2 points, and LayoutReader-T by up to 8 points.

Data assimilation models state dynamics conditioned on sequential observations, and has wide-ranging scientific applications. In the filtering setting, the goal is to model the posterior over the current state given all observations so far. Classical solutions typically make simplifying distributional or functional assumptions, e.g., linear-Gaussian systems, which can be inaccurate in many scenarios. In principle, particle filters (PFs) remove these assumptions, yet often collapse in high dimensions. Recent generative approaches learn conditional state transitions, but without principled Bayesian updates they do not recover the correct filtering posterior and can accumulate error over long horizons. In this work, we introduce Flow Proposal Particle Filters (FPPF), which learn a conditional generative model based proposal approximating the variance-minimizing optimal proposal for particle propagation. Conditioning on observations steers particles toward high-likelihood regions before weighting, reducing weight variance and delaying degeneracy. Since our proposal admits tractable likelihood evaluation, FPPF computes accurate importance weights and retains a Bayesian update step. We further extend FPPF to high-dimensional problems through localization strategies, adressing another standard PF failure mode. Extensive experiments on a variety of dynamical systems show that FPPF outperforms statistical baselines and other generative methods in non-linear, non-Gaussian, and high-dimensional regimes.

The success of deep learning models in classification and regression is widely attributed to the low-dimensional structure that real-world data tend to exhibit, despite their high-dimensional representation. This work attempts to provide a mathematical framework for binary classification on low-dimensional data, building on Cover's (1965) function-counting theory. With our framework, we aim to address the question of how the low-dimensional structure of the data affects the classification capabilities of learning models. Cover's theory relies on a general position assumption that blinds it to the underlying data structure. We refine this assumption to account for the low-dimensionality of the data and derive dichotomy counts that reflect the data structure. We further extend Cover's separation capacity and problem of generalization to the low-dimensional setting, enabling the impact of the underlying data structure on both to be analyzed.

Large Language Models (LLMs) represent one of the most significant advances in AI and natural language processing in recent years. Still, many pressing questions about their mechanisms, capabilities, and relationship to human cognition remain highly debated. This chapter aims to outline our current understanding of LLMs by discussing recent evidence on emerging capabilities and their mechanistic implementation within processing layers. We begin with a concise overview of the Transformer architecture, emphasizing how the attention mechanism enables training on massive datasets, allowing LLMs to function as generalist rather than specialized models. Next, we examine emergent LLM capabilities that appear to resemble aspects of human cognition, including symbolic reasoning, theory of mind, and deception strategies. Several studies provide evidence that LLMs can solve tasks previously thought to require human-like cognition. Other studies reveal insightful failure cases that shed light on the differences between human and LLM cognition. Alongside these findings, we review explainable AI approaches ranging from neuron activation analysis to circuit tracing. In the final section, we address current debates concerning what LLMs genuinely understand versus what they merely appear to understand. Prominent arguments against AI anthropomorphism point to the simplicity of LLM training objectives, claiming that LLM behavior is better explained by pattern memorization of training data than by genuine cognition. We argue that this standpoint is guided by misconceptions about optimization processes and cognitive capacity, and advocate for a more nuanced discussion of LLM cognition that neither dismisses the differences between humans and LLMs nor precludes the possibility of AI cognition through overly simplistic reductionist arguments.

In long-context use, large language models frequently synthesize answers from the meaning of a relevant context span rather than literally copy-pasting them. Identifying which attention heads perform this synthesis matters for interpreting long-context model behavior. Yet existing detectors miss these heads by construction: they reward heads whose attended token matches the generated token, a literal-copy criterion that captures where a head reads but not what it writes through its output-value (OV) circuit, the very mechanism that carries non-literal retrieval. We introduce Logit-Contribution Scoring (LOCOS), a write-aware detector that scores each head by the projection of its OV-circuit output onto the answer-token unembedding direction, contrasting needle and off-needle source positions in a single forward pass. Across three model families (Qwen3, Gemma-3, OLMo-3.1), mean-ablating the top LOCOS heads on the NoLiMa non-literal retrieval benchmark collapses ROUGE-L at lower head counts than prior attention-based detections; on Qwen3-8B, ablating 50 heads drives ROUGE-L from 0.401 to 0.000 while the strongest baseline still retains 0.292. The selected heads are retrieval-specific: parametric recall and arithmetic reasoning stay at baseline under the same ablation. On Qwen3-8B, the same ablation also drops MuSiQue from 0.55 to 0.08 and BABI-Long from 0.62 to 0.20, while a random-heads control stays within 0.05 of baseline.

Radiomics is the established approach for CT-based lung cancer phenotyping, yet comparisons with foundation models rarely isolate contributions of feature extractor, classification head, and segmentation choice, or test cross-cohort robustness. We benchmark five feature extractors (Curia, Curia-2, DINOv3, Radiomics2D, Radiomics3D), seven classification heads (TabPFN, TabICL, XGBoost, CatBoost, Random Forest, logistic regression, Ridge), and three segmentation regimes on five tasks: tumor volume and stage classification, 2-year survival prediction, histology classification, and age prediction. Models are trained on LUNG1 (n=338) and evaluated on an internal test set (n=84) and the external LUNG2 cohort (n=211), with worst-case cross-cohort performance as the primary metric. The dominant design factor is task-dependent: segmentation drives volume and stage classification, while classifier choice drives survival, histology, and age prediction. Radiomics is competitive for tumor volume, tumor stage and survival (partly due to label-derivation effects for the former); Curia variants reach comparable peak scores for survival; DINOv3 falls slightly short across tasks. Patch and slice aggregation have negligible impact. We recommend Curia with tumor segmentation and a CatBoost head as a safe default, achieving the best mean rank across the three primary clinical tasks, though task-specific selection consistently outperforms any cross-task default. When tumor delineations are unavailable, Curia-2 with lung segmentation and logistic regression offers a competitive alternative. All pipelines use a two-stage design suited to small cohort sizes where end-to-end fine-tuning would risk overfitting.

Recent research has increasingly focused on understanding how Transformers store and process knowledge, as well as how this knowledge can be edited. Research work in this area is often conducted in two phases: first, phenomena are explored on individual samples. Then, when results appear promising, more statistically robust experiments follow. To support the first phase, we propose KnowledgeDebugger, a GUI-based exploration tool for knowledge localization and editing in Transformers. Our tool - inspired by LM-Debugger - offers no-code access to the methods in EasyEdit, a widely used library of state-of-the-art Knowledge Editing approaches. We demonstrate the tool's effectiveness through case studies of recent findings in this field.

Most existing multitask learning approaches are limited by their reliance on task-specific loss functions tailored to the scale and type of each outcome. When outcomes differ across tasks, these losses are generally not directly comparable, which makes it difficult to formulate a unified objective and may limit information sharing across tasks. We propose a multitask transformation framework in which task-specific responses may differ through unknown monotone transformations. Motivated by high-dimensional biological applications in which the predictor dimension may diverge with the sample size while only a common subset of predictors is informative, we consider shared sparsity across tasks. Under this framework, we estimate the target functions and identify important predictors by optimizing a smoothed rank-based criterion with a group-Lasso penalty, implemented through a multitask deep neural network with a shared first layer. We establish the nonasymptotic excess-risk bounds, and variable-selection consistency for the proposed estimator. Simulation studies show that the proposed method achieves competitive prediction and variable-selection performance compared with competing approaches. Analyses of gene-expression studies with continuous, binary, and mixed outcomes further illustrate that the proposed method improves prediction and identifies biologically meaningful shared predictors.

Large language model (LLM)-based software engineering agents are increasingly developed to resolve software issues by generating patches from issue reports and code repositories. Bug reproduction tests (BRTs) are an important building block for such agents and have been shown useful for patch validation. However, it remains unclear whether BRTs can also help the more central stage of patch generation. We first conduct a preliminary study and find that directly using advanced BRT generators to guide patch generation is not beneficial: fail-to-fail BRTs can mislead agents, while even fail-to-pass BRTs bring limited or negative gains. Our analysis reveals two reasons: fail-to-pass BRTs may cover only one manifestation of the reported issue, leading to partial patches, whereas fail-to-fail BRTs are unreliable as direct patch-generation targets. Motivated by these insights, we propose SWE-Doctor, a software issue resolution agent that guides patch generation with runtime diagnoses derived from multi-faceted BRT executions. SWE-Doctor first generates multi-faceted BRTs for different behavioral requirements stated in the issue, then executes and debugs these BRTs to construct runtime-grounded diagnosis records, and finally uses the diagnoses together with localization information inferred during BRT generation to guide patch generation and reduce partial patches. We evaluate SWE-Doctor on Python bug-fixing issues from the widely adopted SWE-bench Verified and SWE-bench Pro across five LLM backends. SWE-Doctor consistently outperforms existing agents across all 10 LLM-benchmark combinations, achieving average resolution rates of 75.7% on SWE-bench Verified and 59.4% on SWE-bench Pro. In particular, on the more challenging SWE-bench Pro, SWE-Doctor improves the average resolution rate by 8.0-8.9 percentage points over the baseline agents.

Semantic trajectory analysis has recently emerged as an approach for modeling human movement by capturing implicit patterns and behaviors through semantic information (e.g., visitors' profiles and goals) beyond raw spatial paths to better understand why people move in certain ways. However, analyzing semantic trajectories in real-world scenarios remains challenging, as collecting high-quality data is costly and often lacks rich semantic information. Meanwhile, existing simulation tools require substantial technical expertise, which makes them difficult for practitioners to adopt. To address these limitations, the paper proposes ${SenseWalk}$, an interactive system that supports simulating semantic trajectories by LLM-powered agents. We develop a simulation workflow that combines LLMs and the social force model to balance physical plausibility and semantic coherence. A user-friendly interface is designed to facilitate users in customizing the simulation configuration and analyzing simulation outputs. We also conduct a quantitative experiment to evaluate the effectiveness of our simulation workflow, and a user study (n=12) to assess the usefulness and efficiency of our system.

Automatically detecting stress in speech provides an unobtrusive way to gain insights relevant to behavioral research or clinical assessment. This study investigates the automatic differentiation between a stressful and non-stressful situation, and the prediction of physiological and affective stress responses. Speech data was collected from 50 participants who either completed the Trier Social Stress Test (TSST) or a non-stressful control condition. With a processing pipeline that included speaker diarization and machine learning models, we achieved stress detection performance significantly above a mean baseline. Moreover, relevant physiological and affective stress responses were partially predictable from acoustic-prosodic features. Feature-importance analyses identified the most informative predictors contributing to model performance. The findings demonstrate that speech can serve as a meaningful and unobtrusive indicator of multiple dimensions of the human stress response.

Chest X-ray multi-label classification is a core task in intelligent medical imaging diagnosis. However, real clinical data often exhibit extreme long-tailed distributions, leading to degraded performance on rare diseases in tail classes. This issue is not only driven by data scarcity but also by two intrinsic factors:1) attenuation of tail-class lesion representations under complex anatomical backgrounds, and 2) dominance of head classes in modeling label co-occurrence relationships. To address these challenges, we propose TRCGL-Net. First, a learnable text-guided conditional diffusion model is employed to generate high-quality tail-class chest X-ray image samples under disease semantic constraints, improving data diversity and realism of rare disease patterns while alleviating class imbalance and preserving pathology-consistent semantics.Second, a channel reweighting mechanism is introduced to perform feature recalibration by emphasizing disease-relevant feature channels, thereby improving feature discriminability under long-tailed distributions.A class-aware attention mechanism is further applied to generate class-specific attention maps, enabling the model to localize disease-relevant regions and focus on fine-grained lesion areas.Finally, a graph convolution network based on label co occurrence is introduced to establish an information propagation mechanism among categories. Experiments on the PadChest dataset show that the proposed method achieves a tail-class mAP of 0.4904, an overall mAP of 0.4408, and an mAUC of 0.8989, outperforming state-of-the-art methods. TRCGL-Net effectively improves recognition performance for rare diseases under long-tailed distributions and mitigates the impact of extreme class imbalance in chest X-ray multi-label classification.

Trustworthy deployment of Agentic Retrieval-Augmented Generation (RAG) systems requires mechanisms for estimating when multi-stage reasoning pipelines may fail. This paper presents an uncertainty-aware Agentic Retrieval-Augmented Generation (RAG) framework in which planner, evaluator and generator stages produce uncertainty signals derived from semantic divergence and generator self-evaluation. These signals are propagated through a Bayesian Network (BN) to estimate system-level uncertainty and provide node-level indicators of potential failure points across the workflow. The approach is evaluated on StrategyQA and HotpotQA using GPT-3.5-Turbo and GPT-4.1-Nano, with Area Under the Receiver Operating Characteristic Curve (AUROC), Area Under the Accuracy-Rejection Curve (AUARC), Expected Calibration Error (ECE), and Brier Score used to assess discrimination, selective prediction and calibration. Results show that Bayesian propagation is more effective on HotpotQA, where uncertainty accumulates across multi-hop reasoning stages, while StrategyQA exposes limitations caused by miscalibration and unreliable upstream signals. The study positions Bayesian uncertainty propagation as a promising but preliminary mechanism for monitoring Agentic RAG systems, with future validation required in industrial domains such as Offshore Wind (OSW) maintenance decision support.

This paper introduces Svarna, a free, open-source, web-based corpus workbench for modern Greek. Svarna integrates five databases covering various registers, institutional, literary, dialectal, social media, and historical, to provide a total of more than 507 million words and around 29 million sentences. This platform addresses the chronic gaps in Greek language technology. Although various corpus resources exist, they are scattered across different platforms, and in many cases, institutional access is restricted or they are no longer available online. Svarna integrates these resources into a single interface that can be used without logging in, installation, or specialized training. This system provides a concordancer with KWIC marking capabilities, frequency analysis including register-by-register normalization, collocation extraction using mutual information, a dictionary of 93 Greek discourse markers providing distribution profiles, text-level analysis tools including n-grams, variants, and collocation networks, register comparison using log-ratio, regular expression search, and an optional LLM layer for pragmatic annotation and free research mode. This platform is built upon SQLite FTS5 full-text indexes provided via a FastAPI backend, deployed as Docker containers on Azure, and released under the MIT license. Source code, build scripts, and deployment configurations are publicly available on GitHub. Users can add their own corpora and deploy their own instances. This document describes the system design, corpus structure, and use cases demonstrating the various queries supported by the platform. Svarna serves as the first step in exploring available data and is expected to lay the foundation for more comprehensive research in the future.

Visual analytics (VA) plays an increasingly important role in supporting machine learning (ML) workflows. In the field of visualization, such approaches and techniques are referred to as VIS4ML. While ML models are mostly learned automatically, the corresponding ML workflows receive a variety of human inputs, such as data labelling, feature engineering, model architecture designing, hyper-parameter tuning, and so on. In this work, we surveyed over 200 VIS4ML papers to gain an understanding of how humans inject their knowledge into ML workflows through interactive visualization. We collected a corpus of VIS4ML papers from the IEEE VIS conferences in the past decade. We developed a coding scheme to facilitate the literature research from four perspectives: characteristics of ML, visualization, interaction, and actions. The analysis of the coded dataset allows us to observe different pathways that transfer human knowledge to ML workflows via interactive visualization. Building on the analysis, we explain the phenomena of VIS4ML using the conceptual model that views VA as model building and the information-theoretic cost-benefit analysis that reasons VA as for optimizing ML workflows. This work provides unequivocal evidence showing the merits of using VA in ML workflows. The full list of surveyed papers, along with all analysis results and figures, is available at https://vis4ml4hd.github.io/ml-knowledge-inject-va/.

Emotion recognition in natural language is a foundational challenge in affective computing, with critical implications for human-computer interaction, mental health support, and conversational AI. This paper presents a rigorous, unified zero-shot evaluation of three leading commercial large language models: Claude (claude-sonnet-4-6), ChatGPT (GPT-5.4), and Gemini (gemini-2.5-flash). The models were queried through their respective production APIs as of April 2026 on a fine-grained 13-class emotion classification task. Using a stratified 1,000-sentence sample from the boltuix/emotions dataset, which comprises 131,306 sentences across 13 categories, a single uniform prompt with no exemplars was applied identically across all models. Gemini achieves the highest accuracy (39.9%) and macro-F1 score (0.363), followed by GPT-5.4 (38.8%, macro-F1 = 0.291) and Claude (38.0%, macro-F1 = 0.159). All models excel on sarcasm and desire while consistently failing on love, confusion, and shame. McNemar tests reveal no statistically significant pairwise differences (p > 0.10), suggesting convergence at a shared zero-shot ceiling. Claude's markedly lower macro-F1 score exposes a class-imbalance prediction bias. These findings highlight the current limitations of frontier AI systems in zero-shot fine-grained emotion classification.

Realizing quantum neural networks (QNNs) in industry requires knowing which quantum computing paradigm suits which task. Motivated by AI accelerators and high-bandwidth memory, where die stacking makes wafer-level defect screening central to yield, we study WM-811K wafer-map defect classification (eight classes), comparing the dominant paradigms, continuous-variable (CV) and discrete-variable (DV), under controlled conditions. To isolate the quantum circuit as the sole variable, a shared convolutional backbone (~4.3M parameters) feeds interchangeable heads (classical dense, CV-QNN, or DV-QNN) as the only structural difference; each quantum head is scaled over three sizes (3, 4, 8 qumodes/qubits). The CV head consistently outperforms the DV head: at four qumodes/qubits it reaches 79.7 +/- 1.8% accuracy versus 61.6 +/- 1.4%, a non-overlapping 18-point gap. The advantage is sharpest on the spatially localized Edge-Loc class, easily confused with Scratch, which CV recovers with recall 0.66 +/- 0.06 while DV fails at every size (<=0.05), showing the structured CV layer better captures fine spatial distinctions between defect types. Training curves show the DV limitation is a representational-capacity ceiling, not an optimization failure; at the Fock cutoff used here (d = 2) the CV advantage reflects two intrinsic properties, a structured, neural-network-analogue layer and continuous phase-space encoding, not Hilbert-space dimensionality. On IBM hardware, DV accuracy holds at shallow depth, degrading only at the deepest circuit. Both quantum heads remain below the classical baseline (85.0%), but the controlled setting isolates where a structured head already helps and, as noise and scale improve, which paradigm can deliver practical advantage.

Time series are central to modern data mining applications, from industrial telemetry and server metrics to finance and physiology, yet time-series self-supervised learning often depends on view and augmentation choices that encode domain-specific invariances. We study how an SSL recipe behaves when its method-specific configuration is reused unchanged after the pretraining signal family changes, framing this as a fixed-recipe stress test rather than a comparison against optimally tuned methods. We introduce Latent Euclidean Next-Embedding Prediction Architecture (LeNEPA), a no-augmentation next-latent-token objective with a causal backbone. LeNEPA replaces the stop-gradient/EMA stabilization used by vanilla NEPA with SIGReg-based isotropy regularization and computes the predictive loss in a lightweight projected space that is discarded for evaluation. We compare LeNEPA with an ECG-tuned JEPA recipe under a fixed-horizon frozen-probe protocol on PTB-XL and Diag, a synthetic diagnostic corpus generated with Aionoscope. Both methods are retrained independently on each dataset while keeping their method-specific recipes unchanged. In this protocol, the ECG-tuned JEPA recipe is strong in-domain on PTB-XL but weaker when reused unchanged on Diag, whereas LeNEPA preserves useful frozen-probe gains on both datasets. Learning curves suggest faster early representation acquisition: LeNEPA reaches 80% of its final AUROC/AUPRC gain after 2--5k updates, compared with 5--10k updates for the faster JEPA readout. As a separate external frozen-encoder check, a CauKer-pretrained LeNEPA variant reaches 77.65% mean UCR-128 Random-Forest accuracy in a single-seed, best-checkpoint run, within 1.16 points of Mantis and within 0.24 points of MOMENT (77.89%). Overall, the results support no-augmentation latent prediction as a useful candidate recipe for low-retuning time-series SSL.

Time-series models are often evaluated by what they can forecast or classify, but those scores do not show whether their representations preserve the process state a user may want to inspect: event timing, phase, amplitude, frequency, or regime variables. We introduce Aionoscope, a generator-based diagnostic tool for debugging latent-state accessibility in frozen time-series representations. Aionoscope separates process generation from observation rendering, producing seeded synthetic streams with exact categorical and dense labels across mixture complexity and nuisance variation. We instantiate Aionoscope as Primitive Process Mixtures and evaluate 37 model-plus-adapter systems with a common pooled linear-probe protocol. The main result is a mismatch between coarse and fine-grained accessibility. Most systems make component presence easy to recover, but expose dense process state much less reliably: the highest observed dense-probe row reaches 0.689 mean masked $R^2$, while a dense-feature oracle reaches 0.999. This is the failure mode Aionoscope is designed to surface: a representation can look informative at the level of "what kind of signal is present" while hiding the timing, phase, amplitude, frequency, or regime variables needed for debugging.

Implicit neural representations (INRs) are well suited to cardiac motion estimation, providing continuous, compact representations of motion fields. However, fitting an INR to each image sequence is time-consuming and sensitive to the optimisation trajectory. Learned priors can help guide optimisation towards plausible motion fields and enable faster adaptation, but learning priors for cardiac motion INRs remains under-explored. In this work, we compare four strategies for learning cardiac motion priors, including a population prior learned by joint optimisation, a consensus prior obtained by weight averaging, auto-decoders, and meta-learning. Using short-axis tagged cardiac magnetic resonance images from the UK Biobank, we evaluate their impact on tracking accuracy, motion behaviour, and adaptation trajectory. All learned priors substantially improved early adaptation performance compared with random initialisation. While the simple consensus prior was effective, auto-decoders recovered large deformations faster during early adaptation. Meta-learning achieved strong early performance and maintained the best adaptation trajectory over 50 iterations.

Generative models learn data distributions that reside on a low-dimensional manifold within a higher-dimensional ambient space. Optimizing differentiable objectives on this manifold is challenging: the ambient loss landscape is high-dimensional, rugged, and non-convex. Direct gradient descent, blind to the manifold's geometry, quickly drifts off it. Diffeomorphic optimization starts from the observation that diffusion and flow models provide a map from the data manifold to a much simpler base space in which we perform gradient descent. Using differential geometry, we show this is equivalent to Riemannian gradient descent on the data manifold up to $\mathcal{O}(λ^2)$ corrections, keeping trajectories on-manifold by construction and yielding a smoother optimization surface. For protein design, we extend diffeomorphic optimization to the matrix Lie groups $\mathrm{SO}(3)$ and $\mathrm{SE}(3)$, deriving an autograd-compatible $\mathrm{SO}(3)$ gradient and a generalized adjoint-state method for backpropagation through Lie-group ODE solvers. Diffeomorphic optimization improves over tuned guidance on secondary-structure targeting with FrameFlow ($91.3\%$ vs. $63.3\%$ of residues in the Ramachandran target), outperforms OC-Flow on peptide binding affinity at $2\times$ the speed, and reduces Rosetta energies by thousands of units across the PDB test set for structures with hundreds of residues.

While prior work has explored emotion control in hybrid text-to-speech systems, the geometric properties of these modules, and their implications for steerability, remain poorly understood. We present the first comparative study of speech language model (SLM) and conditional flow-matching (CFM) modules as activation steering sites for mixed emotion speech synthesis. We first characterize emotion representations using linear probing and local intrinsic dimensionality (LID), and then evaluate single-site and joint steering for mixed-emotion synthesis. Our results show that SLM offers a clean, low-dimensional emotion-specific subspace with strong speaker--emotion disentanglement, while CFM exhibitspoor cross-speaker generalization due to speaker--emotion entanglement. Joint steering increases emotion intensity but degrades proportional control and speech quality on in-distribution data. These findings provide practical guidance for multi-site activation steering in hybrid TTS systems and highlight the importance of representation geometry in controllable speech generation.

Quantum computing is increasingly explored for software engineering (SE) optimization, but translating natural-language (NL) task-level requirements into executable quantum applications still demands substantial quantum and programming expertise. We present QPipe, a large language model (LLM)-based multi-agent architecture that autonomously turns NL requirements into traceable quantum-application workflows through specialized agents for requirement parsing, formulation, code generation, review, execution, and verification. We evaluate QPipe on 20 NL requirements, each associated with a real-world benchmark and a test-optimization problem. QPipe successfully completes the key stages of quantum-application generation across requirements, achieving average rates of 100% for code compilation and 96.7% for application execution and final-result combination, with average generation costs of 260.1 seconds and 1.89M tokens per requirement. Among the generated quantum applications that execute successfully, the returned solutions outperform the offline genetic algorithm baseline in most cases. Ablation results further show that QPipe's advantage depends on retaining code-generation skills, task knowledge, review feedback, and multi-agent decomposition. These results indicate that agentic coordination can support generation of executable quantum applications for tackling test optimization problems from real-world benchmarks.

Persona-driven generations (PDGs) have seen prolific use in research and industry applications, where a large language model (LLM) takes on a 'persona' while completing some task. While persona expressed through free-form text (like dialogue) has substantial work investigating stability or consistency, relatively, persona expressed in non-text-heavy outputs (like in multiple-choice question answering, or MCQA) is often overlooked. We work to address this gap, seeking to understand the instability of LLM PDGs in MCQA tasks. We develop three metrics investigating the performance, outcome, and question correctness stability, evaluating three distinct dimensions. Using these metrics, we find that instability varies consistently between model families and model size, and across question domains, with math/commonsense questions leading to greater instability. We also find task prompt format introduces more prediction instability than other hyperparameters, like temperature. Finally, we find that instability is related to task accuracy, and using our instability metrics, find different experimental settings that result in different best and worst personas for tasks, despite their similarity. This reveals the importance of checking hyperparameter instability in PDGs.

Predicting cancer drug response from transcriptomic profiles is a cornerstone of precision oncology, yet the scientific value of machine learning models hinges not solely on predictive accuracy, but also on their capacity to generate reliable biological insights. Current explainability approaches in this setting are computationally costly, lack robustness, and reduce complex drug response to univariate gene importance scores, overlooking the coordinated gene activity that drives sensitivity and resistance. In this work, we present ILLUME+, a scalable post-hoc explainability framework that moves beyond single-gene assessments to capture multiple, complementary forms of explanation. Integrated into our end-to-end pipeline, ILLUME+ produces more stable gene importance scores than existing baselines, recovers established drug-gene associations and mechanisms of action, and enables AI-assisted hypothesis generation to uncover novel interaction-driven molecular signals in cancer biology.

Delta debugging provides an automatic way to minimize a program input while preserving a certain property. However, its effectiveness fundamentally relies on the availability of test oracles to determine whether a reduced input still preserves the specific property. Consequently, the oracle problem substantially limits the applicability of existing delta debugging techniques, particularly for oracle-deficient programs where output correctness cannot be directly determined. To address this problem, this paper proposes a novel approach, DDMT, to enhance the applicability of delta debugging, especially facilitating its application to oracle-deficient programs. Our key insight is to redesign an oracle-independent test function and incorporate it into the reduction procedure of delta debugging such that the property-preservation validation can be accomplished without requiring a test oracle. To this end, DDMT employs the technique of metamorphic testing, which is a property-based and oracle-independent testing method. It establishes a metamorphic testing-based test function, using it as a replacement for the original test function adopted by delta debugging. The experiments evaluate DDMT on 66 subjects across both oracle-available and oracle-deficient scenarios, with different delta debugging approaches. The results positively confirm that DDMT can enhance the applicability of delta debugging while often preserving or improving reduction effectiveness and query efficiency. Furthermore, compared to the relevant delta debugging approaches, DDMT is also able to achieve performance improvements with proper configurations.

Diffusion Transformers (DiTs) have demonstrated impressive performance in image generation but suffer from substantial computational overhead and resource consumption. Post-training pruning offers a promising solution; however, due to DiTs' unique architectural design and parameter distribution, traditional pruning methods are inapplicable, leading to significant performance degradation. Specifically, prior methods developed for LLMs, which derive metrics through a series of approximations, amplify the relative contribution of weights in the saliency metric. In addition, weights in DiTs exhibit significantly larger magnitudes than those in LLMs. Moreover, existing pruning granularity overlooks variations in model structures. In this paper, we propose DiT-Pruning, which improves pruning performance by introducing customized saliency criteria and pruning granularity. We design a novel metric that balances the contributions of weights and activations from an energy-based perspective, enabling more effective identification of important elements. Furthermore, we observe distinct clustering patterns in the two-dimensional weight space. Accordingly, we adopt a clustering-aware pruning granularity, enabling effective sparse allocation. Extensive evaluations on various DiTs show that our method consistently preserves image quality, especially under high sparsity. For FLUX.1-dev at 512x512 resolution on MJHQ, DiT-Pruning achieves only a 0.001 loss in CLIP score at 50% sparsity, dramatically outperforming recent pruning methods.

Generalizing machine learning models to environments that differ from their training distribution remains a critical hurdle, particularly when data from the target domain is entirely or partially unavailable. We propose Generative Meta-Learning with Human Feedback (GMHF), a novel framework that bridges this domain gap by leveraging expert intuition to guide data synthesis. Grounded in a theoretical analysis of generalization error, we derive bounds demonstrating that aligning the distribution of generated data with human beliefs regarding the target physics significantly mitigates risk. GMHF operationalizes this insight by employing a Conditional Neural ODE (cNODE) as a generative digital twin, coupled with a Reinforcement Learning (RL) agent. The agent iteratively refines the latent physical parameters of the generated trajectories based on feedback, effectively steering the meta-learner toward the unobserved target distribution. Empirical validation on a nonlinear Duffing oscillator shows that GMHF substantially reduces deployment loss as expert reliability increases, and that the divergence between generated and target data falls under reliable feedback, directly corroborating the divergence-minimisation mechanism predicted by our theory. Further experiments on a non-dynamical probabilistic model confirm that the framework extends beyond ODE-governed systems, establishing human-AI collaboration as a rigorous catalyst for robust generalisation under distribution shift.

Accelerating materials discovery requires AI systems that can generate scientifically valid hypotheses through multi-step, domain-grounded reasoning. Standard large language models often produce fluent but weakly traceable responses to open-ended materials design problems, making it difficult to determine whether final answers are supported by coherent intermediate reasoning. We develop Graph-PRefLexOR, a family of graph-native reasoning models fine-tuned with Group Relative Policy Optimization (GRPO) to organize reasoning into explicit phases for mechanism exploration, graph construction, pattern extraction, and hypothesis synthesis. This design links neural language generation with symbolic relational structure, enabling causal connections to be constructed, inspected, and reused. On 100 open-ended questions from materials science and mechanics literature, Graph-PRefLexOR achieves 40-65% improvements over corresponding base models, with the largest gains in reasoning traceability. Embedding analyses show broader semantic exploration and approximately 2-3 times greater semantic diversity than baselines. Semantic backtracking and layer-wise hidden-state analyses further show stronger alignment between structured reasoning and final answers. Finally, test-time graph expansion reveals that additional compute primarily increases long-range conceptual recombination within a bounded semantic space, rather than simply expanding semantic coverage. These results establish graph-native reinforcement learning as a pathway toward interpretable AI systems for scientific hypothesis generation in materials design and other scientific applications.

Although large language models (LLMs) have demonstrated impressive creative fiction generation, they struggle to maintain narrative consistency and coherent plot lines in long-form stories. In this work, we introduce a unified framework for long-form narrative generation and verification. MAGNET, a multi-agent goal-driven narrative engine for storytelling, generates stories with persona-grounded character agents that propose actions based on a shared world state and evolving story goals, while ATLAS is a graph-based pipeline that compares scene-level world representations across a generated story to detect hallucinations. By evaluating MAGNET using an LLM editor, pairwise rubric scoring, and ATLAS, we show that our framework produces coherent narratives compared to single-model prompting and IBSEN. At 100 pages, MAGNET reduced annotations and hallucinations by 41 and 50%, respectively, compared to the single model baseline and by 34 and 45%, respectively, compared to IBSEN, with pairwise rubric evaluation showing similar results. These results suggest that long-form narratives can emerge from explicit world-state tracking and goal-driven multi-agent generation, providing a foundation for controllable and structurally coherent long-form narrative generation.

World models can enable Model Predictive Control (MPC), but this requires dynamics prediction that is both fast enough for online use and expressive enough to represent uncertain futures. Diffusion models offer a natural mechanism for modeling uncertain dynamics, yet their iterative inference procedure makes them difficult to use for low-latency latent planning. We bridge this gap with Value Diffusion World Models (Valdi), combining end-to-end online training for MPC with a latent diffusion dynamics model. In preliminary experiments on the CarRacing environment, we show that Valdi, using a single diffusion step at both training and inference, matches a deterministic MLP baseline. Our experiments expose a trade-off between predictive multimodality and control performance in this setup. Code is available at https://github.com/Kit115/ValueDiffusionWorldModels.

As exponential compute scaling continues, will the capabilities of frontier AI models outstrip what is accessible to developers on a small fixed budget? Or will capabilities converge, with "meek models inheriting the earth"? Building on Gundlach et al. (2025b), we show that the answer depends on how we value and measure AI capabilities. We discuss conventional performance measures and show that, while validation loss shows a shrinking gap, on other metrics frontier models grow their lead forever. Classifying performance metrics by their functional forms in relation to training (and inference) compute, we provide tight mathematical conditions for determining which metrics favor meek models, and show that bounded performance metrics always do. But careful interpretation of performance metrics is essential: we show that many common bounded metrics have closely-related counterpart metrics that are unbounded (and vice versa). Determining the apt metric in a domain is a prerequisite for policy, since bounded and unbounded metrics may suggest opposing policy responses. If a particular capability -- like software engineering, synthetic biology, or rhetorical persuasiveness -- is unbounded when measured in the terms we care about, frontier-level capability will likely be concentrated in the hands of a few wealthy actors. Conversely, if that capability is instead bounded, frontier-level capabilities proliferate through meek models into the hands of the many.

AI coding agents increasingly rely on skills: structured context bundles, typically a SKILL.md file with a YAML header and Markdown body, loaded on demand for domain knowledge, workflows, and scripts. Public registries such as skills.sh now host tens of thousands of skills, making them an emerging unit of reuse in agent-based software engineering. Yet skills have largely been viewed as agent capabilities rather than software artefacts whose content and evolution shape agent behaviour. We present the first empirical study of AI agent skills as engineered artefacts that are authored, reused, customised and maintained, across public registries and personal-use repositories. We mined 18,463 skills from skills.sh and 23,199 personal-use skills from 5,876 GitHub repositories, identifying 3,709 reuse links. LLM-based classification into SWEBOK knowledge areas (KAs) shows Software Construction dominates alongside a long tail of specialised areas. A thematic analysis of 180 skills identifies six content categories. Qualitative coding of 444 modifications reveals six themes, of which reworking operational specifications and adapting knowledge and resources are the primary target of change. Our findings show that reuse is largely a one-time copy operation: most reused skills remain near-verbatim, 53\% are never modified after adoption, and subsequent local maintenance is overwhelmingly additive. Customisation primarily adapts skills to local environments, whereas evolution accretes new inline domain knowledge. Across both, a stable behavioural contract -- how a skill interacts with users, monitors runtime state, and recovers from failures -- remains almost untouched. These results suggest maintenance effort should focus on project-specific bindings, and that registries and tool support should enable consolidating the domain knowledge skills re-author in isolation.

Generative Synthetic Populations (GSP) -- the convergence of population synthesis, agent-based modelling, and LLM agents -- are attracting growing interest for urban simulation and institutional communication research. Before any GSP instrument is used on a real population, a more basic question must be answered: does it respond to stimuli of known valence in an ordered, replicable, group-structured way? We call this controllability. We ask not whether a synthetic population tracks humans, but whether it tracks itself: whether the latent structure we impose on it is recovered in its own responses. This internal-validity question is logically prior to any claim about external validity, just as characterising an instrument's response function must precede using it to test a theory. We report SIVE (Synthetic Instrument Validation Experiment): a fictional municipality (Montelago) with 120 synthetic personas of known latent structure, exposed to seven conditions spanning strongly positive to strongly negative institutional communications about a water network. Seven pre-registered criteria, evaluated across a temperature sweep, jointly assess fidelity, stability, noise floor, specificity, sensitivity, and ordering. All seven pass at every temperature. A central finding turns a calibration failure into a diagnostic success: a message designed as "weakly positive" was identified by the instrument as functionally negative, traced to unresolved problems, uncertainty, and institutional passivity in its text; a redesigned version restored the expected ordering and interacts with agents' latent trust in unanticipated ways. A noise sub-experiment shows the instrument's intrinsic noise is roughly half the cross-agent estimate and stable across temperatures. Individual trajectories reveal coherent micro-dynamics that summary statistics obscure. Full data are available via an interactive explorer.

Mixed-precision quantization (MPQ) has become a key technique for deploying large language models under stringent memory and compute constraints. We first identify a phenomenon that we term the Perplexity Illusion: layers ranked as important by perplexity-based sensitivity show little rank correlation with those that are most influential for complex reasoning performance, with Kendall $τ\approx 0$ in our analysis. We further reveal an Alignment-Diversity Tradeoff: using only target-task calibration data can degrade post-quantization performance, whereas incorporating general-domain data stabilizes sensitivity estimation and improves robustness across tasks. Based on these observations, we propose TASA (Task-Aware Sensitivity Analysis), a two-level framework that jointly optimizes calibration-data composition and mixed-precision bit allocation. Specifically, TASA searches for a calibration-data mixture using a training-free gradient-trace alignment criterion, and then aggregates perplexity and reasoning-oriented sensitivity signals to guide both inter-layer and intra-layer bit allocation. Experiments on LLaMA-3-8B and Qwen2.5-7B reveal a precision inversion: appropriately allocated 3.5-bit models can match or surpass less task-aware 4-bit baselines. At an average precision of 3.5 bits, TASA matches or outperforms several competitive 4-bit uniform baselines in aggregate accuracy, and improves over the strongest W3 baseline on GSM8K by more than 20 absolute points on LLaMA-3-8B. These results show that calibration-data composition substantially affects task-sensitive quantization, a factor underexplored in prior work.

Hallucination detection for retrieval-augmented generation (RAG) is usually evaluated on natural-language document evidence. However, grounded generation systems increasingly rely on structured inputs: source code, developer-tool output, markdown documents, tables, and repository metadata. We introduce a unified benchmark for span-level hallucination detection over code, tool output, structured documents, and existing natural-language RAG datasets. The benchmark is built by starting from grounded correct answers, injecting localized hallucinations with exact character labels, and validating the code test split with evidence-based review. Our fine-tuned Qwen3.5-2B detector reaches 0.689 span-F1 on the unified test set and 0.60 on the code-agent source, where it substantially outperforms LettuceDetect-large (0.17) and the strongest zero-shot LLM judges we evaluated (at most 0.22). The same model remains competitive on established natural-language benchmarks, with 81.8 RAGTruth example-F1 and 0.724 English PsiloQA IoU.

Open web-scale pre-training corpora remain concentrated in English, limiting multilingual LLM development. We introduce MultiSynt/MT, an open synthetic parallel corpus with approximately 4.8 trillion target-language tokens across 36 European languages, produced by translating 100 billion high-quality Nemotron-CC tokens with Tower+ and OPUS-MT/HPLT-MT systems. For many medium- and lower-resource European languages, this is the largest openly available pre-training resource. On a broad multilingual benchmark suite, reference LLMs trained on MultiSynt/MT reach the final score of HPLT 2.0, a native-data baseline, using roughly 72% fewer pre-training tokens, and outperform it by approximately 15% relative at a matched 100B-token training budget. Our analyses also identify evaluation blind spots: standard multiple-choice benchmarks miss translation-quality differences that a fluency-sensitive LLM-as-judge evaluation cleanly recovers on the trained LLMs (with no fluency deficit in MultiSynt itself), and Norwegian idiomatic and culturally grounded tasks remain better served by native data. We release the corpus, including row-aligned translations from multiple systems, to support controlled research on multilingual pre-training data and evaluation.

We present DeWorldSG, a novel framework that generates spatio-temporally robust 3D Semantic Scene Graphs from RGB-D sequences. Existing methods often struggle to construct reliable 3D scene graphs due to unstable 3D object representations and missing relations caused by frame-wise inference. DeWorldSG addresses these issues by estimating instance-level geometric 3D Gaussian distributions through depth-guided filtering and representing each object as a probabilistic 3D node rather than a single projected point. To mitigate relational sparsity from frame-wise inference, our framework further aggregates spatiotemporal evidence across object pairs and refines relations using contextual priors derived from a world model (V-JEPA 2). Experiments on the 3DSSG and ReplicaSSG datasets demonstrate state-of-the-art (SoTA) performance in both object and predicate prediction, while producing temporally consistent scene structures. In particular, our method improves triplet recall by 77.4% and predicate recall by 23.2% over prior SoTA approaches, making it suitable for robotic manipulation and AR applications. Our code and models are open-sourced.

Recent advances in neural rendering have established 3D Gaussian Splatting (3DGS) as a highly efficient representation for novel view synthesis, enabling fast training and real-time rendering with strong fidelity. However, when supervision is limited to sparse input views, 3DGS tends to overfit to the observed images and generalize poorly to unseen viewpoints. We address this challenge from the perspective of flat minima (FM) optimization, which seeks solutions that remain stable under small parameter perturbations. Viewing Gaussian parameters as trainable weights, we adapt FM principles to the geometric and dynamic nature of 3DGS with a lightweight training framework. Our method regularizes optimization with controlled Gaussian perturbations that account for each Gaussian's anisotropy and the training progress, preserving fine details while improving robustness to sparse-view overfitting. To further stabilize this flat minima optimization process, we introduce periodic reinitialization, which temporarily returns non-positional parameters to their initial states for a short window. Together, these techniques integrate seamlessly into existing 3DGS pipelines without architectural changes. Experiments on LLFF and Mip-NeRF360 datasets demonstrate improved quantitative metrics and perceptual quality under sparse-view supervision, producing reconstructions that are sharper, more stable, and better generalized to novel viewpoints.

Private continual counting is a fundamental problem in differential privacy: given a binary stream of length $n$, where each $1$ corresponds to the contribution of one individual, the goal is to release all running counts while protecting the privacy of each individual. The standard algorithm is the binary tree mechanism, whose Gaussian-noise variant achieves expected $\ell_\infty$ error proportional to $\log^{3/2} n$ for approximate differential privacy. Whether this dependence on the stream length is necessary has remained a central open problem. In this work, we resolve the dependence on $n$ by proving that every differentially private mechanism for continual counting must incur expected $\ell_\infty$ error $Ω(\log^{3/2} n)$. This shows that the binary tree mechanism is asymptotically optimal in the approximate-DP setting. As a consequence, we also obtain a largest-possible separation between hereditary discrepancy and private $\ell_\infty$ error for linear queries, showing that the known general upper bound in terms of hereditary discrepancy has the optimal dependence on the number of queries.

Rhetorical strategies and their influence on audiences are often studied through social media posts and comments. However, this focus overlooks the universal audience, which is the majority of readers who remain silent and do not explicitly express how a message affects them. This study investigates how two classical modes of persuasion, ethos and pathos, resonate in the silent audience's interpretations of meaning. Using a dataset of social media sentences paired with human-written interpretations, we label both sources for ethos and pathos and assess whether these rhetorical appeals are preserved. Our analyses show that interpretations diverge from the original sentences in 30% of cases, with rhetorically charged content eliciting greater variability than neutral content. We further find that ethos and pathos in original sentences can predict audience attitudes toward the author, underscoring the subtle ways rhetoric shapes perception beyond visible engagement.

Self-evolving agents violate the assumption behind most learning-theoretic guarantees: the data, evaluator, components, and hypothesis space are produced by the policy being updated. We present \textbf{SEA}, an architecture that confines self-modification to a small steering adapter and a versioned harness around a \emph{frozen} base model and admits each modification only through an anytime-valid gate that emits an auditable certificate against a fixed error budget. Five loop controllers compose published guarantees; because such gates can only \emph{select} among behaviors the frozen base already produces, five verifier-in-the-loop mechanisms -- best-of-$N$, micro-step search, self-authored reproduction oracles, search-layer control, and self-repair -- supply the dense, grader-free signal the gates require, computed from the issue text alone. On a $52$-instance SWE-bench Verified subset across four base models, base capability is the dominant, confound-free effect, and on two strong base models a deliberate no-op-composite control isolates the suite's contribution at $+4$ and $+5$ (\textsc{Glm}~5.2 $24\to28$; \textsc{Gpt} $29\to34$, the $65\%$ best), with event logs confirming that its mechanisms fire and prevent regressions. Results are single-run on expensive evaluations; confirming run-to-run variance and adapting the per-task algorithm mix are future work.

We study inference-time pattern-memory gating in a production-scale clinical natural language processing (NLP) pipeline. The pipeline pairs a generator (Llama-3.3 70B) proposing extractions with a verifier (MMed-Llama-3.1 70B) accepting or rejecting them, over 167,034 PMC-Patients narratives, and adds a lightweight memory that learns at deployment which extractions to filter, so the verifier need not re-examine candidates already seen to fail. We report four findings. First, learning filtering rules directly from the verifier's rejections failed at full scale: the relation-extraction filter stayed empty despite 785,797 logged rejections, because they were spread too thinly across too many distinct forms to accumulate. Second, a simpler rule using a fixed clinical ontology produced the same filtering without the verifier, capturing 49,734 ontology-violating relations on a held-out 5,000-patient set. Third, of five versions of the question-answering filter, four failed for distinct, instructive reasons; the fifth succeeded by checking whether a patient's extracted entities support the question asked, and where it applies was 1.84 times likelier to flag an answer the verifier would reject than one it would accept. Fourth, one pattern held across all five: a filter is selective only when it tests the same evidence the verifier weighs, not when it imitates the verifier's output. Together these give a transferable result for any generator-verifier pipeline: the most natural memory design can fail silently at scale, and whether a pre-generation gate is selective is decided before any engineering effort, by whether its signal probes the question the verifier itself answers. Throughout, the system flags suspect extractions rather than deleting them, so every decision stays visible for clinical review. All code and test artefacts are released openly.

Bayesian Optimisation (BO) under unknown constraints is particularly challenging when feasible regions are small. In such settings, existing methods that typically rely solely on evaluations of the true objective and constraints struggle to efficiently explore the design space. However, many real-world applications offer auxiliary data sources (e.g. surrogate models or simplified simulations) that can support early exploration. Despite this potential, their integration into constrained BO remains largely unexplored. We propose a general multi-source framework that extends constrained Max-value Entropy Search, capturing inter-source correlation while balancing evaluation cost and information gain. Experiments on both synthetic and physics-based benchmarks show that our method efficiently identifies feasible and optimal solutions, even when auxiliary data are only weakly correlated. The proposed approach consistently outperforms existing methods, particularly in early-stage exploration.

Large Reasoning Models (LRMs) have achieved remarkable success on complex tasks by leveraging long chain-of-thought (CoT) trajectories, yet they frequently exhibit overthinking on simple queries, resulting in significant token overhead and reduced inference efficiency. However, existing compression methods predominantly apply uniform length reduction or rely on coarse-grained difficulty estimation, often leading to performance degradation on difficult problems. To address this limitation, we propose Confidence-Adaptive Thinking (CAT), a framework that incorporates the model's intrinsic self-certainty signals as confidence into the preference optimization process, which autonomously modulates reasoning lengths based on problem difficulty. Experimental results show that CAT consistently outperforms state-of-the-art baselines on reasoning accuracy across multiple benchmarks on different base models. Our work enables LRMs to effectively compress confident responses while deliberating on uncertain ones, offering a potentially robust solution for balancing accuracy and latency in practical industrial scenarios.

Millimeter-wave (mmWave) beam alignment plays a critical role in next-generation wireless systems, yet its efficient implementation remains challenging. Meta-learning and transfer learning have been explored to enable deep learning-based beam prediction models to rapidly adapt to unseen environments; however, existing meta-learning approaches adapt the entire network and are trained from random initialization, leading to a large number of updated parameters and a high meta-training cost, while transfer learning approaches restrict adaptation to part of the network but do not exploit episodic meta-learning, which explicitly trains the model over multiple tasks, to optimize the adaptation process itself. To overcome these limitations, we propose MTL-BA, a meta-transfer learning framework for beam alignment in millimeter-wave multiple-input single-output (MISO) systems that freezes a pre-trained convolutional backbone and meta-learns only lightweight Scale-and-Shift (SS) adapters together with a classifier head. Warm-starting from the pre-trained model and restricting adaptation to the SS adapters and classifier head reduce both the adaptation cost and the meta-training budget without sacrificing prediction performance. Simulation results on the DeepMIMO ray-tracing dataset show that MTL-BA matches the accuracy and spectral efficiency of full fine-tuning across various SNR levels despite updating approximately $17\times$ fewer parameters than both full fine-tuning and Model-Agnostic Meta-Learning (MAML), outperforms last-layer fine-tuning while updating a comparable number of parameters, and approaches MAML's performance while requiring $60\%$ fewer meta-training epochs.

Contrastive pre-training has propelled video-text alignment, yet models often inherit the critical limitations of their image-text predecessors like CLIP, resulting in entangled representations. These challenges are severely exacerbated by two fundamental properties in the video domain: Temporal Misalignment, where textual descriptions often correlate only to specific, constrained temporal windows, leaving other frames text-irrelevant; and Semantic Asymmetry, which dictates a sparse, bidirectional, and non-equivalent relevance between frame-level visual details and caption-level concepts. This failure persists whether captions are short and temporally disjoint, creating ambiguity, or long and detailed, fostering entanglement between static objects and their temporal evolution. In this paper, we establish theoretical conditions that enable flexible alignment between video and text representations across the temporal dimension and at varying levels of granularity. Building on these theoretical insights, we introduce MoVA, Modular Long Video-Text Alignment, which learns dual asymmetric projections: a text-side projection that adaptively selects frame-aware subspaces of the caption, and a video-side projection that disentangles text-relevant visual concepts. Our framework ensures that the model can preserve global cross-modal semantics while disentangling evolving, frame-specific concepts and scale naturally to long captions and videos. Empirical evaluations show that MoVA outperforms existing methods in multiple video-text alignment tasks, demonstrating the effectiveness of our method.

In recent years, large language models have achieved remarkable success and have seen growing adoption in financial applications. At the same time, explainability remains critical in finance, a domain characterized by high stakes and strict regulatory requirements. Although numerous methods have been proposed to explain black box machine learning models, the majority of these approaches are designed for general purpose tasks and do not incorporate domain specific knowledge. In this work, we study the explainability of financial textual data modeled by large language models through the lens of the Shapley value. Specifically, we investigate whether Shapley based attributions align with established financial domain knowledge. Through rigorous theoretical analysis and extensive empirical evaluations, we demonstrate that Shapley values can yield explanations that are consistent with financial reasoning and can offer meaningful insights into the model's behavior in text based financial applications.

This work studies the hidden-state inversion problem: recovering the original input token sequence of a decoder-only language model from its last-layer hidden states. Rather than treating inversion as a one-shot reconstruction, we study it as a continuous embedding-space optimisation in which a soft proxy is driven towards the leaked target without any hard-token projection during the search, and a token is committed only once, at the end of the inner loop. This design choice has two consequences which are the main focus of this paper. First, keeping the optimisation entirely in continuous space exposes a rich set of internal signals: rank trajectories of the ground-truth token, per-position loss curves, and a discrete loss measured at commit time. Second, the discrete loss allows assessing the correctness of recovery via cumulative discrete loss. We further analyse which tokens break the reconstructions and find a sharp categorical asymmetry: space-prefixed, high-frequency function words in dense regions of the embedding matrix dominate the failures, while content-bearing tokens are recovered almost perfectly. On 10-token C4 prompts the exact-match rate rises from 66.9% to 97.5% (mean similarity 0.994) as the candidate window is widened, confirming that most errors are recoverable near-misses rather than genuine ambiguities. A comparison with the released SIPIT reference situates these findings: per-step hard projection is faster, but the continuous formulation is what makes the optimisation observable and its failures detectable. The results show that last-layer hidden states of GPT-2 are as sensitive as the original text.

Most self-supervised learning (SSL) methods encourage invariance across augmentations, but strict flip invariance can suppress informative left--right correspondences in approximately bilateral data such as medical images and human faces. We propose Mirror-Fusion-Augmented Self-Supervised Learning (MFASSL), a Vision Transformer framework that injects a soft reflection prior into standard SSL without redesigning the backbone. MFASSL constructs mirror-paired views aligned to an estimated symmetry axis and introduces a lightweight Mirror-Fusion Attention (MFA) module for adaptive token-level interaction between mirrored regions while preserving asymmetric cues. The base SSL objective is further coupled with reflection-consistency and mid-layer token-alignment losses. Across CheXpert, BraTS, CelebA-HQ, and WFLW, MFASSL improves downstream performance, calibration, and reflection robustness over MoCo-v3, DINO, and MAE baselines under matched ViT-B/16 settings. It also achieves stronger and more consistent gains than recent equivariant SSL approaches with only approximately 2.7\% additional parameters. These results show that lightweight geometry-aware priors can effectively complement invariance-based SSL.

News articles are an important source of information on disaster impacts and adaptation. A key methodological challenge in socio-environmental studies is how to select a representative data sample. Two approaches are common: querying news databases top-down with the aid of an existing disaster inventory or using NLP methods to cluster news texts bottom-up based on temporal and spatial features. Using a dataset of German news about landslides worldwide, we compare these approaches and discuss variations in event coverage. Such research design decision can influence the resulting news sample, affecting its use in studies of inequality in media coverage, disaster monitoring and inventory enrichment.

In this opinion paper, we propose MetaHOPE, an error severity-aware annotation framework for evaluating metaphor translations. Metaphors present challenges for machine translation (MT) and natural language understanding and processing (NLU, NLP), because it presents the features of semantic complexity, contextual dependency, and cultural embeddings that can lead to ambiguity issues for NLP models. To investigate how state-of-the-art NLP models perform on translating metaphors, we select three representative systems, i.e., GoogleMT, GPT5.4, and Hunyuan-7b as Neural MT (NMT) models and LLMs. We used two human-annotated metaphor corpora, including VUAMC and PSUCMC for English-to-Chinese and Chinese-to-English translation purposes. The original corpora we used are monolingual, where we carried out error annotation using the MetaHOPE framework, and also produced the human post-edited gold reference for bilingual use as a new resource. We believe the MetaHOPE evaluation framework for metaphor translation annotation, the parallel corpora resources, and the error analysis on SOTA automatic translation models can be useful and shed some light for the field of metaphor translation study. We share our resources publicly upon paper acceptance.

This paper studies whether the Metabolic Multi-Agent Optimizer (MMAO) can be credibly derived into a dynamic-optimization method without replacing its core metabolic control loop by external adaptation modules. The proposed MMAO-Dyn maps private energy, communal budget, role drift, success feedback, and lifecycle turnover to a nonstationary setting in which environmental changes repeatedly invalidate previously useful local structure. We evaluate MMAO-Dyn on an 18-scenario synthetic dynamic continuous benchmark matrix covering shifted sphere, shifted Ackley, and shifted Rastrigin landscapes at $10D$, $20D$, and $30D$, with two change severities and 12 seeds per scenario. The comparison layer includes a generic MMAO variant without dynamic derivation, dynamic random search, dynamic PSO-lite, dynamic DE-lite, and three endogenous ablations. Across the full 216-run matrix, MMAO-Dyn attains mean offline error $28.07$, improving over Generic-MMAO ($29.36$), Dynamic-PSO-lite ($34.65$), Dynamic-DE-lite ($67.09$), and Dynamic-RandomSearch ($111.37$). The gains are clearest in aggregate robustness on sphere and Rastrigin families and in 10-step post-change recovery relative to the generic backbone, whereas the seed-aligned comparison with Dynamic-PSO-lite remains unfavorable in win-loss count and the \texttt{NoMemoryRefresh} ablation stays very close to the full method. We therefore position MMAO-Dyn as a credible family-expansion result for MMAO: the metabolic loop can generate meaningful dynamic behavior, but the strongest current value lies in recovery-oriented resource redistribution rather than in universal dominance or in a fully optimized submechanism design.

World models are increasingly used in embodied intelligence and generative simulation, yet their scope remains ambiguous across communities. This tutorial presents a design-space view of world models as action-conditioned predictive models that estimate the future evolution of task-relevant observations or states. We categorize existing methods into observation-space and state-space world models, comparing their trade-offs in visual fidelity, spatial structure, physical interpretability, and control usability. We further introduce world action models, which connect predicted futures with executable robot actions, and summarize four representative paradigms: imagine-then-execute, video-feature-conditioned action prediction, joint video-action modeling, and auxiliary video prediction for policy learning. The goal of this tutorial is to clarify the conceptual scope of world (action) models and provide a structured taxonomy for embodied prediction and control.

Near-Infrared (NIR) spectroscopy has emerged as a promising alternative to traditional soil analysis methods, offering advantages such as speed, low cost, and non-destructive testing. This work proposes a machine learning (ML) approach to calibrate predictive models for carbon (C) and nitrogen (N) content in Oxisols and Inceptisols, utilizing NIR spectral data acquired with a portable MyNIR device. Various preprocessing methods were evaluated, with the most effective being the Savitzky-Golay (SG) filter and a robust outlier removal method based on the Nonlinear Iterative Partial Least Squares (NIPALS) algorithm coupled with a Huber loss function. Multiple validation strategies were compared, including 10-fold cross-validation, leave-one-out, and holdout via the Kennard-Stone method, followed by standardization. Stacking ensemble learning models were employed, using Partial Least Squares (PLS), Support Vector Regression (SVR), and Ridge as base models, with linear regression as the meta-model. The models were evaluated using R2, Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Ratio of Performance Deviation (RPD) metrics. The performance gap between soil types suggests the influence of pedological characteristics. Furthermore, the models achieved an RPD > 2.0 with low overfitting, validating the potential of this approach for rapid C and N quantification. This study contributes to the optimization of sustainable agricultural practices, aligning with the demand for efficient and environmentally friendly analytical methods. The developed technique enables faster decision-making for producers and consultants based on organic matter content, fertility indicators, and nutrient availability.

A single panorama captures the full visual sphere from one camera center, yet confines users to looking around in place without enabling true scene exploration. Converting a single panorama into a persistent, renderable 3D representation for free-viewpoint navigation has attracted growing interest; existing methods either adopt iterative per-view completion that propagates inpainting results to update the underlying geometry, leading to progressive error accumulation and cumbersome multi-step pipelines, or leverage the temporal consistency priors of video generation models, yet the continuous-trajectory constraint intrinsic to such models limits their flexibility in covering scenes from multiple directions simultaneously. We present Pano2World, which takes a single indoor panorama as input and directly outputs a persistent, explorable 3D Gaussian scene. Given the source panorama, Pano2World first reconstructs a coarse 3D Gaussian proxy and renders it at adaptively sampled nearby poses to obtain geometrically aligned guidance panoramas; a panoramic diffusion model then jointly denoises all target views via View-Aware Attention Routing, where each target view simultaneously receives geometric constraints from its corresponding guidance panorama and global semantic guidance from the source panorama, naturally enforcing cross-view consistency. To avoid the information loss incurred by decoding the multi-view hidden features formed during joint denoising back to the pixel domain via VAE, we introduce Latent Feature Adapter, a geometry-aware bridge module that directly distills these hidden features into a scene latent, subsequently decoded into the final 3D Gaussian scene. Experiments demonstrate that Pano2World significantly outperforms existing methods on the multi-position panoramic novel-view synthesis benchmark.

The landscape of automated formal verification is populated by techniques that make prominently different trade-offs: some focus on expressiveness and precision, supporting the verification of complex properties; others favor scalability and practicality, so that they are applicable to larger programs using different features. This paper presents Petrify, a novel automated verification technique for concurrency properties that achieves a distinctive trade-off. Petrify encodes the semantics of Java bytecode programs into Petri nets (PNs), which can be analyzed by state-of-the-art model checking tools such as LoLA. As our experiments demonstrate, Petrify's approach offers an interesting combination of expressiveness and practicality: PNs are a fairly precise encoding of the concurrent behavior of programs; at the same time, Petrify's PN encoding is succinct, so that its analysis remains quite insensitive to parameter size. Another practical benefit of targeting bytecode is that jPetrify, the prototype tool that implements the Petrify technique, is applicable to programs written in any version of Java and even a subset of Kotlin (another language that compiles to Java bytecode) while other similar tools are limited to older versions of Java. While this paper's experiments focus on analyzing fundamental properties like deadlock, Petrify's approach lends itself to be extended to other kinds of concurrency analysis, which we plan to tackle in future work.

Large language models (LLMs) are increasingly used to generate queries, invoke tools, and construct analytical workflows. Although recent advances have substantially improved workflow generation and execution, the semantic information required to operationalize analytical concepts often lies beyond what is explicitly represented in database schemas and data values. We present a cross-domain formative study of operationalization failures in agent-generated analytical workflows. Across 236 analytical intents spanning finance, human resources, and public safety domains, we identify 153 recurring failures despite successful workflow generation and execution. Our analysis reveals five recurring classes of failures: comparative grounding, process reasoning, quantitative reasoning, role confusion, and policy grounding. These findings suggest a semantic gap between user-level analytical concepts and the information available to workflow-generation systems. More broadly, they raise questions about the admissibility of analytical operations and suggest that future agentic data systems may require richer semantic representations to bridge the gap between analytical intent and executable computation.

Standard gate-level transpilation introduces significant physical noise and overhead for high-precision quantum algorithms, such as the Quantum Singular Value Transformation (QSVT), on near-term trapped-ion hardware. Current compilers treat quantum operations as discrete units, forcing the physical control layer to execute highly fragmented laser pulses. To address this hardware-software disconnect, this work introduces a holistic pulse synthesis strategy that bypasses discrete gate-stitching to compile algorithms directly into continuous compound pulse gadgets. As a proof-of-concept, we target Hamiltonian simulation of the $H_2$ molecule, block-encoding the problem into a QSVT circuit to approximate the time-evolution operator $U = e^{-i H t}$ across 3 computational ions (2 system, 1 ancilla). We utilize the Gradient Ascent Pulse Engineering (GRAPE) algorithm to generate these compound gadgets and evaluate our methodology using noisy Lindblad master equation simulations. Preliminary observations indicate that the proposed strategy achieves significant temporal compression, reducing the total pulse schedule duration compared to standard compilers. Furthermore, synthesizing operations holistically eliminates the control-layer latency associated with discrete pulse lookup overhead. By streamlining the physical control schedule, this methodology offers a promising pathway to execute operations faster, highlighting the potential for compound gadgets to increase the computational depth achievable within fundamental $T_2$ decoherence limits.

Frontier foundation models have changed the math on vulnerability discovery, but the bigger challenge is how the remediation side keeps up. Despite recent progresses in Automated Vulnerability Repair (AVR), current solutions struggle to reliably identify the root causes of vulnerabilities, and insufficiently utilize the prior fix knowledge to guide the patch generation process, thus undermining their effectiveness in practice. To address this gap, we propose KeaRepair, a novel agentic AVR approach that grounds patch generation in verified program facts and high-level vulnerability knowledge. Specifically, KeaRepair first extracts multi-dimensional vulnerability knowledge from historical vulnerability-patch pairs from dual complementary views, and constructs dedicated retrieval knowledge bases. It then employs a tool-augmented agent that performs ReAct-style reasoning to collect verified program facts for vulnerability diagnosis. Finally, based on the diagnostic results, KeaRepair performs knowledge-level retrieval-augmented patch generation and iteratively refines patches through a closed-loop validation process involving compilation, PoC replay, and test-suite execution. Experimental results show that KeaRepair significantly outperforms existing AVR approaches on 55 reproducible C/C++ vulnerabilities. When paired with Gemini-3.1-Pro, KeaRepair successfully repairs 46 vulnerabilities, achieving a repair rate of 83.64%. Moreover, KeaRepair fixes six unique vulnerabilities that none of the baselines can address, and further demonstrates strong cross-language generalizability.

Energy consumption is a key limitation in high-performance computing on heterogeneous CPU-GPU systems. This work studies how hardware configuration affects energy-to-solution under realistic workloads. We study energy efficiency regimes using molecular dynamics benchmarks (GROMACS and AMBER) and a stress-test benchmark (FIRESTARTER) on systems with A40, A100, H100, and H200 GPUs and Intel Ice Lake CPU, varying frequency scaling and power cap. We show that energy-to-solution exhibits workload- and architecture-dependent transitions between efficient and inefficient regimes, driven by nonlinear GPU power-frequency scaling. We introduce an interpretable analytical model that decomposes GPU power into linear and nonlinear components, identifying a workload- and architecture-dependent transition frequency beyond which efficiency degrades. The model fits empirical data with low error and highlights the role of baseline power, nonlinear power behavior, and transition frequency as the dominant parameters governing energy efficiency. Power capping is generally less effective for efficiency tuning than frequency reduction, especially for workloads that operate far from thermal design power. Overall, energy-efficient HPC execution is a configuration-dependent problem with identifiable regime shifts, and we provide model-driven guidance for selecting operating points.

SAT solvers settle combinatorial problems beyond the reach of interactive theorem provers and produce LRAT certificates for independent verification. We present LRAT-Catcher, a standalone, general-purpose tool that imports a DIMACS formula together with an LRAT certificate into Lean 4 as a theorem. LRAT-Catcher runs the formally verified LRAT checker from Lean core as compiled native code via reflection. This scales to instances where Mathlib's explicit proof-term import exhausts memory. LRAT-Catcher also composes cube-and-conquer solving runs entirely inside Lean. Per-cube refutations are combined with a cover-completeness certificate, itself an LRAT proof, into a single unsatisfiability theorem. Verified encodings connect CNF-level results to the original combinatorial problems. We evaluate the tool against Mathlib's proof-term import and the external checker cake_lpr on establishing the Schur number S(4) = 44 and the Ramsey number R(4,4) = 18 as Lean theorems.

Unsupervised pre-training on large-scale datasets has demonstrated significant potential for improving the sample efficiency and performance of Reinforcement Learning (RL). Given the large-scale action-free internet videos, existing methods utilize single-step transition prediction and image reconstruction to learn representations. However, these methods prefer to preserve large-proportion stationary information in the pixel space, neglecting small but crucial information. To preserve enough information in the representation, it is essential to pay equal attention to each element in videos. Specifically, we propose a temporal correlation space to distinguish each element. For implementation, we introduce the Multi-scale Temporal Contrastive Learning (MTCL) method to model multi-scale temporal correlations separately. This approach can balance the attention of different elements and yield more informative representations, effectively supporting policy learning in various downstream tasks. Experimental results demonstrate that our method improves sample efficiency and asymptotic performance across various downstream tasks.

Pre-training on large-scale videos to improve reinforcement learning efficiency is promising yet remains challenging. Existing methods typically treat the agent as an indivisible entity, modeling motion patterns globally. Such global modeling is tightly coupled with the morphology, hindering transfer across domains. In contrast, despite the vast disparity in global motions, the local components exhibit similar motion patterns across different agents. Building on this insight, we propose a novel Deconstruct-Recompose Paradigm (DRP) for learning transferable local motion representations. Specifically, in the Deconstruct phase, we identify multiple local points and track their frame-wise motions, defining each as an Atomic Action. We introduce a Dual-Attention Encoder (DAE) to learn local motion representations from these Atomic Actions, capturing their spatiotemporal relationships. In the Recompose phase, we compose local motion representations with a learnable Motion Aggregation Token [MAT] via latent dynamics model learning. Additionally, an adapter bridges local motion and downstream action-specific dynamics to accelerate policy learning. Extensive experiments demonstrate that our method effectively transfers to diverse robotic control and manipulation tasks, significantly improving sample efficiency and performance.

Visual Reinforcement Learning (VRL) has achieved considerable success in solving control tasks. However, generalizing learned policies to new environments remains a major challenge, as agents often overfit to task-irrelevant features in the training environment. To solve this problem, we introduce the concept of decoupling observations into task-relevant and task-irrelevant representations. Building on this idea, we propose a self-supervised Task-Relevant Representation Decoupling (T2RD) algorithm for VRL. This algorithm consists of three components: task-relevant representation consistency, cross-reconstruction, and cross-dynamic prediction. The first two components achieve the decoupling of content and style features, but the resulting content representations are not necessarily task-relevant. To further refine task-relevant features from content representations, we design the third component that introduces dynamic prediction. T2RD achieves State-Of-The-Art (SOTA) generalization performance and sample efficiency in the DeepMind Control Suite and Robotic Manipulation tasks.

For predictive models, the often-reported performance metrics are the loss and accuracy. In synchronous Brain- Computer Interface (BCI) systems, these metrics are informative for most BCI paradigms; however, for Event-Related Potential (ERP) applications the spelling rate, which measures the number of characters correctly selected is more important as it influences the estimation of information transfer rate (ITR) and any related metric measuring spelling performance. Moreover, ERP-based BCIs hold imbalanced data class distributions, which require reporting metrics that can handle the imbalance, such as the area under the receiver operating characteristic curve (ROC AUC). In this work, we study the correlation of the spelling rate with 13 metrics to identify which among them best reflect user spelling performance and how they are affected by trial repetition. The Results of two datasets (a private LARESI ERP dataset and the public OpenBMI ERP dataset) favor the Brier score, Matthews Correlation Coefficient (MCC), and the metrics that account for class imbalance in binary classification: ROC AUC, area under the Precision-Recall curve (PR AUC), Average Precision (AP), and partial AUC (pAUC). These findings encourage researchers and practitioners to report those metrics in ERP-based BCI experiments.

Vision-language pretraining remains dominated by contrastive objectives, whereas vision-only self-supervised learning has largely adopted non-contrastive methods. At the same time, the role of vision-language encoders has shifted: they are increasingly deployed not as zero-shot classifiers but as the frozen visual backbone of vision-language models and dense prediction systems, which consume the full grid of patch tokens rather than a single pooled embedding. We introduce LeVLJEPA, the first fully non-contrastive end-to-end vision-language pretraining method. LeVLJEPA learns through cross-modal prediction with stop-gradient targets and per-modality distributional regularization, without negatives, temperature, momentum encoder, or teacher-student schedule, and trains stably at large scale. We find that the resulting encoder provides markedly stronger dense semantic features for downstream use: as a frozen vision-language-model backbone, LeVLJEPA is the strongest of the evaluated encoders across GQA, VQAv2, and POPE under two distinct language models, and outperforms contrastive baselines on semantic segmentation, while remaining on par on global readouts such as linear probing. These results establish non-contrastive pretraining as an effective means of producing dense semantic vision features.

Recognizing jazz standards from audio is a challenging form of tune-level music retrieval: different performances of the same standard may vary in tempo, key, arrangement, instrumentation, improvisational content, and even whether the head melody is present. We study this problem using a curated subset of the Jazz Trio Database designed for cross-performance standard recognition. We compare a from-scratch trained Harmonic CNN baseline against frozen pretrained music representations from recent music understanding foundation models, using both supervised probing and nearest-neighbor retrieval. Our results suggest that from-scratch spectrogram models overfit strongly to training performances, while pretrained embeddings provide better top-$k$ results but are sensitive to performer identity, which can be partially reduced with a lightweight contrastive projection. Our findings motivate jazz standard recognition as a useful stress test for music representation models and as a step toward retrieval-based standard identification. Project page: https://github.com/cagries/tipofmyear.

Recent recursive Transformer studies have primarily reused shared parameters across computation steps to construct compact, parameter-efficient models. In this work, we leverage recursion to build effectively deeper Transformers with stronger representational capacity. However, in Vision Transformers, simply increasing recursion depth does not reliably improve performance, as existing recursive approaches do not fully utilize the intermediate representations produced throughout recursive computation. We propose Soft Mixture-of-Recursions (SoftMoR) and its Vision Transformer instantiation, Soft Recursive Vision Transformer (SR-ViT). SoftMoR learns token-wise mixture weights to softly combine outputs from all recursion steps, allowing intermediate representations to be utilized in a learnable and flexible way. Across diverse vision tasks, SR-ViT consistently improves as recursion depth increases with minimal parameter overhead. On ImageNet-1K, increasing recursion depth from 1 to 4 improves SR-ViT-S top-1 accuracy from 79.83% to 82.48% with only 1.7M additional parameters, outperforming the substantially larger DeiT-B while using approximately 27% of its parameters. These results demonstrate that SoftMoR provides a parameter-efficient path to deeper and stronger Vision Transformers through recursion.

Discrete diffusion models are widely used for learning and generating discrete distributions. As the generation process is inherently sequential, the acceleration of sampling is of significant importance. In this work, we parallelize the mainstream $τ$-leaping algorithm for absorbing discrete diffusion in a Continuous-Time Markov Chain (CTMC) framework. By leveraging the continuous-time stochastic integral form of the $τ$-leaping algorithm and the Picard iteration method, we achieve parallel-in-time sampling acceleration and provide a proof of exponential-factorial convergence for our algorithm. We improve the overall time complexity of $τ$-leaping under absorbing settings from ${\mathcal{O}}(d \log S)$ to ${\mathcal{O}}(\log (d\log S)\cdot \log d)$ with respect to NFE. Empirically, our method shows consistent acceleration across synthetic and real-data settings. The new sampler achieves at most $7$--$9\times$ runtime speedup for synthetic distribution, and maintains the same quality with $50\%$ fewer NFE and $1.45$--$1.86\times$ runtime speedups in image/text tasks on a single GPU. Our research expands the potential of discrete diffusion models for efficient parallel inference, with broader implications for applications such as molecular structure and language generation.

Digital forensic investigations of network intrusions require analytical outputs that are traceable, reproducible, and court-defensible - requirements existing machine learning pipelines do not satisfy, since they treat original evidence as training data and produce opaque classifications without instance-level justification. This paper presents a forensic-oriented intrusion detection framework resolving both problems simultaneously, integrating synthetic data generation, binary classification, and explainability within a single pipeline governed by ISO/IEC 27037, 27041, 27042, and NIST SP 800-86. The framework operationalises the ISO/IEC 27037 requirement for strict separation between original digital evidence and derived analytical artefacts. Original datasets are treated as immutable, hash-verified artefacts; all training operates on parameterized synthetic derivatives via SDV + CTGAN. XGBoost binary classification provides high-performance detection on tabular network flow data, and SHAP TreeExplainer produces instance-level feature attributions mapping statistical predictions to observable network behaviour for forensic reporting. Train-on-Synthetic, Test-on-Real (TSTR) evaluation on CICIDS2017 achieves F1-macro = 0.96, within cross-validation variance of the real-data baseline (0.97). Kolmogorov-Smirnov testing confirms synthetic privacy preservation (mean |KS| = 0.38) alongside operational utility. Cross-dataset validation on UNSW-NB15 and Kitsune identifies feature space dimensionality as the primary determinant of synthetic training effectiveness, establishing a practical deployment boundary of approximately 30 numeric flow-level features. SHAP attributions for Brute Force, Port Scan, and DoS attacks are consistent across real and synthetic instances, confirming synthetic training preserves forensically relevant attack fingerprints required for expert witness testimony.

Evolutionary multitasking (EMT) has shown strong capability in solving multiple optimization problems simultaneously by exploiting latent inter-task consistency, such as similarities in promising solutions or search directions. However, most existing EMT studies remain focused on objective-driven optimization, where such consistency is mainly used to accelerate convergence toward predefined optima. In this paper, we move EMT from consistency to collaborative discovery and propose a multifactorial evolutionary algorithm with collaborative discovery (MFEA-CoD) for multitask novelty search. Unlike conventional EMT, MFEA-CoD coordinates multiple novelty search tasks to collaboratively discover behaviorally novel solutions rather than merely transferring consistent search information for faster convergence. Specifically, a multitask repulsion operator encourages different tasks to explore distinct regions of the unified search space, thereby reducing redundant behavioral discoveries. Meanwhile, an adaptive inter-task transfer mechanism exploits shared discovery opportunities in overlapping novelty-improving regions by adjusting the transfer probability according to the online contribution of transferred information. Furthermore, MFEA-CoD is extended to multitask novelty-augmented optimization, where behavioral novelty is jointly considered with objective information to alleviate premature convergence caused by deceptive objectives. Experiments on synthetic basin-type problems, deceptive maze navigation problems, MuJoCo policy optimization problems, and generative novelty search problems demonstrate that MFEA-CoD improves the efficiency of discovering diverse novel solutions and shows clear advantages in deceptive objective landscapes.

Long-context LLM services now sustain prompts with hundreds of thousands to millions of tokens, making the key-value (KV) cache a first-order serving cost. Because the cache grows linearly with context length, it can exhaust GPU memory, force smaller batches, and reduce serving throughput. Prior KV cache compression techniques typically target only the sequence dimension or only the channel dimension, which leaves limited headroom as context windows scale. Compressing both dimensions promises higher memory reduction, but applying the two forms of compression directly leads to significant accuracy loss. This paper introduces MosaicKV, a dynamic two-D (dimensional) KV cache compression system for extremely long-context serving. MosaicKV uses dynamic two-D compression to address the accuracy challenge, exploiting the non-uniform importance distribution of elements within the KV cache. Instead of applying one compression pattern globally, MosaicKV identifies important elements for each KV vector and selects compression strategies at the granularity of KV cache segments. To address the performance challenge, where fine-grained sparsity and compression management overhead can offset the gains from compression, MosaicKV introduces compressed KV cache management. This mechanism uses underutilized GPU and CPU resources to maintain compressed KV caches and accelerate attention computation. Evaluation on an H800 GPU with multiple LLMs shows that MosaicKV delivers up to 16x attention speedup, 4.8x lower decode latency, and 7.3x higher throughput than the uncompressed baseline. At the same time, it reduces memory usage by 3x and incurs only 1.76% average accuracy loss on LongBench and RULER.

Table extraction from business documents relies on a cascaded pipeline where Table Detection (TD) first localizes tables and Table Structure Recognition (TSR) then recovers their internal layout. Building task-specific training sets for this pipeline is costly, particularly for TSR which requires fine-grained structural annotations. Active learning (AL) can reduce this annotation burden, yet most AL strategies are designed for single-model tasks and do not account for inter-stage dependencies in cascaded architectures. In this work, we present the first adaptation of Uncertainty Herding (UHerding), a hybrid coverage-uncertainty sampling method originally proposed for image classification, to cascaded object detection pipelines. We propose two pipeline-aware extensions that exploit the TD-to-TSR dependency: RankFusion adds dual-manifold coverage over both detection and structure representation spaces, while CAPA further incorporates stage-dependent gating and per-task uncertainty calibration. Extensive experiments across two public (PubTables-1M and FinTabNet) and two private table extraction datasets, with various annotation budgets (from 71 to 500 documents) show that UHerding generalizes well to table extraction, outperforming each baseline. Among pipeline-aware variants, RankFusion achieves higher expected gains but at the cost of greater variance, while CAPA emerges as the most consistent strategy, outperforming standard UHerding on three out of four datasets.

The bird's-eye view (BEV) representation enables multi-sensor features to be fused within a unified space, serving as the primary approach for achieving comprehensive 3D perception. However, the discrete grid representation of BEV leads to significant detail loss and limits feature alignment and cross-modal information interaction in multimodal fusion perception. In this work, we break from the conventional BEV paradigm and propose a new universal framework for multi-modal fusion based on 3D Gaussian representation. This approach naturally unifies multi-modal features within a shared and continuous 3D Gaussian space, effectively preserving edge and fine texture details. To achieve this, we design a novel forward-projection-based multi-modal Gaussian initialization module and a shared cross-modal Gaussian encoder that iteratively updates Gaussian properties based on an attention mechanism. GaussianFusion is inherently a task-agnostic model, with its unified Gaussian representation naturally supporting various 3D perception tasks. Extensive experiments demonstrate the generality and robustness of GaussianFusion. On the nuScenes dataset, it outperforms the 3D object detection baseline BEVFusion by 2.6 NDS. Its variant surpasses GaussFormer on 3D semantic occupancy with 1.55 mIoU improvement while using only 30% of the Gaussians and achieving a 450% speedup.

Prenatal anomaly classification and localization is of critical importance for fetal health and pregnancy management. Although ultrasound (US) is the primary modality for prenatal screening, accurate diagnosis remains challenging due to the low prevalence and high heterogeneity of anomalies. Existing deep learning methods for prenatal tasks rely on large-scale annotated datasets, which are difficult to obtain in practice. Although few-shot learning alleviates data scarcity, it typically requires fine-tuning for new categories, limiting its practicality in resource-limited clinical settings. To address these challenges, we propose a training-free framework for multi-class prenatal US anomaly classification and localization that operates with only a few reference images per class, representing the first exploration of this setting. Our framework comprises three key components: (1) a memory bank with multi-granular prototypes that explicitly models both class-level semantics and anomaly characteristics; (2) a prototype-driven soft merging mechanism that aggregates discriminative features to detect the anomaly region; and (3) a class-aware refinement strategy that leverages prototype consistency to improve category prediction. Extensively validated on a multi-center prenatal US dataset containing 1,149 cases, with a total of 2,357 images and 9 categories, our proposed method outperforms the competitors.

Microservice availability is commonly assessed by fault injection and chaos experiments, but such experiments are costly, operationally risky, and difficult to repeat for every architectural change. Distributed tracing and deployment metadata provide cheaper evidence, yet they usually remain descriptive: they show which services interacted, not what endpoint-level availability property follows. This paper proposes a formal runtime availability model based on stochastic connectivity for resilience-oriented analysis of microservice endpoints. It treats endpoint availability under explicit fault scenarios as a measurable facet of microservice resilience, combining a typed service-dependency graph, a replication map, a probability measure over node and edge states, and request-specific success predicates. Its semantics separates computational failures of service replicas from communication failures of logical dependencies, showing that replication cannot compensate for bottleneck dependencies. The model can be reconstructed from traces and deployment artifacts, parameterized for architectural what-if analysis, and analyzed by Monte Carlo simulation before or alongside fault injection. We define the model, its trace-to-model construction, elementary semantic properties, and a synthetic adequacy study. The study matches closed-form oracle cases within sampling error and exposes boundaries caused by edge bottlenecks, correlated failures, missing traces, and time-dependent failures.

Large language models can generate polished scientific text that includes unsupported claims, allowing hallucinations to enter the archival record. Assessing this risk via technical statements is difficult and often requires expert judgment, but citations provide a more auditable surface: a reference either resolves to a real scholarly work with compatible authorship, or it does not. We measure citation hallucination in peer-reviewed proceedings using a conservative definition limited to identity-level failures: non-existent works and substantial author-list mismatches. We explicitly exclude ordinary bibliographic drift (e.g., venue/year differences, publication-status updates, minor name variants). To audit citations at scale, we build RefChecker, a verification pipeline that resolves bibliography entries against multiple bibliographic sources and escalates unresolved cases to web-search re-verification. We apply RefChecker to accepted camera-ready papers from ICLR, ICML, NeurIPS, and USENIX Security. Hallucinated citations have entered the archival record. While reference-level rates are usually below 1%, proceedings are large enough that paper-level failures are visible: in 2025, roughly one in twenty NeurIPS and USENIX Security papers contains at least two likely hallucinated academic-paper-like references under our strict definition. We also observe post-ChatGPT increases in several venues, including a tail of papers with 5+ failures in a single bibliography, and likely hallucinated citations even among award-winning papers. These results suggest peer review alone does not reliably enforce citation integrity, yet auditing is tractable (about 0.04$ per paper in one venue-scale scan). We open-source RefChecker for routine, reproducible citation verification before publication (https://github.com/markrussinovich/refchecker).

Table Structure Recognition (TSR) aims to recover the row and column layout of tables from document images, a key step in document understanding pipelines. Accurate TSR depends on precise boundary localization: small errors in row or column boundaries can propagate into incorrect cell assignments and structural inconsistencies. Yet detection-based approaches treat table elements as generic objects, ignoring a fundamental property of table layout: rows and columns play structurally distinct roles and their boundaries carry unequal importance. We propose an Edge-constrained Fine-grained Localization loss (EFL) that formalizes this structural asymmetry by encoding table-specific geometric priors into the training objective: row-like elements are supervised with emphasis on their horizontal boundaries, while column-like elements prioritize vertical boundaries. Implemented within a real-time detector with distribution-based boundary refinement (D-FINE), EFL operates during training only and guides boundary refinement toward structurally meaningful adjustments with no change to the inference pipeline. The proposed approach, ConRTF, is also data-efficient, maintaining robust accuracy with as few as 2k--3k annotated tables. Experiments on PubTables-1M and two private datasets show consistent improvements over the optimized baseline and several real-time detectors including RT-DETRv2 and YOLOv10-11, with gains of up to +1.6 GriTS points at equal inference speed.

Mechanistic modeling via ordinary differential equations (ODEs) provides interpretable descriptions of complex dynamics and enables inference of underlying mechanisms, which is particularly valuable in clinical settings. However, in rare diseases, both the structure and parameters of the model are typically unknown, while individual-level data is scarce, noisy, heterogeneous, and subject to privacy constraints. In such settings, population-level summary statistics provide a practical privacy-preserving data representation, while capturing heterogeneity further requires modeling parameters as distributions rather than fixed values. Yet no existing method jointly discovers ODE structure and refines parameter distributions solely from summary statistics. We present AgentODE, an end-to-end framework that addresses this gap. An LLM proposes candidate ODE structures, while a tool-augmented inference agent iteratively refines parameter distributions through a diagnosis--update loop, operating on population-level summary statistics alone. We evaluate AgentODE on three benchmark problems across different fields and two clinical datasets, including the rare disease recessive dystrophic epidermolysis bullosa (RDEB), with only 231 observations across 46 patients. AgentODE recovers functionally consistent ODE structures across all settings, and experiments on RDEB demonstrates that in sparse and noisy data settings reasoning from summary statistics promotes mechanistically principled structure discovery, whereas baselines with individual-level data access recover implausible structures despite better predictive performance. AgentODE opens new possibilities for mechanistic modeling of rare diseases directly from population-level summary statistics, where data scarcity and privacy constraints have traditionally limited such analyses.

Retrieval-augmented generation (RAG) under a fixed reader-context budget forces a selection problem: of the evidence retrieved, only a fraction can be shown to the reader. We argue that document recall -- the standard retrieval metric -- is the wrong quantity to optimize in this regime, and we make two contributions. First, as a general contribution, we introduce answer-in-context, a diagnostic that measures whether a gold answer survives as a contiguous span in the packed reader context (not the retrieved set). It predicts answer F1 better than recall (r=0.39-0.55 vs. about 0.31), separates answer quality roughly five-fold (0.60 vs. 0.12 on HotpotQA), and carries information beyond retrieval: it adds Delta R squared=0.17 over recall and shows a 4.6x EM gap even among questions where all gold was retrieved. We also confirm it interventionally: on 2WikiMultiHopQA a packing change that raises coverage but not answer-in-context yields no accuracy gain. Second, as a conditional contribution, we cast reader-context construction as budgeted monotone submodular maximization and build a packer that jointly optimizes relevance, query coverage, representativeness, and diversity. On HotpotQA with a 160-token budget and a 3B reader it beats a strong focused heuristic, MMR, and naive packing -- by up to +5.1 F1 at equal-or-lower token cost, across three seeds. Crucially, we map the scope of this win honestly: it requires the conjunction of (i) multi-hop complementary structure, (ii) retrieval that surfaces the evidence, (iii) a binding but not extreme budget, and (iv) a reader weak enough that evidence density, not reading capacity, is the bottleneck. A quantization-controlled reader-scale ladder (3B to 7B to 14B) shows the edge over the heuristic is absorbed by 7B and significantly reverses by 14B, while the diagnostic explains every boundary with a single variable.

Multilingual fluency often invites a stronger assumption: a model that can speak a user's language must also understand the culture encoded by that language. We call this the Illusion of Cultural Alignment. To test this assumption directly, we introduce MSQA, a benchmark of 1,064 natively sourced questions across 11 language groups, five cultural dimensions, and three difficulty tiers. Unlike translated benchmarks, MSQA targets locally grounded knowledge and reduces shortcuts from English-centric cross-lingual transfer. Evaluating 18 LLMs, we find substantial cultural degradation and a pronounced Locality Effect: cultural competence tracks pre-training exposure more closely than general reasoning ability. We further show that common inference-time remedies do not dissolve the illusion. Models remain overconfident on unfamiliar cultural questions, repeated sampling yields unstable rather than reliable correctness, and retrieval augmentation helps unevenly on long-tail facts. These findings indicate that cultural alignment cannot be inferred from multilingual ability alone and requires deeper intervention than calibration, sampling, or retrieval at inference time

Despite the increasing sophistication of industrial AI systems, the ability to reliably detect subtle and noisy anomalies in complex time series data remains a critical yet unresolved challenge. In large-scale industrial applications, labeling time series data is often prohibitively expensive and time-consuming, making unsupervised learning a practical and widely adopted approach. However, existing unsupervised methods frequently struggle to distinguish near-normal anomalies from normal patterns and are vulnerable to noise contamination within normal samples. To address these limitations, we propose a novel framework that leverages active learning to iteratively enhance the performance of unsupervised models. Our framework's core contributions are (1) a masked time-series reconstruction feedback strategy that forces the model to learn robust temporal dependencies, and (2) a minimax learning strategy that promotes robustness by differentially treating normal and abnormal samples. This process encourages the model to better capture the dynamics of subtle and noisy patterns. The proposed framework is evaluated across 28 test cases involving four multivariate time-series datasets and seven unsupervised backbone models. Experimental results demonstrate a 12.39% improvement in AUC compared to the original models, confirming that our method can be readily integrated into existing unsupervised reconstruction-based anomaly detection systems to significantly enhance their performance.

Skeleton-based action recognition has achieved remarkable success by exploiting joint coordinates and their topological connections, yet prevailing methods overwhelmingly assume complete and clean skeleton inputs. In real-world deployments, such as egocentric vision, crowded surveillance, wearable devices, or edge robotics, limited field-of-view (FoV) frequently causes substantial joint visibility dropout, leading to severe performance degradation that existing models are largely unprepared to handle. To bridge this critical yet underexplored gap, we introduce PartialVisGraph, a novel hypergraph framework tailored for robust skeleton action recognition under constrained FoV. We first construct highly expressive hypergraphs by introducing learnable virtual hyperedges that form a soft incidence matrix, capturing flexible high-order dependencies beyond conventional pairwise graphs. We then propose the Single-Head Sample-Adaptive Transformer, which adaptively aggregates joint features onto hyperedges while explicitly incorporating a visibility prior. This prior selectively gates information flow, preventing occluded or out-of-view joints from corrupting reliable feature propagation. We further establish rigorous evaluation protocols with realistic FoV simulation benchmarks on NTU RGB+D 60 and 120. Extensive experiments demonstrate that PartialVisGraph consistently achieves state-of-the-art accuracy under partial visibility, with gains of up to 68.8\% on subsets with severe FoV restrictions compared to recent strong baselines, while remaining superior on full-visibility settings. Our approach offers a principled and practical pathway toward deployable skeleton-based action understanding in unconstrained environments.

Self-conditioning is a core technique that enhances continuous flow-based language models, where the model learns to denoise generated text by conditioning on its own denoising estimate. While empirically successful, its performance improvements are poorly understood. Moreover, there is growing interest in the use of few-step generators based on flow maps, for which how to leverage self-conditioning is unclear. Here, we show that flow language models with self-conditioning solve a fixed-point iteration that bootstraps the performance of the learned denoiser. We use this viewpoint to formulate fixed-point flows, a two-dimensional class of self-conditioned flows, where the first dimension represents the flow process and the second represents the fixed-point iteration. We show that fixed-point flows define valid flow maps, and show that they can be distilled from self-conditioned flow models by compressing both fixed-point iterations and the flow process, the former with fixed-point distillation and the latter with flow map distillation. Our resulting flow map language model, FMLM$^\star$, outperforms state-of-the-art self-conditioned models and few-step models in one- and few-step generation on OpenWebText. Code is available at https://github.com/Ugness/self-conditioned-fmlm.

Large Language Models have emerged as programming assistants. However, the efficacy of code generation is constrained by the quality of input requirements, which are frequently ambiguous, incomplete, or underspecified. While LLMs excel at one-shot code synthesis, their ability to proactively clarify intent remains underexplored, as a critical trait for robust software engineering. Existing benchmarks largely overlook this interactive bottleneck, assuming perfectly specified prompts that do not reflect the iterative nature of requirement elicitation. To bridge this gap, we introduce ClarifyCodeBench, a novel interactive benchmark for evaluating LLMs' capability in resolving requirement ambiguity. Constructed from real-world programming tasks, ClarifyCodeBench features high-quality manual annotations, including N unique ambiguity types, associated clarification questions, and corresponding ground-truth answers. Furthermore, we formalize two rigorous metrics to assess the interaction quality: Turn-discounted Key Question Rate, which penalizes inefficient questioning, and Optimal Round Adherence, which measures the precision of the elicitation process. We conduct a systematic evaluation of six state-of-the-art LLMs using ClarifyCodeBench. Our empirical results yield three critical insights: 1) Capability Decoupling: Strong code generation performance does not inherently translate to effective requirement clarification; 2) The Reasoning Paradox: While increased computational thinking enhances code correctness, it yields marginal gains in identifying ambiguities; 3) The Multi-ambiguity Ceiling: LLMs' clarification performance degrades sharply as the density of ambiguities increases, revealing a significant bottleneck in handling complex, real-world specifications. Our work underscores the necessity for future AI4SE research to transition from static synthesis to interactive elicitation.

Well-designed autonomous driving datasets have fundamentally shaped research progress, yet existing literature primarily describes what datasets contain rather than how to strategically design impactful ones. This is especially limiting for small and medium-sized labs and startups that cannot afford to misallocate scarce resources. We argue that impactful dataset creation begins with a diagnosis: whether a research question is blocked by a data problem or an evaluation problem, and proceeds by selecting the minimal data operator(s) that closes the resulting gap, recording new data only when no cheaper operator(s) suffices. We analyze the evolution of major autonomous driving (AD) datasets through this lens and distill a strategic framework spanning gap identification, operator choice, sensor suite design, and annotation strategy. We ground the framework in a running case study of our KITScenes dataset family. The datasets are available at: https://kitscenes.com/

LLVM is a widely used compiler infrastructure whose scale and complexity make issue resolution labor-intensive and challenging. Although large language models (LLMs) have recently achieved remarkable success in issue resolution, their effectiveness on complex system-level LLVM compiler remains largely unexplored. To address this gap, we introduce LLVM-Bench, the first large-scale benchmark for LLVM issue resolution, containing 423 real-world, validated tasks collected from the LLVM project. We further develop LLVM-Gym, a scalable evaluation platform that automates issue reproduction, patch application, compiler building, and test execution. Using LLVM-Bench and LLVM-Gym, we conduct a comprehensive study of four representative LLMs, six retrieval configurations, and three agents. Our results show that current LLM-based issue resolution techniques remain limited on LLVM-Bench, with patch invalidity and build failures as the dominant failure modes. We further reveal a strong complementarity among different LLMs and agents, motivating LLVM-Ens, a lightweight ensemble approach that expands the patch space through integrating the patches generated by diverse techniques, filters incorrect and redundant candidates, and identifies the most promising solution. Our results show that LLVM-Ens achieves a resolution rate of up to 21.99%, further improving LLVM issue resolution.

Long-horizon LLM agents accumulate tool results, files, plans, and user constraints that are too structured to be treated as a disposable text suffix. Current systems mostly rely on in-run heuristics such as chronological pruning and tool-output masking, or on final self-summary near a context limit. Heuristics are cheap but blind to future dependencies; summaries preserve narrative state but often hide exact evidence, locators, and editable artifacts. We present Self-GC, where GC denotes self-governing context while deliberately echoing garbage collection: the system does not merely reclaim unused tokens, but governs the lifecycle of agent context objects. Self-GC turns user turns, tool spans, and skill state into indexed objects; asks a side-channel planner to propose fold, mask, and prune actions; and lets the harness enforce recoverable sidecars, safe commit boundaries, and cache-aware commit. On a 33-session Hard Set, Self-GC prunes 43.95% of prefix tokens while leaving 84.85% of future continuations unaffected, compared with no-impact rates of 54.55% to 69.70% for heuristic baselines. On a 332-session production-derived suite, three planner backbones reach no-impact rates of 91.27% to 94.58%, while baselines remain at 77.71% to 87.46%. In production, an online account-level split reduces daytime average input tokens by 10% to 15%, with peak reductions near 20%. These results point to context management as runtime lifecycle control over indexed, recoverable objects rather than post hoc text cleanup.

Black-box optimization is a fundamental science and engineering tool that makes it possible to optimize objectives without gradient information. Unfortunately, as it often requires many function evaluations, it can be challenging when each one is costly. This is especially true when the evaluation function is noisy or failure-prone, and when high-performing solutions are confined to thin, curved, or disconnected regions of the search space. Existing methods leveraging generative models to navigate these subspaces are built to sample from reward-aligned distributions. As a result, they require a large number of evaluations to align their sampler effectively, making them impractical in low-budget settings. We propose SPARROW, an algorithm that completely decouples the generative prior from the reward signal. SPARROW can use any sampler with a known corruption process and trained on unevaluated data, as a fixed, structured proposal operator. Optimization proceeds by rank-based guidance over an archive of evaluated candidates. SPARROW can navigate complex geometries, handle unreliable reward signals, and perform effective optimization under very low evaluation budgets. We provide asymptotic convergence guarantees over the sampler support and demonstrate strong empirical performance on problems with unreliable rewards and geometrically complex landscapes.

Comparing transformer backbones for image segmentation is confounded: each is paired with a different decoder, recipe, and pretraining, so reported differences rarely reflect the backbone itself. We introduce the Lightweight Universal Mask Adapter (LUMA), a lightweight, backbone-agnostic mask-transformer head that treats any backbone as a black-box feature extractor, letting a set of queries read from its features through cheap cross-attention. LUMA matches the accuracy of EoMT, the state-of-the-art efficient ViT-segmenter, at lower cost, while attaching unchanged to isotropic, hierarchical, convolutional, and mixture-of-experts backbones alike. Holding this head fixed, we benchmark 20 backbones, 11 pretraining schemes and a range of resolutions on ADE20K and Cityscapes under one modern recipe. We find that ``efficient'' token mixers fail to deliver efficiency even at the high resolutions that motivate them, with plain ViT holding the throughput Pareto-front at every resolution. Additionally, the pretraining objective, not the architecture, the lever the field has tuned hardest, governs segmentation quality.

Vision Language Models (VLMs) have demonstrated remarkable capabilities in multimodal reasoning tasks, yet they still suffer from recurring failures, such as skipping key visual checks, misapplying domain rules, and hallucinating unsupported concepts. Most existing solutions rely on supervised fine-tuning (SFT) and reinforcement learning (RL), which are expensive to iterate and can be brittle under distribution shift. To this end, we propose Multimodal Mistake Notebook Learning (M2Note), a training-free continual evolution framework that externalizes learning into an editable memory. M2Note transforms failed trajectories into compact subject-guidance notes: the subject summarizes the underlying domain and concept, while the guidance provides actionable verification steps that can be reused in future inference. At test time, M2Note retrieves relevant notes via multimodal retrieval-augmented generation (RAG) and appends them to the model context, steering reasoning away from previously observed pitfalls. To stabilize continual evolution, we adopt batch-level post-verification with rollback, which commits notebook edits only if they improve performance on the same batch, reducing noisy updates and preventing regressions. M2Note supports both self-evolving, where the same VLM acts as solver and supervisor, and cross-model evolving, where a stronger supervisor guides a weaker solver, enabling capability transfer without weight updates. Experiments on six multimodal reasoning benchmarks show consistent improvements across domains and backbones, while achieving strong cost and sample efficiency and remaining complementary to Chain-of-Thought (CoT) prompting.

The classification accuracy of pretrained Vision-Language Models (VLMs) relies on the quality of the text prompts. Handcrafted templates and Large Language Model (LLM)-generated descriptions not only make predictions more interpretable, but also enable reuse of the same prompts across heterogeneous VLMs. Recent works construct task-adapted text prompts with a small number of labeled images. However, existing few-shot text prompting methods do not explicitly focus on misclassified examples during prompt construction, leading to only marginal improvements even as more shots become available. To fully exploit few-shot supervision, we propose Text Prompt Boosting (TPB), an AdaBoost-inspired framework that treats each text-prompt-based classifier as a weak learner and sequentially aggregates them into a strong ensemble by explicitly targeting hard, misclassified examples. Extensive experiments show that TPB preserves task-intrinsic, model-agnostic cues in text space, enabling robust cross-model transfer. Across eleven classification benchmarks, TPB improves accuracy on the source model and preserves shot-driven gains when transferred to larger, more capable VLMs, where existing methods struggle to sustain such improvements.

We study distributed online submodular maximization under partition matroid constraints, in which multiple agents select a limited number of actions from their own subsets sequentially to maximize the cumulative value of a sequence of objective functions. We develop a unified algorithmic framework that accommodates full-information and bandit feedback models. For both feedback models, we prove that the proposed algorithms achieve sublinear $(1-1/e)$-regret guarantees, which are comparable to those achieved by existing centralized counterparts. Furthermore, to tackle the sampling violation issue caused by continuous relaxation and rounding, we develop a bounded stochastic pipage rounding scheme and show that the probability of sampling violation vanishes asymptotically. As a result, the cumulative sampling violation remains sublinear in $T$, which is further shown to be not improvable under certain conditions. Numerical results validate the theoretical findings in this paper.

Graphs are a complex and versatile data structure used across various domains, with possibly multi-label nodes playing a particularly crucial role. Examples include proteins in PPI networks with multiple functions and users in social or e-commerce networks exhibiting diverse interests. Tackling multi-label node classification (MLNC) on graphs has led to the development of various approaches. Some methods leverage graph neural networks (GNNs) to exploit label co-occurrence correlations, while others incorporate label embeddings to capture label proximity. However, these approaches fail to account for the intricate influences between labels in non-Euclidean graph data. To address this issue, we decompose the message passing process in GNNs into two operations: propagation and transformation. We then conduct a comprehensive analysis and quantification of the influence correlations between labels in each operation. Building on these insights, we propose a novel model, Label Influence Propagation (LIP). Specifically, we construct a label influence graph based on the integrated label correlations. Then, we propagate high-order influences through this graph, dynamically adjusting the learning process by amplifying labels with positive contributions and mitigating those with negative influence. Finally, our framework is evaluated on comprehensive benchmark datasets, consistently outperforming SOTA methods across various settings, demonstrating its effectiveness on MLNC tasks.

Vision-Language-Action (VLA) models often fail to perform the same learned tasks under environmental shifts, such as changes in camera pose and shifts to a different but similar robot (e.g., from Panda to UR5e). Adapting these models to the shifted environment (i.e., target domain) often requires training on multiple demonstrations for each task, which are costly to collect. To reduce the burden of data curation and training, we propose an analogy-based method that adapts VLA models under environmental shifts through weight vector arithmetic with domain-specific information addition, named Domain ARiThmetic (DART). Unlike prior approaches, DART requires collecting only a single demonstration, enabling efficient adaptation. To accurately isolate domain-specific information for addition, DART performs subspace alignment between singular components in weight vectors to filter out noisy components. In both simulated and real-world experiments, DART outperforms existing VLA adaptation methods in one-shot scenarios across diverse visual and embodiment shifts. Code is available at https://github.com/snumprlab/dart.

We propose YOMI-Bench, a benchmark for evaluating kanji reading and phonological understanding of large language models (LLMs) for Japanese. In Japanese, a single kanji character often has multiple possible readings, making it difficult to infer the correct reading from surface-level text alone. Due to these linguistic characteristics, it is empirically known that LLMs exhibit low performance in kanji reading for Japanese. The proposed YOMI-Bench consists of four tasks specifically designed to evaluate kanji reading performance in Japanese. In our evaluation using YOMI-Bench, we assessed one multilingual open LLM, four Japanese-specific open LLMs, and five commercial LLMs. As a result, we found that even Japanese-specific models show low performance, and that commercial models also perform poorly on generation tasks that require consideration of kanji readings.

Explanations for emotion classifiers are usually produced post hoc, with no guarantee that they reflect the computation behind the label. We present an explication interface for event-based emotion analysis. A parser maps the input text to an explication, a short script in the closed vocabulary of Natural Semantic Metalanguage organized into twelve typed slots, and a fixed decision list of rules transcribed from published semantic definitions computes the label from the explication alone. The faithfulness guarantee is therefore causal and definitional, while all empirical risk lives in the learned parser, which the per-line entailment interface makes auditable against the input. On crowd-sourced event descriptions, our fine-tuned parser reaches 0.33 accuracy and 0.48 selective accuracy on a small held-out set, suggesting that the interface trades insignificant accuracy difference to a black-box model for a verifiable, inspectable decision basis for first-person event-based emotion analysis. We also release EmoExpl-1200 with per-line verification metadata and the full rule set.

Reinforcement learning has proven to be a valuable tool in the creation of advanced AI and robotic systems, contributing to everything from game playing to robotics to foundation models. Through trial-and-error, these AI systems typically learn one, near-optimal behavior to solve their tasks. However, there are many use cases in which one would like to assert some level of control, preferably in real time, over how the task is solved. We refer to these modifications of a core task as styles. We combine universal value function approximators (UVFAs) with carefully selected training scenarios, learning algorithms, and data augmentation to create a framework for coaching agents that exhibit styles in complex domains. We demonstrate the framework's application in the AAA video games Horizon Forbidden West and Gran Turismo, and in an open-source humanoid test domain. Despite the different nature of the domains -- car racing, stylized game combat, and humanoid walking -- each agent shows strong coherence to the style requests while still satisfying the main task in its domain. Importantly, the techniques outlined in this paper allow an end user to choose the final behavior at run time, giving them flexible control over the final executed performance.

Advances in generative AI are rapidly increasing the quality and commercial value of generated music, and this progress depends on large catalogs of creators' recordings. This raises a central question for platform design: how should creators be compensated when their work is used to train generative AI models that in turn produce commercial outputs? We develop a framework for fairly compensating creators in generative-music markets, where each creator's payment depends on a data-attribution score estimating their contribution to model outputs. Compared to past compensation frameworks, our framework has two unique considerations: (1) attribution is traced to entire creator catalogs, not individual songs, and (2) the informativeness (signal-to-noise ratio) of the attribution score is an input to the payment mechanism. The framework yields a closed-form payment rule per creator and measures the welfare cost of inaccurate attribution for both creators and the platform. Whether the welfare-optimal contract is royalty-based or takes the form of fixed-fee licensing depends on how informative attribution is for that creator's catalog. We show that better attribution translates directly into welfare gains for both creators and the platform, yet under multi-platform competition a platform only captures gains from attribution improvements when its signal becomes the most precise in the market. To ground our framework in empirical behavior, we train acoustic and symbolic music generation models and measure the informativeness of scalable attribution techniques against a leave-one-catalog-out ground truth. Our experiments reveal that noisy attribution signals push payment toward fixed-fee licensing and diminish welfare for both creators and the platform, providing an economic motivation for further research on improved attribution.

A growing class of tools recovers a program from observations of its behavior using an untrusted generator, a neural model or a search, that proposes candidates with no correctness guarantee. We study how to make such recovery trustworthy, in the concrete setting of recovering a runnable Scratch program from a recording of its execution. The recording shows what the program does but never its code; many programs produce the same video, so the source cannot be recovered, and the right target is a program that behaves the same as far as the camera can tell, made precise with a lens. The core is a two-tier validation oracle with a deliberate verdict asymmetry. A static checker proves lens-equivalence to a reference and issues a certificate that, granting the partial-order independence quotient adequate, never accepts a wrong program; a renderer can only refute or witness finite agreement, never certify. Around it, Vid2Prog reads each sprite's motion, visibility, and timing from the video and a known-asset manifest and synthesizes a candidate source-free; a closed loop renders and runs recovery again for ground truth. Under the exact lens the oracle makes no false accept on 246 labeled differing pairs, including an adversarial battery built to trap its concurrency quotient; on inputs outside the vocabulary and on real projects it abstains or refutes, accepting none we test. In-vocabulary recoveries reproduce their source frame for frame and 80% earn a static certificate, while whole real projects, mostly outside the vocabulary, recover at 14%, a vocabulary-bound rate the system never inflates with a wrong answer. A frontier vision-language model recovers none of the matched programs single-shot, which oracle-in-the-loop repair lifts only to a few while the structured pipeline recovers all, the gap a sound checker makes for an untrusted generator.

In settings such as fine-tuning and reinforcement learning, neural networks are often adapted under distribution shift. Standard adaptation methods typically optimize the target objective directly, inducing an abrupt change from the source training objective. This abrupt transition can distort learned representations, including features that may still be useful for the new task. We investigate whether a more gradual transition can improve adaptation. We propose loss smoothing, a simple approach that interpolates between the source and target training objectives at the start of adaptation. This smooth transition helps to preserve useful features from the source distribution while still enabling the model to specialize to the target distribution. Across controlled supervised shifts, pretrained vision adaptation, offline-to-online and online reinforcement learning, and language model fine-tuning, we find that loss smoothing consistently improves performance, suggesting that smoother objective transitions are a broadly useful tool for model adaptation.

Large language models (LLMs) are powerful pattern-completion systems, but their default operating mode - predicting the next token from a static context - does not reliably produce persistent, manipulable representations of an external world. Many tasks that look like "reasoning" in text become substantially harder once the environment is partially observable, stateful, and requires memory and structured hypotheses about hidden state. AGI Maze is a lightweight framework for building such environments without requiring high-dimensional sensory inputs. It provides a family of grid-based maze tasks with a clean API and multiple difficulty regimes. The goal is to create benchmarks where agents must learn and use world state representations, not just infer a local rule over readily provided observations. We provide an initial evaluation of several vanilla LLMs on simple mazes showing that they fail to represent mazes internally at LLM inference time. We also introduce a baseline agent, which is allowed to use its message history as a working memory to construct descriptions of observations at agentic runtime. Although this can improve performance, it is still insufficient for an LLM agent to reliably solve even small mazes within a step budget that is more than enough for humans.

Terminology such as "whitelist/blacklist," "master/slave," "man-hours," or "dummy value" has long been part of the technical vocabulary used in software artifacts, including source code, version histories, and documentation. In recent years, however, many of these expressions have been recognized as potentially non-inclusive and unwelcoming to groups historically underrepresented in software development, such as people of color, women, and individuals with disabilities. Consequently, a growing movement within the software industry has sought to replace these terms with more inclusive alternatives. Despite these initiatives, little is empirically known about how software developers perceive such terminology or how its continued use may influence their professional experiences and sense of belonging. This paper addresses the knowledge gap by examining how software developers perceive non-inclusive terminology in software and its perceived impact on team dynamics, productivity, belonging, and well-being. We surveyed open-source contributors and received 1,517 responses, of which 1,212 were complete and analyzed. On average, respondents reported low negative workplace impact overall; however, perceptions and impacts varied by demographic group. Women and non-binary participants, as well as respondents residing in the United States, were more likely to view the terms as non-inclusive. Among those who considered the terminology non-inclusive, non-binary participants reported higher overall negative impacts than male respondents, and female participants reported higher impact specifically on their sense of belonging.

Block-based languages such as Scratch let beginners assemble interactive programs from sprites and scripts. These programs are concurrent in practice: green-flag scripts, broadcasts, and clones run as cooperatively scheduled threads over shared sprite and stage state, and their authors never write a thread. We show that such programs contain schedule-sensitive behaviors whose observable result depends on an execution order the language leaves open. Editing, saving, or remixing a project can produce a copy with the same blocks but a different layer order, changing the order the virtual machine starts scripts. We formalize the schedule space a Scratch virtual machine can realize as the permutations of the initial executable-target order, and define schedule-robustness against a lattice of observation lenses over a fixed horizon. A partial-order exploration runs one schedule per dependence-equivalence class, and on projects small enough to enumerate, an independent oracle confirms it recovers every realizable outcome. On larger projects, representatives stand in for the factorial under the validated dependence model. SchedCheck implements this on the production Scratch VM. Across 224 real student projects, at least 21% of the concurrent ones are schedule-sensitive at the grading lens, and a uniform random sample of public projects replicates the rate at 17.6%, with two real remixes of a deployed animation arranging its letters differently. On hand-built fault pairs and a generated benchmark of 32 spec-defined faults across four classes, the tool detects and localizes every schedule fault, with a logic-fault control reporting clean. The oracle exposed four unsoundness gaps in the dependence model, all repaired. The method is parametric in the execution model, instantiating unchanged on a second cooperative event loop.

Post-quantum cryptography and privacy-preserving technologies are expected to play a central role in future secure communication systems. Lattice-based PQC schemes such as ML-KEM (CRYSTALS-Kyber) and ML-DSA (CRYSTALS-Dilithium) rely heavily on large-degree polynomial arithmetic, making the Number Theoretic Transform (NTT) a key computational primitive. Although existing hardware accelerators exploit parallelism and pipelining to support both NTT and INTT, their efficiency is often limited by the overhead of modular reduction and correction steps, inverse-transform scaling operations, and suboptimal FPGA implementations. This work addresses these limitations by proposing parallel iterative NTT/INTT accelerators based on optimized unified butterfly units. We introduce a novel redundant number representation that eliminates conditional corrections for both Montgomery modulo multiplication and combined subtract-multiply operations, and integrate inverse-transform scaling into existing arithmetic hardware to avoid dedicated scaling units. Furthermore, we design hierarchical Montgomery multipliers that map efficiently onto FPGA DSP resources, reducing hardware cost while enabling high operating frequencies. FPGA-based experimental results demonstrate higher clock frequencies, reduced execution times, and competitive resource utilization, supporting efficient NTT acceleration for PQC and related privacy-preserving applications.

Generalized Category Discovery (GCD) aims to recognize known classes while autonomously discovering novel ones in open-world settings. However, current approaches primarily focus on designing clustering objectives, often overlooking a critical bottleneck: standard vision backbones yield high-rank, entangled token representations that are ill-suited for unsupervised discovery of latent concepts and structures. In this paper, we propose Compositional Primitive Fields (CPF-GCD), a novel representation learning framework that reshapes the feature space to make such latent structure identifiable by enforcing a low-rank compositional organization. Our core hypothesis is that all categories, whether known or novel, can be expressed as compositions and spatial arrangements of a finite set of learnable visual primitives that capture reusable concepts. CPF instantiates this geometric constraint via a spatial field mechanism. Inserted between the backbone and the head, it rewrites noisy patch tokens through low-rank primitive mixtures, effectively decomposing images into reusable atomic parts and their spatial layouts. By explicitly modeling the spatial distribution of primitives, CPF enables novel categories to emerge naturally as new activation patterns over a shared vocabulary. This shifts the focus of representation from merely partitioning global embeddings to constructing a structured and separable primitive field. Extensive experiments demonstrate that CPF serves as a generic, plug-and-play module that consistently boosts performance across diverse GCD baselines, validating that identifying and leveraging low-rank compositional structure is a crucial inductive bias for open-world recognition.

Limited Memory Language Models (LMLMs) externalize factual knowledge to a database to enable deletion-based unlearning without retraining. Existing evaluations measure post-deletion correctness in aggregate and cannot tell whether a deleted fact persists through residual parametric memory, alternative retrieval paths, or near-neighbor retrieval artifacts. We propose a causal auditing framework that holds the model fixed and varies the database state at inference time across three interventions: FULL, DEL-ON, and DEL-OFF. The framework decomposes post-deletion behavior into parametric leakage L(f), retrieval-mediated correctness R(f), and a retrieval artifact rate grounded in the inference-time retrieval trace. We apply it to 12,228 alias-closure deletions across thirteen databases, including four adversarial topologies (Base, Alias, Noise, Collision) we construct in three domains, and six prompt formulations. Parametric leakage is near zero in every variant and every prompt style: the model rarely returns the deleted answer in the absence of retrieval. The residual that does survive lives in the retrieval graph: retrieval-mediated correctness and the retrieval artifact rate match within rounding everywhere, so post-deletion correctness is, in our audit, predominantly reconstituted from near-neighbor retrieval. This residual ranges from 0.7% on the released LMLM database to 13.6% on the most adversarial variant, and prompt formulation does not independently control how much of a deleted fact survives. These results suggest that, for this class of LMLM and deletion procedure, the unlearning boundary is drawn primarily by the database administrator rather than by the model.

We give a descent-free, alignment-free measurement of singular structure on trained networks. At a single frozen checkpoint the read recovers the order $k$ of each dead direction from the directional-Fisher rate, the master invariant from which the per-direction learning coefficient $1/(2k)$ follows exactly, in whatever basis the optimizer left. The same read classifies each direction, separating a genuine singularity, whose order the architecture fixes, from a flat gauge symmetry; the directional-Fisher magnitude settles the cases the order cannot. A pluggable detector supplies the directions for transformer, convolutional, and normalisation layers. The read recovers the architecture-predicted order across constructed cells and trained networks, including a fine-tuned vision transformer whose dead structure is the LayerNorm-kernel gauge and a from-scratch one whose compressed MLP forms a node-death at its activation order. Where the singular structure enumerates, the per-direction orders assemble, through the typed intersection of the loci, into the global coefficient $(λ, m)$ matching the closed form. The method removes the canonical-alignment and descent preconditions of the underlying rate result, turning order-recovery into a deterministic, architecture-general reading. We then map its reach into the Watanabe triple: the order determines the universal singular fluctuation $ν(k)$, though a trained network's realized $ν$ falls below it as the live structure absorbs the dead direction's data fluctuation, and the multiplicity recovers from the dominant structure under a single-locus assumption.

The game of Taboo requires describing a target word without using a set of forbidden words, so that other players can guess it. This deceptively simple task combines strict lexical constraints with the need for communicatively effective descriptions, making it a compelling playground for examining how LLMs navigate competing demands at inference time. We evaluate two open-weight models under conditions that intervene at progressively deeper levels of the generative process, from prompting to generation-time constraints to internal representations manipulations. We assess their outputs through forbidden word violation detection, LLM-as-a-judge measuring the degree to which generated descriptions successfully evoke the target concept for both human and machine guessers, and examining whether the strategies models adopt under constraint align with those of human players. Our results show that compliance with the rules of the game and communicative effectiveness trade off differently across conditions, and that models remain substantially weaker than humans as guessers, suggesting that lexical grounding under constraint is an open challenge for current language models.

Scientific literature search often requires more than retrieving papers from a single query: users' intents are underspecified, preference-dependent, and evolve through interaction. Existing search agents typically rely on fixed pipelines or implicit language-only reasoning, making their search strategies difficult to control, inspect, and refine. We introduce PaperPilot, a multi-turn literature search agent that frames scientific search as workflow induction. Given an anchor paper and a user query, PaperPilot constructs an executable DAG of paper-search operators, including keyword search, citation expansion, filtering, scoring, reranking, and evidence extraction. User feedback is then used to refine both the query and the workflow itself. We train PaperPilot with supervised workflow imitation and preference optimization over controlled workflow corruptions. Experiments show that PaperPilot-9B improves over the base Qwen3.5-9B toolset agent under multi-turn interaction, increasing Hit@5 from 58.0 to 77.0, MRR from 47.5 to 59.4, and nDCG@10 from 26.8 to 32.5, while reducing workflow execution errors from 9.5% to 0%. These results show that explicit, editable search workflows provide an effective and controllable interface for aligning literature search agents with complex scientific intent.

With the increasing deployment of heterogeneous robot fleets in industrial environments, efficient coordination remains a critical challenge. Real-time path planning must simultaneously accommodate high robot densities and heterogeneous motion capabilities, while communication delays, execution uncertainties, and other disturbances may cause robots to deviate from the temporal assumptions underlying planned paths. Such deviations can lead to excessive waiting and congestion propagation across the fleet. This paper presents SCALE, a reactive online coordination framework that enables real-time planning while maintaining robust execution. Within this framework, we introduce a motion-induced conflict reduction mechanism to support the online generation of feasible paths for online conflict resolution. To mitigate the effects of disturbances, we further design a generalized Conjugate Action-Precedence Hypergraph (CAPH) that adaptively adjusts precedence relations among robots. Extensive validation experiments, together with a three-day deployment in a warehouse, demonstrate the

Continuous diffusion language models such as ELF report record-low generative perplexity (Gen-PPL). We find a catch: these models repeat far more than human text, and Gen-PPL rewards rather than penalizes that repetition, so its low scores overstate quality. Strip the repetition and ELF-B's Gen-PPL rises from $19.5$ to $27.7$; the smallest model even posts the best Gen-PPL because it repeats most. We trace the repetition to its source: a contractive attractor along a \emph{single direction} in the self-conditioning feedback loop, the loop that feeds each step's clean estimate into the next. Because the failure is one-dimensional, a one-dimensional fix suffices, and we propose one. \textbf{ACE} (Attractor-Contrast-Escape) subtracts that single, label-free direction from the feedback at each step. Estimated once on the $105$M model, the direction cuts repetition to near the human level while keeping quality competitive, and transfers near-unchanged to the $342$M and $652$M models and across samplers; the same recipe recovers useful directions on other architectures. Since Gen-PPL itself rewards repetition, we instead measure the compute each fix needs to produce human-clean text, where ACE is $1.5$--$5\times$ cheaper.

We present a simple, yet general approach to study the scaling properties as the dimensionality of Metropolised MCMC sampling algorithms increases. The study relies ultimately on the symmetry of the Metropolis-Hastings formula. Our findings contain, as particular cases, many known results for the Random Walk Metropolis, MALA and other algorithms. In addition, they provide, in an easy way, new optimal scaling results for a variety of proposal mechanisms, including implicit proposals and proposals generated with the help of differential equation integrators. The analysis applies to targets that are products of a given, not necessarily univariate distribution, and also to cases where the different terms in the product are scaled differently. We show how to construct gradient-based MALA-like proposals where the variance of the proposal as the dimension $d$ increases may be taken as $O(1/d^μ)$, with $μ>0$ arbitrarily small, to be compared with the values $μ= 1$ for Random Walk Metropolis and $μ=1/3$ for MALA.

This study presents a comprehensive analysis of bird diversity across Sri Lanka by integrating spatial, temporal, and environmental data. Bird observation records were combined with environmental variables, including weather conditions, air pollution, the Normalized Difference Vegetation Index (NDVI), land cover, elevation, and Artificial Light At Night (ALAN), and rigorously preprocessed to ensure data quality. Spatial analyses were conducted on multiple grid scales (2 km, 5 km, 10 km) to evaluate patterns in species richness while minimizing sampling bias through spatial thinning. Temporal trends were assessed using effort-corrected metrics including rarefied richness and occupancy rates to account for variations in observation effort over time. Environmental drivers of bird diversity were examined using multivariate statistical models, including Poisson Generalized Linear Models (GLMs) and correlation analyses, to identify key associations between ecological factors and species richness. Additionally, community structure, dominance patterns, and beta diversity were analyzed to understand variations in species composition across regions and time. The study found that land-cover type is a stronger predictor of bird diversity than individual continuous variables such as NDVI or temperature alone. Urbanization, measured by ALAN, exhibits nuanced scale-dependent effects, supporting high abundances of a few generalist species while reducing overall richness. The findings provide actionable insights into the patterns and drivers of avian diversity in Sri Lanka, offering a scalable and reproducible framework for biodiversity research and conservation planning.

Sparse tangent portfolio optimization aims to learn an interpretable, low-cardinality portfolio in the tangency direction of the mean-variance frontier. However, the associated cardinality-constrained formulation is NP-hard, and standard predict-then-optimize pipelines often misalign forecasting accuracy with downstream portfolio quality. We propose an end-to-end decision-focused learning framework that reformulates Sharpe ratio maximization as a Disciplined Parametrized Programming (DPP)-compliant convex programming layer and replaces discrete selection with a smooth top-$k$ operator enforcing an exact cardinality $k$. This enables gradient flow through prediction, asset selection, and re-optimization, allowing the predictive model to directly optimize portfolio performance. Across four major equity markets, our method achieves competitive and often superior out-of-sample Sharpe ratios compared with historical and prediction-focused baselines, with particularly strong gains in larger asset universes. Our \href{https://github.com/feuerwerksh/Diffble-card-SR}{code} is publicly available.

Multi-image content has become an increasingly prevalent form of visual communication in social media, giving rise to a new safety issue, multi-image implicit toxicity (MIIT), where each image appears benign in isolation, but harmful semantics emerge when the images are interpreted jointly. MIIT is particularly challenging for existing commercial moderation APIs and models due to the lack of explicit risky cues in each image. This paper aims to study how to identify MIIT. We first provide a formal definition of MIIT and analyze three key challenges for its detection. To alleviate the scarcity of data in this area, we construct MIIT-dataset, an image-only multi-image safety dataset covering seven representative risk categories through an automatic generation pipeline. Finally, we train MiShield with progressively distilled reasoning supervision, enabling it to produce safety judgments accompanied by explicit analyses of the correlated entities that result in the hazards. Experiments show that MiShield-8B models outperform representative moderation services and even larger-scale models, revealing its effectiveness and practical value for this widely used visual format. Warning: This paper contains potentially sensitive content.

Understanding how aligned LLMs internally represent safety is critical for diagnosing alignment vulnerabilities, as it explains why jailbreaks succeed and informs the design of robust alignment strategies. Prior work shows that aligned LLMs encode harmfulness and refusal as separable directions in the residual stream at prompt-side token positions. We show that jailbreaks succeed at prompt encoding by suppressing either the refusal or harmfulness direction before any token is generated, with distinct attack classes occupying separable regions of the harmfulness-refusal plane. Extending the analysis to response-token positions, we find that the model recognizes harmful content while it is generating that content, even when it failed to recognize the input as harmful at the prompt side. Motivated by our findings, we introduce HARC (Harmfulness-And-Refusal Coupling), a fine-tuning method that pairs the two directions across both prompt and response positions. Since the intervention is confined to the harmfulness-refusal subspace, it leaves the rest of the residual stream intact and does not degrade general capability or inflate over-refusal. Across extensive experiments, HARC achieves the strongest robustness-capability-usability trade-off among six baselines spanning the major training-time and inference-time safety methods. The harmfulness and refusal directions at prompt and response positions transfer across the five model families and two scales we tested without architecture-specific tuning.

Retrieval-augmented generation (RAG) increasingly requires models to answer questions from multiple retrieved documents, where only some sources are relevant and the retrieved bundle may contain stale, noisy, or conflicting evidence. Existing contrastive decoding methods primarily focus on resolving conflicts between the model's internal memory and the retrieved context. In contrast, we study the complementary problem of intra-context conflict in multi-document RAG. To evaluate this setting, we introduce DRQA, a factual-conflict question answering benchmark derived from enterprise deep-research scenarios, where answers are grounded in synthetic enterprise-specific facts that are designed not to be recoverable from the model's internal memory. We further propose Dual-Confidence Contrastive Decoding (DCCD), a training-free decoding method that combines document-level confidence, which estimates whether a document appears sufficient for answering the question, with token-level confidence, which estimates whether that document supports a confident next-token prediction. DCCD selects positive and negative document-conditioned streams using these dual-confidence signals and scales a document-level contrast by their confidence margin. Across DRQA and standard multi-document QA benchmarks, DCCD achieves the best average performance among full-context and contrastive decoding baselines, with the largest gains on DRQA. These results highlight the importance of source-aware, confidence-gated decoding when retrieved evidence is internally conflicting.

Debugging the Linux kernel remains a formidable challenge due to its vast codebase, complex architecture, and low-level programming intricacies. Effective fault localization (FL) is thus essential for efficient kernel debugging and maintenance. While existing FL techniques (both traditional and LLM-based) have shown promise in general-purpose software, they are ill-suited for the kernel context. In particular, recent LLM-based techniques often treat bug reports and source code as plain text, lacking deep integration of kernel-specific knowledge, which limits their ability to identify root causes and achieve fine-grained localization. We present CoHiKer, a novel LLM-based FL technique tailored to the Linux kernel. CoHiKer introduces two key innovations: (1) contrastive reasoning, which identifies root causes by analyzing the behavioral divergence between carefully mutated passing and failing test cases, and (2) hierarchical context analysis, which systematically narrows the localization scope from files to methods by integrating crash reports, syscall semantics, inter-file dependencies, and kernel-specific features. Unlike prior techniques that rely on static understanding and full-code input, CoHiKer decomposes the localization task and enables structured LLM prompting to reason semantically over meaningful contexts. We evaluate CoHiKer on an extended Linux kernel bug dataset against five state-of-the-art baselines. CoHiKer consistently outperforms all competitors, improving Top-1 localization accuracy by up to 26.07% at the file level and 56.85% at the method level over state-of-the-art LLM-based baselines, while achieving up to 8.84% and 28.9% reductions in token consumption, respectively. Furthermore, CoHiKer demonstrates strong generalizability on the non-kernel dataset, with comparable gains (15.5% and 5.3% in Top-1 at file and method levels).

As Artificial Intelligence(AI)-based applications take off, a clear understanding of AI patterns can uplift the quality of AI applications. Many AI patterns have been proposed in the literature; however, their prevalence in real-life code has not yet been validated. Understanding the actual use of those patterns in practice can clarify our understanding both of the significance of these patterns and their utility. In this paper, we present a methodology to a) identify relevant patterns by mining the literature and then to b) validate their presence and prevalence in actual code repositories using active learning. To that end, we identify 14 AI pattern classes by mining 44 published AI pattern-related sources. Then we use an active learning approach to determine the prevalence of the most common pattern class across 100 GitHub open AI repositories. Using prevalence estimation, we propose bounds on the accuracy of the occurrences. The model achieves 56\% accuracy and 55\% recall in an 8-way classification task, significantly outperforming the 11\% random-chance baseline. Furthermore, the prevalence estimation offers usable bounds for analyzing pattern applications. This methodology provides a robust foundation to start understanding how AI patterns are used in practice, a field that currently lacks empirical data.

While recent advancements like the Poincaré ResNet have demonstrated the potential of learning visual representations directly in hyperbolic space, their optimisation remains hampered by the computationally intensive nature of Riemannian gradients and the strict boundaries of the manifold. Furthermore, standard hyperbolic networks treat spatial transformations of the same object as distinct hierarchical concepts, leading to redundant parameter usage and vanishing signals. We propose Equivariant Poincaré ResNets, combining hyperbolic geometry with discrete symmetry groups ($C_4$ and $D_4$). We identify critical roadblocks in applying Euclidean equivariance to hyperbolic space and propose geometrically safe tensor reshaping, left-regular permutations for hyperbolic group convolutions, and joint-orientation Poincaré Midpoint Batch normalisation. Empirically, embedding equivariance drastically reduces the optimisation space, accelerating convergence while accelerating convergence while respecting the boundary constraints of the Poincaré ball and preserving spatial-group equivariance.

Deep learning (DL) frameworks are critical AI infrastructures that often hide bugs with serious security implications. While dynamic approaches such as fuzzing are effective in uncovering these bugs, they require real test execution and incur high computational costs. Static analysis is a natural complement because it can detect bugs without runtime execution, offering fast and scalable testing. Unfortunately, there is still limited work targeting static analysis for DL frameworks due to their multilingual architectures and tensor-related program state. We present Phoenix, the first LLM-based static analysis technique for DL frameworks. Our key insight is that cross-language tensor flows in DL frameworks can be modeled, together with concrete code context, as a structured semantic bridge intermediate representation (SBIR) that LLMs can analyze for potential bugs in tensor semantic propagation. We implement this insight through a multi-agent workflow. A summarization agent first distills bug summaries from historical bug-fix patches and CWE rules. Guided by each summary, an extraction agent identifies bug-relevant repository symbols for code retrieval, and a generation agent synthesizes grounded SBIRs from the retrieved context. Finally, an analysis agent is leveraged to check SBIRs and report potential bugs. Our evaluation shows that Phoenix is a practical complement to dynamic DL framework testing for bug finding. To date, Phoenix has found 31 real new bugs in PyTorch for different heterogeneous hardware backends (Intel CPU, NVIDIA CUDA, and Apple MPS). Among them, 20 submitted bug-fixing patches have been merged into upstream.

Lightweight machine learning models are increasingly proposed for intrusion detection in Industrial Internet of Things (IIoT) networks due to their suitability for resource-constrained edge deployment. Most reported results evaluate these models only within their training network, leaving behavior on unseen networks unverified. This study trains four lightweight architectures on one IIoT dataset and evaluates them, without retraining, on two structurally distinct IIoT datasets using a feature representation restricted to attributes available across all three sources. Explainability analysis across two top-performing models shows both rely overwhelmingly on coarse port-category features; the most influential category occurs in source-domain attack traffic at 96 to 435 times the rate in the two target domains, indicating that coarsening port resolution relocates rather than removes a documented shortcut. Evaluation under naturally imbalanced class distributions reveals a further effect: the evaluation protocol used can reverse which target network appears to pose the greater generalization challenge. Adversarial robustness and recovery through limited target-domain exposure are also assessed; robustness to adversarial perturbation is unrelated to cross-network generalization, and recovery through adaptation varies considerably by architecture. These findings suggest deployment readiness should be assessed using cross-network evaluation under realistic class distributions, rather than within-domain accuracy alone.

Existing egocentric benchmarks have primarily constructed the egocentric setting from first-person-view data, which makes it difficult to evaluate egocentric perspective itself in isolation. However, understanding first-person-view input and taking an egocentric perspective are separable abilities, especially when first-person body cues are absent or when other agents are present. To isolate egocentric perspective understanding, we introduce EgoGapBench, a diagnostic benchmark for measuring action selection in multi-agent egocentric scenes. We define the ability measured by this benchmark as Egocentric Action Selection (EAS): selecting an appropriate action from the agent's perspective in the presence of other agents. On EgoGapBench, humans answer reliably, whereas both open-source and proprietary MLLMs perform substantially worse and systematically select actions performed by other visible agents. Fine-tuning on existing egocentric data fails to close this gap and can even be detrimental. In contrast, fine-tuning on EgoGapBench training data improves accuracy but does not reach human performance. These results show that EAS is difficult to acquire from first-person-view data alone, and that MLLMs should be evaluated and trained not only for scene understanding but also for egocentric action selection.

Few-step flow-map generators, such as consistency models and MeanFlow, accelerate sampling by directly learning long-range transport maps between noise and data. However, these models are typically deterministic, which makes them difficult to optimize with reinforcement learning (RL) post-training methods that require stochastic trajectories and well-defined likelihood ratios. Existing SDE-based stochasticization techniques are designed for velocity-based samplers with infinitesimal or finely discretized transitions, and therefore do not directly apply to long-range flow maps. In this work, we propose Flow-Map GRPO, an online RL post-training framework for deterministic few-step flow-map generators. The key component is Anchored Stochastic Flow Map Composition (ASFMC), a path-preserving stochasticization mechanism that introduces randomness through anchor-based conditional resampling while preserving the original marginal probability path of the deterministic flow map. We derive GRPO objectives for both single-time and two-time flow-map parameterizations. Experiments on few-step FLUX-based text-to-image generators, including MeanFlow and sCM, show that Flow-Map GRPO improves pretrained deterministic flow-map models across reward-based, perceptual, and task-level evaluation metrics. Our results demonstrate that deterministic few-step flow-map generators can be effectively aligned with RL post-training without modifying their original model parameterization or retraining them as native stochastic models.

As AI takes on more software work, the line between human and AI effort is shifting. Where developers draw that line around AI autonomy bears on how we design tools and roles that preserve meaningful work. Drawing on cognitive appraisal theory, work design, and automation research, we conducted a mixed-methods study of 448 professional developers at Microsoft to investigate their accepted levels of AI autonomy across software engineering work. Most developers accepted AI producing work under their oversight, although accepted autonomy varied substantively across tasks and individuals. Acceptance was lowest for identity-defining, human-facing, and design-oriented work, and higher among developers with more AI experience and risk tolerance. Task accountability was associated with lower odds of allowing AI to act on developers' behalf, whereas task identity was associated with lower odds of granting AI decision-making autonomy. Task demands had the opposite effect, increasing willingness to delegate decision-making to AI. Our findings suggest that preferences for AI autonomy reflect how developers cognitively experience their work, highlighting important considerations for designing meaningful work.

Scientific reasoning is an increasingly important capability of large language models, yet improving the robustness and efficiency of training such reasoning remains a key open challenge. We study this problem in instruction-based molecular optimization, where answer-only supervised fine-tuning (SFT) collapses multi-step reasoning and reinforcement learning with verifiable rewards (RLVR) suffers from sparse feedback. Reference-guided Policy Optimization mitigates both by anchoring policy updates to dataset-provided references, but its effectiveness is tightly coupled to reference quality: weak or misaligned references impose a performance ceiling. To overcome this ceiling, we propose active reasoning, a paradigm in which the policy actively decides, on a per-instance basis, when to imitate a reference and when to reinforce its own discoveries, while continuously upgrading what it imitates. We instantiate this paradigm as Active Group Relative Policy Optimization (Active-GRPO), realized through two coupled mechanisms: active imitate-reinforce and active referencing. The former performs imitation learning when the reference still outperforms the policy's own candidates, and shifts to self-improvement via reinforcement learning once the policy has generated molecules that surpass the reference. The latter continuously upgrades the reference itself by replacing it with the best policy-generated candidate discovered so far, progressively raising the imitation target and ensuring that reference guidance remains informative-rather than restrictive-throughout training. Across TOMG-Bench MOLOPT, Active-GRPO improves average SRxSim from 0.0959 for GRPO and 0.1665 for RePO to 0.1773 under matched three-seed evaluation, with statistically significant gains on LogP, MR, and QED.

Improvements in the technical performance of human--robot interaction (HRI) systems do not automatically translate into differences that human users can detect during live interaction. This paper investigates whether a 15 percentage point gain in end-to-end task success (from 75% in a multimodal baseline system to 90% in an improved configuration identified through a prior ablation study) is sufficient to produce consistent and measurable differences in user perception. The baseline system combines Whisper for speech recognition, Florence-2 for open-vocabulary object detection, LLaMA 3.1 for action extraction, and an interval Type-2 fuzzy logic controller for motion execution. The improved configuration replaces the perception and language modules with Grounding DINO + SAM and Qwen 3.5 9B, respectively, while retaining the same controller. A within-subject user study with 24 participants compared both systems on the same tabletop object-grasping task. After interacting with each configuration, participants rated perceived speed, reliability, and overall competence and fluency on a 7-point Likert scale. Results show that 17 out of 24 participants (70.83%) preferred the improved system (exact binomial test, p = 0.043, h = 0.43), and all three perceptual constructs were rated significantly higher for the improved configuration after Holm correction, with large to very large effect sizes (p < 0.001). These findings confirm that the identified technical improvements are perceptible to users in direct interaction and underscore the importance of complementing benchmark evaluation with user-centred evidence when assessing robotic manipulation pipelines.

Generative AI now enables games to produce dialogue, quests, characters, images, and worlds at runtime. Yet generation alone does not make a game AI-native, nor does it guarantee playability. This paper defines AI-native games by whether runtime generative AI is constitutive of the core loop: if the AI component were removed or trivially replaced, the central form of play would collapse or become fundamentally different. This counterfactual criterion separates AI-native games from AI-augmented games, boundary artifacts, chatbots, tavern-style role-play, procedural content generation, and AI-assisted production. Using this definition, we screen candidate artifacts and analyze 53 publicly available AI-native games and prototypes. We introduce a dual-axis G/N taxonomy: the G-axis captures player-facing game type, while the N-axis captures the dominant AI mechanic that makes generative AI indispensable to play. The corpus is concentrated around language-forward designs, especially narrative adventure, epistemic interaction, and generative narrative, while categories such as semantic adjudication, multi-agent simulation, generative construction, and relationship/companion play remain less represented. We argue that the central design problem is organizing semantic openness into stable gameplay. AI-native design depends on mechanical invariants: goals, rules, state, feedback, pacing, and player agency that make open-ended AI outputs interpretable and consequential. We conclude with a roadmap for controllable generation, AI-as-mechanic design, multimodal and multi-agent systems, inference economics, evaluation, safety, and regulation.

Artificial intelligence (AI) is becoming ubiquitous, and across domains, increasingly autonomous systems are carrying out tasks which raise significant ethical and legal challenges which demonstrate a need for strong human-machine teams rooted in trust. In this article, I argue that within highly impactful areas (such as medicine or warfighting) there are grounds for us initially treating autonomous and opaque systems as relevantly analogous to dogs (or other animals with which we have close relationships). Under this analogy, humans making use of these systems are not to be viewed as "users" or "deployers" of these systems, but instead take the role of "handlers". This recasting of roles shifts the way we view humans, AI-enabled and autonomous systems, and the relations between them, and moreover clarifies the clear and traceable lines of responsibility humans have for the outcomes brought about when using these systems. In developing this point, I clarify that the machine-animal analogy does admit disanalogous elements, but that its touch-points ground it as a starting point. I then explore how we can divest the humans-as-handlers approach of those aspects of our relationships with animals which are unfitting for how we engage with and make use of autonomous and AI-enabled systems. I conclude by arguing that the trajectory of human-machine teamings for autonomous and AI-enabled systems should be a state where we authentically view these not as artifacts which we simply make use of, but as collaborators with which we pursue complex goals and carry out complex tasks.

Empirical quantum-software papers often report that one compiler, optimizer, backend, or ansatz outperforms another. Such comparisons are not properties of a tool alone: they can change with benchmark scope, circuit construction, compilation, sampling, backend or noise assumptions, optimizer choices, and resource budgets. Existing testing, benchmarking, and reproducibility methods help assess programs, tools, executions, and platforms, but they do not directly audit whether the reported comparison itself is supported by the evidence exposed in the source paper or accompanying materials. We present CLAIMSTAB-QC, a source-bounded framework for auditing empirical comparisons in quantum software. Given a reported comparison, the framework records the baselines, metric, relation, and admissible evidence; locks the comparison design before outcomes are computed; and reports either a scoped relation outcome or an explicit evidence boundary. For strict scalar-directional comparisons, the reported direction is classified as Sustained, Unresolved, or Reversed within the locked audit scope. We evaluate CLAIMSTAB-QC on 455 comparative claims from 119 quantum-software papers. The central finding is a materialization gap: 175 claims can be represented for audit planning, 79 become scalar-directional planning records, 53 yield lockable audit or diagnostic designs, and only 8 expose enough matched evidence to audit the original comparison without proxy reconstruction. These 8 records yield 2 Sustained, 4 Unresolved, and 2 Reversed outcomes. Controlled diagnostics over 24 benchmark-relevant comparisons further show that simpler checks can preserve apparent directions whose support weakens under locked audit designs.

Automatic understanding of dynamic 4D point clouds, the 3D-point sequences captured over time by depth sensors and LiDAR, is central to robotics and embodied perception. Yet annotating them densely is expensive, making self-supervised pretraining the natural route to transferable representations. Existing pretext tasks, however, are almost entirely intra-modal, and the few methods that transfer knowledge from 2D foundation models rely on a single global embedding per clip, discarding the rich per-patch semantics that these models compute. To address this gap, we propose Cross4D-JEPA, a teacher-student method that distills a frozen 2D foundation model, an image model DINOv2, or a video model V-JEPA 2, into a 4D point encoder. The proposed method combines (1) a dense cross-modal correspondence that maps every 3D point to the teacher patch feature it projects to, and (2) a per-point objective that trains the student to match these features in latent space with no masking, negatives, or decoder. We evaluate Cross4D-JEPA on four benchmarks, MSR-Action3D, DeformingThings4D, NTU-RGB+D 60, and HOI4D, against intra-modal and global cross-modal baselines. Experimental results show that, under a matched protocol, the proposed method consistently outperforms intra-modal and global cross-modal baselines across the four benchmarks and is competitive with heavier published 4D methods; further analysis attributes this gain primarily to the granularity of the correspondence rather than the teacher modality. Beyond recognition accuracy, the dense representation learned by Cross4D-JEPA transfers across domains, improves label efficiency, and improves full-label fine-tuning under the same training budget, while a 13x smaller encoder matches a heavyweight pooling backbone.

Structural equation modelling (SEM) is widely used in survey-based business and information systems research to assess latent constructs and theory-driven structural relationships. However, SEM path significance is obtained within a particular model specification and may not show whether findings remain stable under alternative estimation frameworks. This study develops and demonstrates a staged robustness analysis framework that connects SEM, ordinary least squares (OLS) regression, and Double Machine Learning (DML). SEM is first used to refine the measurement structure and estimate the robustness-baseline SEM model, in which the full theory-specified structural path system is retained for downstream robustness analysis before final structural path evaluation. OLS regression is then applied to SEM-derived construct scores as a transparent regression benchmark. Finally, DML-style residualisation is used to examine whether each tested focal relationship remains stable after flexible machine-learning-based adjustment for observed controls. Learner-sensitivity checks compare Random Forest, Gradient Boosting, and Support Vector Machine learners, and selected reverse-direction diagnostics are used to examine directional sensitivity. The framework is demonstrated using a FinTech Digital Customer Intimacy survey model. The findings identify which relationships are stable across SEM, OLS, and DML-style checks, and which require more cautious interpretation. A reproducible Google Colab workbook and generated result files are publicly available, providing a reusable template that researchers and students can adapt to other survey-based latent-construct studies. The paper contributes a practical robustness workflow and interpretation guide for survey-based researchers seeking to complement SEM with conventional and machine-learning-based robustness checks.

Mutation testing is a powerful technique for ensuring software quality. However, the presence of equivalent mutants introduces unnecessary costs and biases, limiting its practical effectiveness. Although numerous equivalent mutant detection (EMD) methods have been proposed, they often face distinct challenges: pure-code analysis methods can be limited by their reliance on specific compiler infrastructures, while existing machine-learning approaches remain constrained by scarce training data and limited generalization to unseen mutants. Large language models (LLMs) have recently demonstrated remarkable performance across diverse code-related tasks by better capturing program semantics. Yet their potential for EMD remains largely unexplored, particularly in the multi-lingual context. This paper presents the first comprehensive empirical study on LLMs for EMD, using 3,302 Java and 1,088 C mutant pairs to benchmark against state-of-the-art methods, explore strategy variations, assess efficiency, and evaluate cross-lingual generalization. Experimental results show that LLM-based approaches achieve higher F1-scores than the evaluated traditional methods, with fine-tuned code embedding yielding the highest detection accuracy among the tested strategies. Moreover, LLM-based approaches strike a practical balance between effectiveness and efficiency with inference times comparable to existing machine-learning models. Importantly, fine-tuned LLMs demonstrate measurable generalization across programming languages. These findings establish LLMs as a viable and efficient approach for tackling the longstanding challenge of equivalent mutant detection, offering new directions for advancing mutation testing in practice.

Knowing which training examples drive outputs is fundamental to auditing, correcting, and understanding language models, yet for modern LLMs this remains expensive, approximate, and largely post-hoc. Standard language models generate tokens through a dense network pathway, causing training data's influence to be distributed across parameters rather than organized along explicit, traceable components. We introduce a prototype language model architecture, Prototypes for Interpretable Sequence Modeling (PRISM), that forms each prediction via a sparse, non-negative mixture of learned prototypes, trained with clustering objectives that anchor each prototype to coherent neighborhoods of training examples. Across architectures from 130M to 1.6B parameters trained on up to 50B tokens, prototype language models either surpass or remain within 2.5 percentage points on average downstream accuracy of matched dense baselines. We show that sparse prototype structure localizes curvature in the loss landscape, yielding a more tractable Hessian and enabling training data attribution that is ~500x faster than post hoc baselines when consuming equivalent memory. Calibrating linear prototype controllers can improve downstream accuracy by roughly 3 points while tracing those corrections back to training neighborhoods, and targeted prototype suppression can remove model behaviors without finetuning or measurable loss in generation quality.

While long-horizon mobile GUI agents typically rely on thought-action-observation loops, they struggle to separate persistent task states from transient screen observations. As execution histories grow, this entanglement imposes a severe context burden, causing agents to forget initial requirements, hallucinate progress, or repeatedly interact with stale interfaces. To address this, we introduce Task-State Representation (TSR), a training-free framework that explicitly decouples task state from sensory input. Acting as a lightweight external wrapper, TSR maintains three structured components: a global instruction summary, a dynamic progress tracker for subgoals, and a transition-aware action verifier. By continuously updating through pre- and post-action visual comparisons, TSR effectively guides the agent's reasoning without requiring architectural modifications. Experiments across four mobile GUI benchmarks validate TSR's effectiveness, yielding up to a 12 absolute point increase in success rate on complex cross-application and memory-intensive tasks.

We present BaseRT, a native Metal inference runtime for large language models (LLMs) on Apple Silicon, and report the highest inference throughput on this hardware to date. Existing runtimes, including llama.cpp and MLX-based frameworks, incur overhead from abstractions not designed for Metal's execution model or Apple Silicon's unified memory topology. By building natively on Metal with chip-specific kernel fusion, unified memory-aware optimisation, and custom dispatch logic, BaseRT recovers performance that framework-based approaches leave on the table. BaseRT supports a wide range of model families across eight quantisation formats (Q2 to FP16) on all Apple M-series devices. In this paper, we evaluate the Qwen3, Llama 3.2, and Gemma 4 families at Q4 and Q8 quantisation on M3 and M4 Pro devices. BaseRT achieves up to 1.56x higher decode throughput than llama.cpp and up to 1.35x higher than MLX, with substantially larger margins on prefill for mixture-of-experts models, delivering consistent best-in-class throughput from sub-1B to 30B parameter models. These results establish Apple Silicon as a more capable inference platform than previously reported, with direct implications for the emerging edge inference paradigm: as privacy requirements, latency constraints, and cloud cost pressures drive inference toward on-device deployment, performance-optimised local runtimes are a critical enabling layer for this transition. BaseRT is publicly available at https://github.com/basecompute/baseRT

Benchmarks for vision-language models (VLMs) mostly test observational spatial reasoning: models describe relations already visible in the input. Existing what-if tasks typically vary the observer while keeping the scene fixed. Can VLMs instead predict the consequences of hypothetically moving or rotating an object? We introduce MindEdit-Bench, a benchmark of six spatial reasoning tasks built from three-photo smartphone triplets of newly captured indoor scenes via an automatic in-the-wild 3D scene-graph extraction pipeline. Four tasks probe perception and perspective transformation over observed structure; two new tasks, L4 (spatial editing) and L5 (cross-view visibility editing), probe object-level counterfactual reasoning, where correct answers are absent from all input images. Each question provides 8-24 structured answer choices, enabling answer-letter-level diagnosis of spatial and fallback errors. The benchmark covers 120 private indoor scenes not drawn from public datasets, reducing public-data pretraining-overlap risk. Across 15 VLMs on 1,003 human-verified questions, task-wise mean VLM accuracy is only 8%-31%, versus 81%-97% human majority-vote accuracy. The pooled human--best-VLM gap is 53 pp, with at least 39 pp on every task. The structured answer space further reveals non-uniform failures, including weaker camera-depth-axis inference and fallback behavior on difficult visibility-editing cases.

Diffusion models are highly effective at modeling complex data distributions, including images and text. However, in applications like personalized recommender systems, the objective often shifts to modeling specific regions of the distribution that maximize user preferences-initially unknown but gradually uncovered through interactive feedback. This can naturally be framed as a reinforcement learning problem, where the goal is to fine-tune a diffusion model to maximize a reward function based on preferences. However, the main challenge lies in learning a parameterized reward model, which typically requires large-scale preference data-something that is often not feasible in practice. In this work, we introduce Personalized Active Preference Alignment PAPA, a novel method that bypasses the requirement for a parametrized reward model by directly optimizing the diffusion model using real-time user feedback. PAPA enables feedback-efficient preference alignment, drawing inspiration from the variational inference framework. We demonstrate PAPA's effectiveness through extensive experiments and ablation studies across diverse class-conditioned and fine-grained alignment tasks. Additionally, based on theoretical insights, we propose an enhanced fine-tuning strategy, referred to as EPAPA, that requires less computational budget and accelerates the fine-tuning process, further boosting PAPA's suitability for real-world deployment. Our code is made publicly available at https://github.com/NasikNafi/papa.

Large reasoning models (LRMs) have achieved strong reasoning capabilities in English, yet their performance degrades significantly when required to reason in other languages. A natural solution is to transfer the model's English reasoning ability to target languages. However, existing transfer approaches typically rely on distilled target-language reasoning traces from stronger LRMs or online supervision from external judge models, which are costly and difficult to scale. In this paper, we propose PCS (Progressive Code-Switching), a more efficient transfer framework that requires only lightweight translation without any stronger model for distillation or judging. PCS first constructs code-switched reasoning traces by translating a subset of English reasoning steps into the target language, and uses them to initialize the model's code-switching ability via supervised fine-tuning. It then applies reinforcement learning with a step-level language consistency curriculum, progressively raising the target-language ratio until the model reasons entirely in the target language. This progressive design provides a smooth transfer path that avoids the instability and performance degradation commonly observed when directly enforcing target-language reasoning. Experiments on multiple benchmarks and five typologically diverse languages show that PCS substantially narrows the performance gap between target-language and English reasoning, yielding more language-consistent reasoning while maintaining competitive accuracy.

Reasoning language models frequently overthink: generating extended chains of behaviors such as hedging, approach abandonment, and self contradiction that consume tokens without improving answers. We show that these behaviors are not merely a consequence of length; even when controlling for response length, incorrect traces exhibit higher rates of unproductive self-reflection than correct ones. Addressing this requires identifying where self-reflection helps vs hurts, but obtaining these step-level annotations is costly. We observe that intermediate answer commitments within reasoning traces can provide a cheap proxy: by comparing each final answer candidate in the trace to the ground truth, we can determine whether subsequent reflection is productive without any additional supervision. Building on this insight, we propose DASH (Drift Aware advantage SHaping), which assigns segment-level credit based on whether each reasoning segment leads toward or away from correctness. On competition-level math benchmarks, DASH achieves the highest accuracy where overthinking is prevalent (AIME25: 50.8% vs. 45.4% GRPO) while reducing overthinking behaviors and achieving more productive self-correction than baselines.

Jailbreak attacks remain a critical threat to the safe deployment of large language models (LLMs). While prior work has primarily studied attacks and defenses at the prompt level, we show that this prompt-centric paradigm overlooks a structural vulnerability in stateful, function-calling environments. In such applications, developer-defined schemas, structured arguments, and untrusted tool outputs are interleaved into a single shared model context. This architecture expands the attack surface by blurring the boundary between trusted control logic and untrusted data, allowing adversarial intent to be distributed across a multi-turn execution path. We exploit this architectural flaw through SMT, a black-box attack framework based on Simulated Moderation Traces. Departing from purely prompt-based interactions, SMT constructs a multi-turn trajectory that simulates a legitimate moderation-auditing workflow. Within this trajectory, a fabricated moderation frame leverages red-team testing as a pretext to elicit harmful generations. The subsequent validation feedback treats safety refusals as execution failures, prompting refinements that gradually weaken the model's safety constraints and ultimately trigger harmful outputs. Extensive empirical evaluations on prominent commercial LLMs from five different providers across two standardized safety benchmarks show that SMT consistently achieves the highest average attack success rate and HarmScore while requiring a near-minimal number of queries, substantially outperforming existing baselines. These findings demonstrate that prompt-level sanitization alone is fundamentally insufficient for defending tool-enabled LLM systems and highlight the urgent need for context-aware validation across schemas, arguments, tool outputs, and accumulated conversation state. The code is available at https://github.com/liujlong27/SMT.

Transformer-based large models have demonstrated remarkable generalization abilities across different tasks by leveraging a context-aware attention module for in-context learning. With richer context, transformers adapt more effectively to the current use case without any parameter updates. However, the quadratic computational and memory complexity with respect to context length significantly slows data processing in softmax transformers. Linear transformers were proposed to address this issue by reducing the complexity to linear dependence on context length, but the design and understanding of the feature mapping in linear attention, from a theoretical viewpoint, remain unclear. In this paper, we investigate the approximation and generalization abilities of linear transformers under a two-staged sampling process from domain generalization. We show that linear transformers perform in-context learning as learning a mapping from context distributions to response functions. A dimension-independent convergence rate is obtained for our generalization analysis, which also exhibits the tradeoff between the regularities of data distributions and latent features. Guided by our theoretical framework, we propose a new perspective on activation and loss design for linearizing pretrained softmax large language models.

A total of seven categorical encoding methods were tested on the IEEE-CIS fraud benchmark dataset (590,540 records, 3.5% positives, 8 high-cardinality columns). The encoders were evaluated using a stratified 5-fold cross-validation (CV) with three repetitions. Five of the encoders had identical frozen LightGBM learners in the downstream phase, allowing for controlled comparisons of their performance to each other. CatBoost and TabNet were included as comparisons across paradigms using different learners. The entity embeddings produced the highest AUC-ROC (0.9612), with a statistically significant tie with that of CatBoost (0.9602) and statistically superior to tier group encoding (0.9548), whereas target encoding was only 0.0023 worse than tier group encoding and the auditor-friendly tier boundaries were maintained. Off-the-shelf TabNet did not outperform tree-based pipelines and collapsed under data scarcity. On AUC-PR, CatBoost leads (0.822 vs. 0.793); no encoder dominated both metrics. Per-column analysis confirmed the embedding advantage arises from joint multi-column representation.

How many days of early behavior suffice for subscription churn prediction? In the public KKBox dataset, the early indicator of churn is typically an indicator of someone's contract status; however, when looking in the heavily churned manual-renewal segment, having access to early behavior creates a substantial increase in prediction for that specific segment (PR +0.10 at 120 days). A nine-window sufficiency curve shows a diminishing-returns knee in a 45-90 day band. However, stress-testing over three cohort/task designs shows that this curve is singular to the design being tested; for example, in our test with a moving target, the curve inverts and can shift depending on the feature set used. Therefore, any window-sufficiency claim should state its cohort construction, target definition, and feature families. All evidence is from one music-streaming dataset; the mechanism should generalize but the magnitudes may not.

Cardiovascular disease is still one of the main causes of death around the world. Acute myocardial infarction (MI), or heart attack, claims millions of lives each year. MI happens when blood flow to the coronary arteries is blocked or reduced, which causes permanent damage to the heart muscle. Without treatment, this can lead to cardiac arrest, where the heart stops pumping blood to the organs, resulting in organ failure and death. Even survivors often face serious problems like heart failure, pulmonary edema, and asystole. Research shows that 5 to 10 percent of survivors die within the first year after an MI, and nearly half need to be hospitalized again. Early thrombolytic treatment leads to better outcomes, so there is a clear need for faster and more accurate ways to diagnose MI. Right now, doctors usually review patient history and use their own experience to find the causes of MI. This process takes a lot of time and can be inconsistent. Detecting MI accurately and quickly can help patients take better care of themselves and prevent fatal events. In this study, we introduce an automated model to predict deadly outcomes of MI and help doctors understand important biomarkers linked to its complications. This approach aims to make diagnosis clearer, faster, and more affordable. The process includes preparing the data, filling in missing values, and handling imbalanced data using SVMSMOTE, ADASYN, and class-weighted methods. We use wrapper and embedded feature selection to find the most important variables, then scale the features for consistency. The model combines Logistic Regression, Random Forest, Light-GBM, and Bagging SVM, and is further improved with an artificial neural network to increase accuracy. We evaluate all models using precision, recall, and other key measures to find the best option for clinical use.

Low Earth orbit (LEO) satellite constellations are emerging as a backbone for global 6G connectivity, where independent tenant slices share orbital infrastructure, each requiring an ordered chain of security virtual network functions (VNFs). Because onboard computation and networking are scarce, slices cannot be given dedicated VNFs. They must share instances on the same satellites, enlarging the attack surface and exposing tenants to cross-slice side-channel risk. This exposure shifts continually as visibility, orbital motion, and the inter-satellite topology change in time (epochs), making VNF migration a structural necessity that couples resource efficiency, service continuity, and security isolation into a single problem. We formulate this security- and migration-aware security function chain (SFC) placement as a multi-slice mixed-integer linear programming (MILP) whose core is a co-location risk model, grounded in ISO/NIST principles and supported by analytic bounds, in which we separate avoidable migrations from those forced by orbital motion. Because the joint program scales quadratically with the cross-slice co-location terms, we develop an alternating direction method of multipliers (ADMM)-inspired penalized per-slice best response decomposition that recasts the coupling as a linear per-slice penalty, yielding independent subproblems through sequential (S-ADMM) and parallel, collision-repaired (P-ADMM) schedules. Simulations over a Walker-Delta satellite constellation show that the proposed framework eliminates co-location risk, reduces SFC migrations, and sustains full delay compliance, while remaining feasible within the per-epoch budget for slice counts where the monolithic security-aware MILP is intractable.

We investigate a forecasting framework based on a simple discrete-time dynamic model with coefficients varying in time. The parameters of the model are recovered within a deep learning framework, which makes it possible to retain a transparent parametric structure while simultaneously accounting for complex and nonstationary patterns in the observed phenomenon. Our analysis covers two specifications of the noise process. Besides the standard Gaussian setting, we also consider Laplace-distributed noise, which can offer a more adequate description in the presence of heavier tails and sharper local fluctuations. For both cases, we formulate the predictive scheme of the model and analyze the associated uncertainty quantification, including the construction of prediction intervals. The results illustrate that a relatively simple model, when combined with time-dependent parameter estimation, can serve as a mathematically tractable and practically flexible tool for forecasting complex dynamics under different noise assumptions. The general model is stated for TVAR($p$), while the prediction-interval formulas and the numerical experiments are developed for the TVAR(1) case.

In prefill-decode (PD) disaggregated LLM serving, each request is assigned to a decode worker after prefill. Existing decode routers balance only load; for mixture-of-experts (MoE) models this is incomplete: equally loaded workers can differ in latency, since each decode step loads the weights of every distinct expert its batch activates. We present ELDR, an expert-locality-aware decode router for PD-disaggregated MoE serving. From a request's prefill expert activations, ELDR builds an expert signature predicting the experts it will activate during generation. Offline, balanced K-means partitions signature space across decode workers; online, locality-band routing sends each request to the least-loaded worker among those best matching its signature. A signature cache, co-indexed with the KV cache at KV-block granularity, keeps signatures exact under prefix caching. Implemented in vLLM and evaluated on deployments of up to 40 GPUs, ELDR reduces median TPOT by 5.9-13.9% over the strongest of four load-balancing baselines across three MoE models and two workloads, with model outputs unchanged.

Large Vision-Language Models (LVLMs) rely extensively on Visual Instruction Tuning (VIT) to elicit their multimodal reasoning capabilities. However, we find a discrepancy: VIT often packs multiple language tasks about the same image for conversational, multi-turn training, whereas existing benchmarks evaluate LVLMs in isolated, single-turn scenarios. The models can suffer from visual attention decay and contextual overfitting during multi-turn training, making it hard for them to realize their full potential in the mismatched test phase. To close the gap, we propose learning with Stochastic Turn Depth (StochasT), which stochastically groups language tasks for the same image into clusters of varying sizes (turn depth) while preserving their organic order. Hence, while StochasT draws on Dropout and stochastic depth for ResNets, it does not actually drop anything to maximize the utility of the training data. Furthermore, we introduce a challenging, benchmark-agnostic evaluation mechanism based on the Balanced Latin Square to measure LVLMs' robustness under varying contextual dependencies. Extensive experiments demonstrate that StochasT effectively grants LVLMs strong, harmonized capabilities for both single-turn and multi-turn use cases.

Current molecular generation benchmarks emphasize task complexity, molecule novelty, and property alignment; they largely overlook a critical concern: the potential safety risks of AI-generated molecules. In practice, many generative models may produce molecules with toxic, reactive, or otherwise hazardous characteristics - posing hidden dangers that remain insufficiently addressed. To address this gap, we introduce MolSafeEval, a benchmark dedicated to evaluating and analyzing the safety risks of molecular generation. Unlike prior approaches that rely on narrow toxicity predictors, MolSafeEval integrates heterogeneous safety knowledge - ranging from toxicological databases to hazard rules - into a structured molecular safety knowledge graph. This graph serves as a foundation for large language model-based reasoning, enabling systematic detection and explanation of unsafe features in generated compounds. We further categorize molecular generative models into four representative task types - unconditional generation, property optimization, target protein-based design, and text-based generation - and provide standardized datasets and safety evaluation protocols for each. By systematically revealing the safety vulnerabilities of current generative approaches, MolSafeEval offers a new lens for benchmarking molecular models and provides essential guidance toward safer, more trustworthy molecular design.

Predicting complex spatiotemporal dynamics in physical processes often demands computationally expensive numerical methods or data-driven neural networks that suffer from high training costs, error accumulation, and limited generalizability to unseen parameters. An effective approach to address these challenges is leveraging physics priors in training neural networks, known as physics-informed deep learning (PiDL). In this work, we introduce the Multi-Resolution Finite-Volume-inspired network, MuRFiV, designed to capitalize on the conservative property of finite volume on the global scale and the expressive power of deep learning on the local scale. We demonstrate the effectiveness of MuRFiV on several spatio-temporal systems governed by partial differential equations (PDEs), including Burgers' equation, shallow water equations, and incompressible Navier-Stokes equations. By embedding PDE information into the deep learning architecture, MuRFiV achieves strong long-term prediction accuracy and remains stable over very long autoregressive rollouts, significantly outperforming data-driven neural network baselines. This result highlights the promise of combining multiresolution learning with finite-volume-inspired inductive bias for accurate and robust long-term prediction of complex dynamics.

Embodied agents operating in the real world require multi-scale reasoning and knowledge adaptation as conditions change. We identify two challenges in applying Mixture of Experts (MoE) to this setting: routing lacks an explicit notion of scale, preventing targeted updates at specific scales, and a uniform update policy cannot accommodate the different rates at which knowledge at each scale becomes outdated. We present MuSix, a framework that addresses both challenges through scale-aware world model mixture and evolution. A two-stage routing mechanism grounds scale selection in experiential distance, a measure of situational novelty inspired by Construal Level Theory: a meta-router first maps this quantity to a weight over continuous scale space, then per-scale base routers select world models within the identified scale. For adaptation, scale-dependent forgetting rates allow low-scale knowledge to refresh rapidly while high-scale abstractions persist, and gated inter-scale transfer maintains coherence across the hierarchy. Experiments on EmbodiedBench and HAZARD show that MuSix improves over state-of-the-art baselines on multi-scale reasoning and dynamic adaptation.

Purpose: Cloud computing environments are highly dynamic, creating major challenges for resource management. Accurate workload prediction is therefore essential for effective auto-scaling. To address this, we present CloudyGUI, a Python simulation framework with an easy-to-use GUI that allows researchers to test and validate resource management strategies. Methods: This framework employs a three-stage pipeline: workload generation, prediction (utilizing XGBoost and LSTM), and an auto-scaling system based on the MAPE loop. Validation includes internal, intermediate, and external methods to ensure system reliability. Results: CloudyGUI's generated workloads closely match real-world datasets. A two-sample K-S test confirms this alignment, showing strong p-values of 0.19 for CPU and 0.14 for memory. When compared to a command-line tool, the GUI adds only a minimal overhead of 1.4x-4.67x. Furthermore, expert review validates the tool's realism and practical usefulness. Conclusion: CloudyGUI fills a critical gap by providing an accessible and efficient platform for simulating auto-scaling in cloud applications, helping researchers develop advanced cloud management solutions.

Agricultural advisory systems face a fundamental tension: static agronomic guidelines offer consistent, evidence-based recommendations, yet remain blind to in-season variability and dynamic uncertainties. Recent advisory systems powered by LLMs are liable for a different risk of generating recommendations that are agronomically credible but physiologically unconvincing. Agri-SAGE is a closed-loop framework designed to resolve the above two limitations by integrating retrieval-grounded multi-agent LLM reasoning with APSIM-based biophysical simulation, to generate and validate agronomic advisories. To assess this framework, we evaluate three reasoning approaches, namely Plan-and-Solve, Tree of Thoughts, and Reflexion, over a 10-year retrospective analysis. All three significantly outperform static PoP (Package-of-Practice) baselines, with Tree of Thoughts achieving impressive peak yields. At the same time, Reflexion achieves comparable agronomic outcomes at substantially lower computational cost by leveraging cross-seasonal episodic memory.

Actor-critic methods depend on learned critics, but critic quality is often evaluated only indirectly through return, temporal-difference error, or value loss. Critic complexity is introduced as an additional diagnostic and intervention dimension for actor-critic reinforcement learning. The analysis uses spectral effective-rank entropy, a rank-like summary of the singular-value distributions of critic weight matrices, to assess critic model complexity. Across TD3 and PPO experiments, critic complexity is tracked together with return and Monte Carlo value-estimation bias. The results show that critic complexity is measurable throughout training and is systematically associated with training behavior, while also making clear that the relationship is heterogeneous across algorithms, tasks, and hyperparameters. A direct complexity-control intervention is then evaluated by adding a spectral-entropy penalty to the critic loss. This intervention reliably changes the targeted spectral quantity, demonstrating that critic complexity can be controlled rather than only observed. Return effects are treated as task-dependent evidence rather than as a general performance claim, because overall complexity-control results vary.

The two-tower model has been widely used for large-scale recommendation systems, particularly in the retrieval stage. Industry standards for training two-tower models typically involve in-batch and/or out-of-batch negative sampling. However, these methods often produce easy negatives that models can quickly learn, failing to sufficiently challenge the model. To address this issue, a novel self-supervised hard negative sampling technique is proposed that leverages a large language model (LLM) to generate hard negatives from the same cluster during model training. By utilizing the LLM to learn media representations, the proposed approach ensures that the generated negatives are more challenging and informative. This real-time sampling framework is designed for seamless integration into production models, capable of handling billions of training data points with minimal computational complexity. Experiments on public datasets, along with deployment to a large-scale online system, demonstrate that the proposed negative sampling technique outperforms widely used industry methods. Furthermore, analysis in industrial applications reveals that this sampling method can help break inherent feedback loops in recommendations and significantly reduce popularity bias.

Large language models often produce hallucinated answers that violate prompt-level constraints. A key diagnostic question is whether these failures reflect missing knowledge, or whether the model has the relevant information but follows the wrong inference path. We study this phenomenon as inference misalignment: a mismatch between the answer supported by the prompt and the answer favored by statistically salient latent associations. We formalize this view with a latent key-task model, in which pretraining-frequency imbalance can cause a shortcut path to dominate the constraint-sensitive path and induce positive inference loss. The framework predicts two failure modes: task-retrieval bias in entity disambiguation and key-selection bias in action choice. We introduce TrapQA, a controlled diagnostic testbed with two components. ScientistQA tests disambiguation among similar scientists with supplementary factual probes, while Real-Life Constrained QA tests everyday constraint following under salient shortcuts. Our results show that hallucination can arise from biased latent inference rather than absent knowledge alone.

As video corpora continue to expand in both scale and task complexity, there is increasing demand for approaches that retrieve relevant videos from large-scale corpora (inter-video reasoning) and subsequently perform fine-grained, query-conditioned tasks (intra-video reasoning) within the retrieved content, such as temporal grounding. However, existing approaches typically treat retrieval as a preprocessing step, and consequently, when the initial retrieval fails, there is no mechanism to refine the search, leading to the failure of subsequent fine-grained intra-video reasoning. Moreover, while recent agentic frameworks have advanced video understanding, they typically assume that the query-relevant video is already given, focusing exclusively on intra-video reasoning tasks. To address these limitations, we propose VideoSearch-R1, an agentic framework for iterative video retrieval and reasoning through multi-turn interaction with a video search engine. Specifically, we introduce Soft Query Refinement (SQR) to refine search query tokens in a continuous latent space rather than rewriting queries in the discrete text space, enabling more efficient and fine-grained adjustments. SQR and its reasoning process are trained using Group Relative Policy Optimization (GRPO), guided by task-level reward signals derived from retrieval and downstream tasks. Building upon this, VideoSearch-R1 achieves state-of-the-art performance across three datasets on Video Corpus Moment Retrieval (VCMR), iteratively retrieving videos from large-scale corpora, refining search queries, and performing precise query-conditioned temporal grounding within the retrieved content. Our analyses show that SQR effectively refines the original query, requiring significantly fewer generated tokens than explicit text-level query refinement. Code and model checkpoints are publicly available at mlvlab.github.io/VideoSearch-R1.

Spatiotemporal motion planning, especially in multi-robot settings, requires robots to reason about collision-free regions that change over time, which is challenging in continuous spaces when feasible regions are transient and geometrically constrained. We present an algorithmic framework based on graphs of space-time convex sets (ST-GCSs), where collision-free regions are represented as convex sets in space-time and trajectories correspond to paths on the graph together with continuous motions within the selected sets. We formulate time-optimal planning on ST-GCSs as a graph-search problem over path-indexed states and develop a best-first search solver that evaluates partial paths via continuous trajectory optimization, guided by admissible heuristics and dominance checks. We further present an Exact Convex Decomposition (ECD) scheme to reserve trajectory occupancies in space-time, enabling unified handling of dynamic obstacles and multi-robot interactions. For multi-robot motion planning, we integrate ST-GCS planning and ECD into prioritized planning methods and introduce a windowed coordination scheme to improve efficiency. Extensive experiments on single-robot and multi-robot problems demonstrate substantial speedups over various planners while maintaining high solution quality, particularly in environments with narrow and transient feasible regions. Large-scale demonstrations further show that the proposed multi-robot motion planner can solve instances with up to $100$ robots within only a few minutes. Project homepage: https://sites.google.com/view/stgcs

Reinforcement learning (RL) for quadruped locomotion commonly depends on fixed, hand-crafted, and Markovian reward functions that limit both interpretability of learned policies and lack explicit control over gait behaviors. We introduce a framework where distinct gaits are specified using parameterized constraints expressed in Signal Temporal Logic (STL). These include safety bounds, gait synchronization constraints, command tracking, and actuation bounds. From these specifications, we develop a reward shaping mechanism that provides learning agents a dense, continuous reward landscape that encodes desired behavior. We define parametric STL templates for three speed regimes (walking-trot, trot, bound), calibrate their parameters from reference rollouts, and compute rewards from using smooth approximations of STL robustness over the rollouts. The generated rewards can be used to provide shaped gradients compatible with Proximal Policy Optimization (PPO). We instantiate the approach on Google's Barkour quadruped robot in MuJoCo XLA (MJX). We use parallelization within the simulator to improve training speeds and use domain randomization to robustify learned policies. We show that compared to a baseline of hand-crafted rewards, the STL-shaped rewards yield tighter velocity tracking and more stable training. Videos can be found on our project website: https://stl-locomotion.github.io/.

In the current era of great-power competition and the diffusion of emerging disruptive technologies on the battlefield, NATO's approach to coordinating the development, adoption, and standardization of new technologies is changing from its practices during the Cold War, but the nature of these technologies poses additional challenges for the alliance.

Large language model agents are increasingly connected to scientific software, yet it remains unclear when tool access makes scientific computation more reliable rather than merely more complex. We introduce PHREEQC-MCQ-200, a benchmark for evaluating tool-augmented agents on deterministic aqueous-geochemistry simulations. The benchmark contains 200 multiple-choice questions derived from 21 validated PHREEQC scenarios, requiring agents to construct simulator inputs, execute PHREEQC, inspect structured outputs, and commit to final answers. Across multiple frontier and mid-tier model families, simulator access substantially improves aggregate accuracy, confirming that grounded execution is necessary for many scientific-computation tasks. However, the gains are not monotonic: tool-augmented agents also lose items they answered correctly without tools, revealing regressions that average accuracy alone hides. We further show that output-access protocol matters. A table-of-contents interface can reduce token cost while preserving or improving accuracy for stronger models, but it degrades performance for mid-tier models that cannot reliably navigate structured simulator outputs. PHREEQC-MCQ-200 therefore frames scientific tool use as an end-to-end diagnostic problem rather than a simple tool-calling capability. We argue that evaluations of scientific agents should report not only accuracy, but also item-level retention, output-access sensitivity, trajectory failures, and where the computation chain breaks.

Social coding platforms such as GitHub host millions of repositories, yet many suffer from high mortality rates. Despite this, several survival factors remain poorly understood. Human capital is widely recognized as essential. Social attention, while often assumed to be a lifeline, can become a liability. Structural features that improve onboarding, such as code readability and documentation, may also accelerate the cessation of active development when combined with massive visibility. To examine these dynamics, we analyzed more than 73,000 GitHub repositories using an Accelerated Failure Time (AFT) survival framework, which accounts for the time-varying nature of predictors. Our study identifies human capital as the most critical determinant of project survival. In contrast, excessive social attention emerges as a liability, and when coupled with accessibility features, it amplifies the risk of project inactivity. Importantly, when the number of contributors interacts with social popularity, the protective effect of labor becomes visible, highlighting the need for governance strategies that balance visibility with labor capacity to ensure the long-term resilience of open-source projects.

Vision-language models (VLMs) have become a paradigm for multimodal learning, yet remain unstable due to object hallucination, weak visual grounding, and catastrophic forgetting after full-parameter instruction tuning. We claim these failures result from a lack of explicit control over visual representation learning during the standard next-token prediction objective. As a result, visual embeddings thus become passively optimized and prone to injecting redundant or spurious signals. To counter this, we introduce Information-Regularized Attention (IRA), a stochastic attention mechanism that explicitly regulates the amount of visual information injected into the hidden states of intermediate transformer layers. This local reparameterization translates uncertainty about visual representations into local noise that is independent across data points. Beyond evaluating model performance, we also quantify embedding properties, where IRA produces smoother curvature trajectories and suppresses attention-sink across all layers, indicating a more stable transformation of the visual signal. Our results suggest that stochastic attention is not merely a regularizer but a key contributor to representation learning in a generative architecture, offering a new direction for building more reliable VLMs.

Forecasting models for health-signal digital twins must preserve the oscillatory, frequency, phase, and state-transition dynamics of physiological signals, yet the pointwise metrics used to benchmark them cannot detect when these fundamental properties are lost. We show that this blind spot misranks models: across 11 architectures, models with comparable pointwise error diverge by up to 53° in phase accuracy, equivalent to roughly 123 ms for a 1.2 Hz cardiac rhythm and invisible to standard metrics. To enable development of models that escape such failures, we introduce TimeSynth, a controlled benchmarking framework with two reusable components: a physiologically grounded generator producing signals with analytically known ground-truth dynamics from parametric models fitted to real electroencephalography, electrocardiography and photoplethysmogram signals, along with diagnostics quantifying amplitude, frequency, phase, and state-transition fidelity. Linear and full-sequence attention models systematically lose frequency and phase information despite acceptable amplitude error, whereas architectures with localized temporal structure better preserve dynamical fidelity and adapt to observable state transitions; none, however, reliably preserves stochastic switching. Because the dominant determinant of fidelity is architectural, model choice becomes a principled, use-case-driven decision rather than a search for a single winner. TimeSynth thus supplies the controlled preclinical stress test missing before models are coupled to patient data, with a reusable generator and diagnostics for fidelity-aware development.

Large language models (LLMs) are increasingly applied to requirements engineering (RE) tasks, yet the prompts guiding them are typically designed manually through trial and error, yielding inconsistent and suboptimal results. Automated prompt construction remains largely unexplored in RE, leaving its effectiveness unclear. To address this, we propose a lightweight Automatic Prompt Engineering approach, Backtracking APE (BT-APE), and apply it to requirements classification. We frame prompt design as an optimization problem, iteratively refining prompts via LLM-generated candidates, backtracking search, and dynamic example selection. Evaluating BT-APE on three benchmark datasets with five instruction-tuned LLMs, we compare it against four classical prompting baselines (zero-shot, few-shot, chain-of-thought, CoT+few-shot) and a state-of-the-art but resource-intensive APE baseline (PE2). BT-APE and PE2 achieve nearly identical accuracy, both substantially outperforming the classical baselines with large effect sizes; however, BT-APE imposes a far lighter computational footprint, consuming roughly 72% fewer input tokens and 66% less wall-clock time at equivalent accuracy, making it better suited to resource-constrained deployment. Our contributions are threefold: (i) a lightweight APE framework with an open interactive tool and replication package; (ii) the first systematic comparison of APE against classical prompting for requirements classification; and (iii) insights into how class definitions and prompt evolution affect performance.

Vision language models (VLMs) demonstrate strong zero-shot performance, but often perpetuate social stereotypes in person-centric queries, yielding skewed demographic distributions. Current debiasing methods apply uniform bias corrections across all input queries regardless of their bias sensitivity, creating a fundamental fairness--utility trade-off. Strong debiasing distorts semantically meaningful information in bias-insensitive queries, while weak debiasing fails to mitigate stereotypes in bias-sensitive ones. This one-size-fits-all approach hampers simultaneously achieving high utility on bias-insensitive queries and fairness on bias-sensitive queries. We introduce Reward-Gated Test-Time Adaptation (RG-TTA), a reinforcement learning-based test-time adaptation framework that selectively applies debiasing based on input sensitivity. RG-TTA adaptively triggers fairness regularization based on the bias sensitivity of each input during test-time policy adaptation, while focusing exclusively on optimizing cross-modal alignment for bias-insensitive inputs. Experiments on fairness benchmarks (e.g., FairFace, UTKFace) demonstrate substantial bias reduction while simultaneously improving zero-shot utility, resolving the trade-off of uniform debiasing.

This paper presents Speech Playground, an interactive speech visualization and comparison tool. While existing tools such as Praat are excellent, it can be cumbersome to integrate them with modern deep learning representations and use them for comparison. Speech Playground addresses this by combining a Python backend with a web-based frontend for interactive exploration of multiple feature types, including continuous, discrete, and variable-length representations. It includes TextGrid and forced alignment support together with configurable distance and alignment settings for visual and auditory comparison. Speech Playground is intended for use in speech research, representation validation, and computer-aided pronunciation training (CAPT)-oriented experimentation.

In the era of satellite constellations, multi-view optical satellite imagery is pivotal for Earth Observation (EO) and high-quality Digital Surface Model (DSM) reconstruction. Although feed-forward 3D foundation models have transformed computer vision, their deployment in satellite remote sensing is inherently constrained by the structural discrepancy between implicit perspective assumptions and explicit orbital pushbroom geometry. This geometric incongruity is further compounded by pronounced view-set heterogeneity. We present EO-VGGT, a framework that adapts a frozen perspective-driven model to orbital observations via explicit physical geometry embedding.First, the Geometry-Correlation Constrained Selection (GCCS) strategy prunes sub-optimal observations by balancing geometric diversity and radiometric consistency to optimize the input sequence. Second, a Sensor-Ray Encoder (SRE) parameterizes pixel-level pushbroom lines of sight derived from the Rational Function Model (RFM) into high-dimensional space-geometric tokens, reconciling the mathematical discrepancy between central projection and orbital kinematics. Third, a lightweight Ray-Pointing-Aware Adapter (RPAA) employs gated residual blocks to integrate these tokens directly into the frozen transformer backbone. Our findings underscore that integrating explicit physical geometry with optimized view selection is essential for robust feed-forward satellite 3D reconstruction.

Authority bias poses a critical safety concern in language models: models systematically prioritize social cues from authority figures over factual consistency, swaying their answers based on source credibility rather than evidence. We mechanistically investigate this phenomenon using a controlled medical QA setting, where hints suggesting incorrect answers are attributed to personas of varying expertise. Across Llama-3.1-8B, Qwen3-8B, and Gemma-2-9B, we find that models respond in a graded manner proportional to perceived authority, a hierarchy that is never explicitly prompted but emerges from training. Logit lens analysis and linear/non-linear probing localize this effect to a critical late layer where correct answer representations are actively erased, an erasure that scales with authority level, resists mean vector intervention, and is only partially reversible through chain-of-thought reasoning. Our findings suggest that authority-induced sycophancy is not a surface-level output bias but mechanistic knowledge erasure, a precise, layer-localized overwriting of correct internal representations by high-status authority signals.

Recent brain decoding studies have made substantial progress in reconstructing externally perceived visual content from neural signals. However, using electroencephalography (EEG) recordings to guide facial expression editing remains largely unexplored and poses a distinct challenge: rather than recovering what a subject sees, it requires identifying facial-action related patterns from noisy EEG signals and grounding them in localized, identity-preserving expression edits. In this paper, we investigate EEG-conditioned facial image editing for fine-grained facial action unit (AU) control and propose MindAU, a unified framework for controlling facial AU edits from EEG signals. MindAU first learns noise-robust and AU-discriminative EEG representations through temporal masked reconstruction and AU classification supervision. It then bridges the modality gap via Dual-Stream Manifold Alignment, aligning EEG features with AU-level text semantics and identity-reduced visual displacement trajectories in the multimodal space of Qwen2.5-VL. Finally, MindAU incorporates EEG-aware Multimodal Rotary Positional Embeddings, landmark-guided reference masking, and AU-aware region supervision into a multimodal diffusion-based editor for high-fidelity identity-preserving editing. We also introduce E-CAFE, a curated benchmark for EEG-Conditioned Action-Unit Facial Editing with paired EEG-face editing samples and standardized evaluation protocols. Extensive experiments demonstrate the effectiveness of MindAU and suggest its potential as a step towards future assistive expression technologies for individuals with facial neuromuscular disorders.

Slide design requires personalizing both deck themes and page layouts. Yet, current AI agent-based methods struggle with fine-grained, page-level design. Solely relying on prespecified templates or user verbose instructions, they fail to capture latent design intents, leaving Page-level Slide Personalization (PSP) unresolved. To close this gap, this work formulates PSP as an inverse planning problem. We propose to learn a design intent without assuming any knowledge of the specific executing tools (e.g., PowerPoint, Beamer) being used. However, relinquishing control over these tools makes the problem intractable to optimize end-to-end. To overcome this, we propose SPIRE, a principled framework to solve PSP approximately. By intentionally corrupting the visual structures of clean slides, SPIRE creates a verifiable task to denoise the corruption, whereby two agents learn to collaboratively refine executable designs via reinforcement learning (RL). We present a proof that structural denoising is a consistent surrogate for PSP, and that the multi-agent formulation strictly reduces policy gradient variance in RL. Extensive experiments demonstrate the superiority of SPIRE.

Safety alignment of text-to-image (T2I) diffusion models aims to suppress harmful generations while preserving utility on benign prompts. Recent methods often appear to deliver high safety with high utility, but this conclusion rests largely on coarse global utility metrics (e.g., FID, CLIPScore) that are insensitive to fine-grained semantic correctness, creating an illusion of high utility. We show that when utility is measured with structured evaluation, this illusion breaks: on TIFA (Text-to-Image Faithfulness evaluation with Question Answering), safety-aligned models suffer substantial drops in semantic fidelity, including failures in object counts, attributes, and relationships. To diagnose the source of this gap, we analyze the text-encoder prompt embedding space and uncover semantic collapse, a contraction of embedding spread coupled with distortion of inter-prompt similarity structure, which strongly correlates with structured utility loss. Guided by this insight, we propose StructureAware Geometric Regularization (SAGE), a safety alignment objective that explicitly preserves embedding spread and inter-prompt relational structure during adaptation. Our method restores structured utility (TIFA +5.0% over prior state-of-the-art) while maintaining strong safety performance and competitive coarse-grained utility scores. Our source code and trained models are available at https://adeelyousaf.github.io/SAGE_ECCV26_Project_Page/.

Simulating two-dimensional frustrated quantum matter is a grand challenge due to the sign problem and exponential Hilbert space complexity. In this work, we introduce the Holographic Quantum Transformer (HQT), a physics-inspired generative architecture that leverages global self-attention to resolve non-local entanglement patterns. We validate HQT on the square lattice $J_1-J_2$ Heisenberg model. On the heavily frustrated $8 \times 8$ lattice at the quantum critical point ($J_2=0.5$), HQT reaches a ground-state energy per site ($E/N$) of $\mathbf{-0.5001(1)}$, consistent with the expected finite-size scaling trend. Beyond numerical accuracy, HQT exhibits intrinsic physical awareness, autonomously recovering the underlying $J_2$ interaction geometry through interpretable attention maps. Our central contribution is ``Holographic Transfer", a zero-shot size-extrapolation protocol with rapid alignment: a model trained on $8 \times 8$ systems is directly projected onto larger $10 \times 10$ lattices via continuous positional-embedding interpolation and head re-initialization, achieving high-fidelity initialization and rapid convergence. This zero-shot protocol yields an energy of $E/N = \mathbf{-0.49782(3)}$, statistically consistent with the variational state of the art while requiring no from-scratch training on the target lattice. Our results establish generative attention as a scalable paradigm for transferable quantum simulation.

Understanding how complex cognitive functions are organized within artificial systems is central to interpreting large language models (LLMs) and relating them to biological cognition. Yet although LLMs exhibit broad cognitive-like behaviours, it remains unclear whether their internal representations form reproducible functional systems that explain behaviour, failure and links to human cognition. Here we present NeuroCogMap, a cognitive neuroscience-inspired framework that organizes internal features of LLMs into functional parcels and links them to interpretable functions, cognitive capabilities and a cognitive hierarchy. These parcels form a stable and semantically coherent organization that is partly conserved across models and functionally linked to model outputs. Within this organization, major LLM failures, including hallucination, bias, refusal failure and sycophancy, correspond to distinct disruptions in representational and behavioural-control systems, yielding internal signatures for mechanism-guided detection and targeted intervention. Beyond model behaviour, NeuroCogMap improves prediction of human cortical responses during naturalistic language comprehension, with the strongest correspondence in higher-order association cortex. At the cognitive level, its internal signatures expose latent strategies that guide refinements of classical models of human decision-making. Together, these findings establish NeuroCogMap as a system-level framework for mapping functional organization in artificial systems and for relating this organization to human cortical function and cognitive behaviour.

LLM agents increasingly rely on retrieval buffers to store and reuse past experience, yet the cache management policies governing these buffers remain largely ad-hoc. We formalize this as an online semantic cache replacement problem with switching costs, where items are matched by embedding similarity and hit quality is continuous rather than binary. Through experiments on two datasets from MemoryBench-Full (LoCoMo, DialSim) with 8 replacement policies, we reveal a surprising finding: classic heuristics (LRU, LFU) \emph{consistently underperform} the naive FIFO baseline on semantic workloads, due to the absence of temporal locality and frequency concentration. We propose SOLAR, a learning-augmented framework that derives modification timing from regret accumulation (achieving $\sim$17\% modification rate) and content selection from Bayesian online learning over implicit retrieval feedback. We prove SOLAR achieves a constant competitive ratio $\leq 3$, independent of cache size and horizon (vs.\ $Ω(K)$ for FIFO), and eviction regret $O(\sqrt{KT\log T})$, matching the $Ω(\sqrt{KT})$ lower bound up to logarithmic factors. Experiments demonstrate 5--75\% relative improvement over FIFO at tight cache sizes, with a clearly characterized phase transition at the working set boundary. Synthetic experiments with 5000-item pools further reveal an inverted-U relationship between pool size and retrieval quality, justifying capacity constraints as a retrieval noise phenomenon rather than a storage limitation.

Unsupervised Reinforcement Learning (URL) aims to pre-train scalable, skill-conditioned policies without extrinsic rewards, serving as a foundation for downstream control tasks. Despite recent progress, we argue that current off-policy URL methods are limited by two critical, overlooked bottlenecks: (1) non-stationary skill semantics and (2) brittle generalization. To address these challenges, we propose GenDa (Generalizable Data-efficient Agent), a unified framework for robust unsupervised reinforcement learning. First, we introduce a skill relabeling mechanism to mitigate non-stationarity and significantly improve data efficiency for pre-training. Second, we propose a Complementary Information Bottleneck (CIB), encouraging the learned skill policy to focus on ego-centric features and become robust to distribution shifts for downstream tasks. Through various experiments, we demonstrate that GenDa significantly enhances the scalability of URL with superior generalizability and data efficiency. Our code and videos are available at https://ihatebroccoli.github.io/official-GenDa.

Automated malaria diagnosis from blood smear microscopy is a critical challenge in global health AI; in resource-limited settings, the scarcity of expert microscopists remains the primary bottleneck to timely and accurate diagnosis. Three compounding failure modes prevent reliable clinical deployment of existing deep learning systems. First, end-to-end detectors treat unannotated cells as background during training, producing recall figures that are strongly influenced by annotation completeness rather than reflecting true cell recovery. Second, Non-Maximum Suppression tends to suppress valid detections in dense smear regions where infection counts matter most. Third, existing whole-slide detection pipelines lack per-cell spatial evidence for clinical audit, despite image-level explainability methods such as Grad-CAM having been applied to malaria image classification tasks. We present MalariAI, a two-stage decoupled framework that addresses all three failure modes in a unified pipeline. Stage 1 applies an annotation-agnostic distance-transform guided watershed algorithm to isolate every cell in a full 1600x1200 blood smear image, recovering 75.95% of ground-truth cells by centroid localisation across the 120-image NIH BBBC041 test set without any ground-truth input. Stage 2 fine-tunes EfficientNet-B0 with Focal Loss (gamma = 2.0, per-class inverse-frequency weights) on 64x64 crops, achieving 98.36% overall classification accuracy with 87.5% and 75.0% per-class accuracy on the rare schizont and gametocyte stages, compared to only 24.57% and 25.95% AP for a Faster R-CNN baseline on the same classes. Grad-CAM++ heatmaps generated per detected cell provide instance-level spatial evidence for clinical audit, enabling microscopists to verify model predictions at the individual parasite level without sacrificing classification performance.

Self-supervised Continual Graph Learning (CGL) aims to successively learn from a graph sequence with different tasks without label supervision - a paradigm that has attracted widespread attention. Most existing self-supervised CGL methods rely on instance-level consistency objectives that enforce stability of individual node (or node-pair) embeddings. Due to optimizing nodes in isolation, these methods fail to maintain global relational structure, causing inter-node correspondences to progressively distort under continual learning. To this end, we propose a novel Structure-Aware Optimal Transport (SAOT) framework that explicitly captures and preserves relational structure within graph representations across sequential tasks. Specifically, SAOT leverages optimal transport theory to capture global inter-node correspondences, thereby facilitating and enhancing graph representation learning. Simultaneously, SAOT incorporates a cross-task knowledge distillation mechanism to preserve the previous structural knowledge. Extensive experiments on four CGL benchmark datasets demonstrate that SAOT outperforms existing self-supervised baselines. In particular, SAOT achieves significant performance gains, improving average accuracy by up to 5% on CoraFull-CL and over 15% on Products-CL compared with state-of-the-art methods in the Class-IL setting.

Composed Image Retrieval (CIR) retrieves a target image from a reference image and a textual modification. While supervised CIR relies on costly triplets, Zero-Shot CIR (ZS-CIR) alleviates this reliance through proxy tasks trained on image-text pairs. However, existing proxy tasks primarily enhance visual and textual representations to accommodate a predefined composition mechanism such as pseudo-word injection into a frozen text encoder or linear feature arithmetic. As a result, the composition function itself remains unlearned, limiting the model's ability to express diverse and fine-grained semantic modifications. To address this, we propose FoCo, which models composition as two coordinated stages: focusing on modification-relevant visual content, and then completing the target semantics. We realize these through two proxy tasks: text-anchored visual aggregation to selectively gather visual content guided by localized textual semantics, and context-conditioned semantic completion to transform these aggregated visuals with the remaining scene context into a coherent composed representation. The tasks are trained jointly with a cross-instance contrastive objective, encouraging semantic diversity and discouraging shortcut composition strategies. Extensive experiments on four ZS-CIR benchmarks show FoCo's state-of-the-art performance and improved generalization.

Visual AutoRegressive modeling (VAR) has pioneered a coarse-to-fine multi-scale autoregressive generative paradigm, demonstrating strong capabilities in image generation. However, VAR still suffers from inherent deficiencies in multi-scale representation learning. Specifically, lower scales primarily capture global semantics, while higher scales focus on fine-grained details. Employing a shared architecture across scales induces optimization conflicts. Moreover, due to the causal autoregressive process, inaccurate semantics at early scales can propagate and significantly degrade the final output. To address these issues, we introduce a scale-aware token-routed Mixture of Experts (MoE) architecture, allowing scale-adaptive expert selection, thereby facilitating decoupled representation learning across scales. In addition, we enhance semantic modeling at early scales by incorporating external self-supervised features. Unlike naive alignment, we analyse and design a residual feature aggregation scheme tailored to the VAR paradigm. Extensive experiments show that our method significantly improves both training efficiency and generation quality. On the ImageNet 256*256 benchmark, our model achieves a superior FID compared to the dense baseline while requiring only half of the default training epochs and a smaller parameter budget, with a merely marginal increase in training cost. Moreover, the performance gap further widens with larger training epochs.

Large language model test-time training (TTT) is often evaluated through local proxy metrics: models are updated on recent tokens, retrieved context, target-domain data, or verifiable task attempts, and then judged by perplexity, future-token loss, long-context performance, or reward. These metrics are well matched to claims about stream adaptation, domain adaptation, context compression, and reward-backed test-time improvement. They are weaker evidence, however, for a capability that TTT results are increasingly used to motivate: deployed assistant memory, personalization, or sparse post-deployment learning, which instead requires behavioral evidence such as later recall, paraphrase robustness, retention, locality, conflict handling, and use in downstream actions after the original support context is removed. We introduce a behavioral evaluation framework that calibrates TTT memory claims to the evidence that supports them. It has two components: a claim-calibrated evidence ladder that separates stream/domain adaptation, bridge internalization, and deployment-time behavioral learning; and an evaluation protocol with matched explicit-memory baselines and mutually exclusive failure categories. We validate the framework by auditing recent TTT and memory-adjacent work and by instantiating it as a controlled diagnostic in which, in a sparse nonce-fact setting, one-step LoRA updates lower support and answer loss across three Qwen3 model scales while generated free-form recall stays at zero, exposing a measurable gap between proxy improvement and deployment behavior. The framework gives authors and evaluators a concrete standard for aligning TTT memory claims with the evidence actually reported.

Artificial intelligence (AI) and quantum information (QI) are rapidly co-evolving. AI is becoming a practical tool for learning, designing, controlling, and verifying quantum systems, while QI offers new computational models, representational structures, and learning-theoretic questions for AI. This survey reviews the interface from both directions. In the AI for QI direction, we organize recent progress around the central tasks of extracting information from limited measurements, training and discovering quantum algorithms, stabilizing noisy hardware, automating experimental and programming workflows, and extending learning-based methods to sensing and networking. In the QI for AI direction, we examine how quantum computation and quantum-inspired structures affect learning through algorithmic speedups, expressivity, trainability, generalization, neural-network design, and tensor-network representations. We close by identifying cross-cutting challenges in reproducibility, scalability, hardware realism, and co-design, arguing that progress will depend on tighter integration of theory, experiment, and hybrid quantum--classical systems.

Flow Matching (FM) has emerged as a powerful paradigm for speech generation but remains constrained by high inference latency and timbre leakage. To address these bottlenecks, we propose a unified guidance framework that enhances generation efficiency and robustness through two complementary strategies. On the data front, we introduce Data-guidance via heterogeneous augmentation, encouraging the model to disentangle linguistic content from acoustic residue. In parallel, we propose an enhanced Model-guidance mechanism that synergizes trajectory rectification with a novel intrinsic guidance objective. This approach distills conditional knowledge into network weights and straightens inference trajectory path, thereby eliminating Classifier-Free Guidance (CFG) overhead. Experiments demonstrate that our framework accelerates inference by nearly three times while effectively improving speaker similarity compared to state-of-the-art baselines.

Mobile on-device AI (MoAI) systems that integrate locally deployed AI models with conventional mobile software components are emerging as a key paradigm for delivering intelligent functionality directly on end-user devices. By moving inference from remote cloud services to the local mobile environment, such systems enable privacy-preserving, low-latency, and offline-capable AI functionality, yet introduce new security risks arising from the local storage of AI models. This paper presents the first comprehensive systematization of knowledge on MoAI security, covering security pillars, attack landscape, and defense landscape of MoAI systems. We further identify unresolved gaps in current attack and defense research and point to promising directions for future research in this emerging area. Our work establishes the first systematic framework for understanding the attack and defense landscapes of MoAI systems, serving as a foundation for building secure MoAI systems and advancing research in this critical domain. Companion resources are available at https://github.com/Jinxhy/Awesome-MoAI-Security.

Electroencephalogram (EEG) captures endogenous brain activity with high temporal fidelity and holds substantial promise for precise emotion decoding. However, channel redundancy and pronounced inter-subject variability remain key obstacles to scalable generalization. To address these limitations, we propose a novel framework termed PRioritized channel Importance with Semi-supervised doMain adaptation (PRISM), enabling label-efficient cross-subject emotion decoding. On the channel side, PRISM assigns differentiable, data-dependent channel weights via a lightweight expert ensemble, amplifying reliable electrodes while suppressing distractors. On the domain side, PRISM leverages unlabeled data through confidence-filtered pseudo-labels to drive consistency regularization and domain alignment, mitigating subject-specific heterogeneity. Extensive experiments show that PRISM surpasses state-of-the-art methods on DEAP, DREAMER, and SEED datasets, achieving robust cross-subject generalization given limited annotations.

Enterprise adoption of LLM agents requires model selection methods that balance quality, reliability, safety, latency, and cost. Evaluation-Driven Development and Operations (EDDOps) positions evaluation as a continuous governing function across the agent lifecycle rather than a terminal checkpoint. This paper presents a practitioner-oriented instantiation of EDDOps on AWS Bedrock AgentCore and proposes a cost-to-performance framework for selecting foundation models in enterprise agent architectures. We make three contributions: a conceptual synthesis explaining why traditional TDD/BDD methods are insufficient for non-deterministic LLM agents; an architectural mapping of the EDDOps reference architecture onto AgentCore Runtime, Evaluations, Agent Registry, and CloudWatch observability; and an empirical cost-to-performance decision framework validated through a proof-of-concept comparing three foundation models across two deployment paths. Using trace data from 30 single-turn invocations across six agents, 9 multi-turn evaluations, and registry-integrated governance, we show how evaluation evidence can convert model selection from a benchmark-ranking exercise into a governed economic decision. The results suggest that managed agent platforms can support EDDOps when they provide trace-native observability, pluggable evaluator frameworks, and governed registry-based discovery.

Large language models achieve strong performance on many reasoning tasks when allowed to externalize intermediate steps as Chain-of-Thought (CoT). However, many questions require the model to internalize the multi-step reasoning within a single forward pass before generating the answer. We study this challenge through two-hop reasoning, a representative task where the model must compose multiple pieces of parametric knowledge within a single forward pass. Standard non-recurrent Transformers suffer from a depth-local storage problem: facts learned in earlier layers are unavailable where second-hop retrieval happens. We found that Looped Transformers mitigate this issue by reusing the same memory, but still generalize imperfectly. We show that the remaining bottleneck is representational. In the two-hop reasoning task, the first loop often makes the correct bridge entity nearly perfectly decodable, yet the corresponding hidden state remains poorly aligned with the bridge token embedding. Surprisingly, an easy training-free realignment intervention nearly closes the generalization gap. Building upon this insight, we propose DiscoLoop, a looping architecture whose recurrence carries both a discrete embedding channel and a continuous hidden-state channel. DiscoLoop achieves near-perfect accuracy with substantially fewer training steps across symbolic and synthetic-language multi-hop reasoning tasks. When applied to real-world pretraining, DiscoLoop attains lower training loss and stronger benchmark performance than looped-transformer baselines, suggesting that the mixed-channel design transfers to practical language modeling.

Conversational data is increasingly used as a persistent source of user state for long-running assistants and AI agents. However, querying this data remains challenging because conversations naturally evolve: plans are revised, preferences change, and later messages frequently supersede or contradict earlier information. Existing long-memory pipelines largely treat memories as independent text or vector objects. This approach often retrieves semantically similar but stale evidence, offering limited support for state-aware reasoning. To address this problem, we present TRACE, a query processing framework over temporal evidence graphs for evolving conversational data. TRACE models conversations as a hierarchical graph spanning events, sessions, and topics, enriched with typed temporal, causal, update, and contradiction relations. Crucially, the framework maintains validity annotations so obsolete facts remain accessible for historical queries but are discounted for current-state answers. At query time, TRACE combines vector-based note retrieval with graph-guided evidence search, generating validity-aware support paths and a hybrid context for answer generation. This design separates lexical recall from evidence reconstruction, enabling bounded query-time reasoning over long conversational histories. Experiments on long-conversation query-answering (QA) benchmarks show that TRACE improves temporal and multi-hop reasoning, with ablations highlighting the importance of hierarchy, update-aware seeding, and path-grounded evidence.

Autonomous agents, whether LLM-driven software agents or robotic physical agents, face a common class of failure modes when operating without continuous human oversight: safety violations from unverified actions, behavioral instability from unconstrained loops, and continuity loss from unhandled error states. We develop \system{}, a discrete-time control system that combines five execution gears (\Gobs{}, \Gsug{}, \Gplan{}, \Gexec{}, \Gint{}) with utility-gated dispatch and event-driven fallback. For the single-agent case, we prove monotonic stability, execution safety, eventual stabilization, fallback completeness, and equivalence to a gear-constrained Markov decision process. For multi-agent cyber-physical systems (CPS), we apply the established \smart{} managed-autonomy lifecycle and map runtime evidence into its four governance states (\Stable{}/\Meta{}/\Assisted{}/\Regulated{}). Consensus gating, swarm-level Lyapunov analysis, per-agent gear authority, and rendezvous control provide distributed safety and stability guarantees, including zero collision under the stated assumptions. We evaluate the resulting runtime on a three-agent UR5 robotic assembly cell using fault magnitudes calibrated from the NIST \emph{Degradation Measurement of Robot Arm Position Accuracy} dataset across 10,000 Monte Carlo episodes. It achieves a 99.6\% anomaly detection rate versus 2.1\% for the single-agent baseline, reduces detection latency by $3.5\times$, and supplies a formal physical-workspace safety certificate. The execution gears act as micro-level permissions beneath the \smart{} runtime governance states, separating action control from autonomy governance.

Recursive Feature Machines (RFMs) are a class of kernel machines that utilize the Average Gradient Outer Product (AGOP) as a mechanism for feature learning. They have been shown to effectively replicate the learning dynamics and feature representations of Feedforward Neural Networks (FNNs) across various settings. However, despite comparable capacity for feature learning and the similarities in the features they acquire, RFMs exhibit significantly lower performance than neural networks in certain data-corrupted scenarios. In this work, we investigate these limitations in mathematical problems. As a solution, we introduce a remarkably effective transformation applied to the training labels which promotes learning in noisy, complexly represented, and class-imbalanced data. This simple yet powerful adjustment enables RFMs to close the performance gap with FNNs and, in some cases, even surpass them.

A growing body of literature suggests that training data membership inference problems are fundamentally hard tasks in modern language modeling settings. We argue that output watermarking techniques are the right gadget to make training membership tests for generative models more tractable, based on prior results showing that language models exhibit residual watermark "radioactivity" under partially watermarked training datasets. We pit a watermark-based dataset inference approach head-to-head against traditional loss-based membership inference methods and show that watermarking can achieve comparable membership detection performance when subset exposure is high enough, under an alternate set of assumptions.

We establish approximation and learning guarantees for Fourier neural operators (FNOs) applied to time-$T$ solution operators of dissipative evolution equations. The analysis builds on the premise that FNOs can efficiently approximate and learn solution operators whenever these operators admit stable and accurate spectral discretizations. To formalize this idea, we introduce classes of evolution operators defined through spectral methods and derive FNO approximation bounds and polynomial sample complexity guarantees for these classes. For equations with polynomial nonlinearities, the learning rates depend primarily on the smoothness of the input space and the dimension of the physical domain. Our results hold uniformly over broad families of dissipative equations, rather than for a single fixed PDE, and apply in particular to the Navier--Stokes, Allen--Cahn, and Cahn--Hilliard equations. For equations with non-polynomial smooth nonlinearities, we prove that polynomial sample complexity still holds with rates that now additionally depend on the smoothness of the nonlinear terms and the dissipation strength. Overall, we connect classical spectral approximation theory with modern operator learning and explain when FNOs can learn nonlinear evolution operators efficiently.

Foundation video diffusion models are increasingly viewed as world simulators for embodied agents, yet their pretraining on internet-scale generic video leaves them poorly aligned with real-world deployment domains. We study parameter-efficient adaptation of a pretrained foundation video world model to retail scenes: when synchronized egocentric and exocentric video of the same activity are available, which viewpoint of training data produces the strongest adapted model? We introduce RetailSMV (Retail Synchronized Multi-View), a corpus of 32,105 captioned retail clips from five supermarkets with synchronized ego/exo capture from the store-staff perspective (stocking, arranging, weighing, managing supply carts, scanning at checkout), rather than the customer-centric framing of prior retail video corpora, and train three matched Low-Rank Adaptation (LoRA) configurations of Cosmos3-Nano (egocentric-only, exocentric-only, combined) under identical hyperparameters. On a 200-clip held-out test set evaluated with seven complementary metrics under a strict paired statistical protocol, exocentric-only adaptation matches or exceeds combined adaptation on six of seven point estimates and is significantly better on LPIPS, PSNR, and DreamSim, despite training on only 15,985 exocentric clips (versus 32,105 for combined). A symmetric paired comparison further shows that adding exocentric data to egocentric-only training helps while adding egocentric data to exocentric-only training hurts. The absolute adaptation gap is largest at the shortest rollout time, identifying the near-horizon prediction window as the regime in which adaptation is most beneficial.

We present a real-time musical interface that converts natural-language scene descriptions into evolving procedural soundscapes. A performer types a prompt such as "warm jazz cafe at midnight" and steers it through direct parameter adjustments - stepping brightness down, switching a rhythm style - each producing a predictable, audible shift without re-prompting. Where GPU-bound text-to-audio systems synthesize monolithic waveforms, our instrument generates human-readable configurations over a categorical schema, enabling fine-grained performer control; most valid combinations are designed to sound musically coherent. Three interchangeable backends - embedding retrieval for sub-second CPU-only use, hosted LLMs via API, and a fine-tuned 270M local model - all emit the same schema. A live generator architecture continuously emits audio while resolving new instructions in the background, crossfading seamlessly when ready; even when an LLM takes 5-12 seconds to respond, the audience hears uninterrupted sound - reframing text-to-music as an ongoing performable stream rather than a one-shot generation. We evaluate text-audio semantic alignment using LAION-CLAP on held-out prompts as a technical proxy, finding that retrieval-based configuration outperforms random valid configurations on this metric, while noting that LAION-CLAP also informed retrieval-map construction. We report performance observations, informal listener feedback, and release materials for the SDK, dataset artifacts, model, and audiovisual performance interface.

The bias-reliability tradeoff conjectures that LLM evaluation systems are constrained in (gamma, H, CV) space, where evaluator coupling (gamma), strategy diversity (H), and small-sample measurement reliability (CV(N)) cannot be simultaneously optimized at fixed sample size N. Prior evidence rests on n=5 conditions with complete metrics from a single study. We expand the empirical base to 11 conditions, measuring gamma and H for all 11 (nine with valid weight vectors) and CV(N=5) for seven with sufficient seeds (N >= 5). Five conditions provide the complete (gamma, H, CV) triple. The data confirm the trade-off: conditions with low evaluator coupling (gamma < 0.2) exhibit high measurement noise (CV(N=5) > 1.0), while conditions with strong coupling (gamma > 0.9) achieve low noise (CV(N=5) < 0.16). The correlation r(H, gamma) = -0.989 (n=5, excluding GPT-4o conditions) confirms that evaluator coupling suppresses strategy diversity. Four GPT-4o conditions show gamma=0.000 and H=1.000 across all seeds -- a pattern we attribute to version drift in the June 2026 GPT-4o API. No condition occupies the region {gamma < 0.2, CV(N=5) < 0.3}. We release all per-condition metrics as a standardized benchmark dataset for evaluator comparison.

Asynchronous Many-Task (AMT) runtimes use futures as placeholders for values produced by other tasks. In the ItoyoriFBC AMT runtime, the existing future-only model binds each future to its producer at creation time and requires the number of tasks that read each future to be fixed at compile time. This prevents directly expressing algorithms that create dependencies dynamically. We extend ItoyoriFBC with an implementation of a promise-future model that lifts these limitations. Thereby, our ItoyoriFBC variant supports dynamic algorithms such as Hierarchical LU factorization (HLU). We experimentally evaluated our implementation using HLU on up to 16 nodes and observed near-ideal scaling with a 15.6x speedup.

The emergence of powerful deep generative models based on diffusion and flow matching has enabled the learning and modeling of complex distributions. Learning quantum distributions, however, remains challenging due to the inherent difficulty of accurately modeling the meaningful physical properties of quantum states. We propose Quantum Flow Matching (QFM), a novel generative model designed to learn quantum distribution by utilizing spin Wigner function and flow matching. By converting density matrix into the spin Wigner function and leveraging functional flow matching to learn distributions in function space, QFM enables accurate and effective learning of multi-qubit quantum distributions. We demonstrate the effectiveness of our method by evaluating physical quantities such as trace, purity, and entanglement entropy of the generated quantum states, accurately capturing the underlying physics of the given quantum distributions.

When LLM agents use evaluator feedback to adapt their behavior in closed loops, evaluator biases propagate through the agent's strategy distribution -- a phenomenon known as evaluator preference coupling. Prior work has documented coupling across multiple evaluator families and model versions, but the field lacks a standardized protocol that enables third-party researchers to (i) reproduce coupling measurements, (ii) compare results across evaluators and time points, and (iii) detect measurement decay as proprietary evaluators silently update. This paper provides the protocol. We specify EPC (Evaluator Preference Coupling) -- a detailed, RFC-style protocol specification for the four-phase isolation paradigm, covering executor and evaluator configuration, strategy and task design, the TTRL update rule, metric computation (gamma, JSD, ECE, Brier), and output schema. We accompany the protocol with a versioned Reference Snapshot v1.0: coupling measurements for eight evaluator conditions (N=122 unique experimental repetitions across GPT-4o, Qwen, DeepSeek, and others) derived from five independent studies, annotated with evaluator version identifiers, API endpoints, and measurement dates. The snapshot is explicitly time-bound: all values are conditional on specific model versions and are expected to decay as proprietary evaluators update. We define a versioning convention (vX.Y-Z, encoding protocol version, snapshot version, and evaluator generation) and provide a usage guide covering adoption, interpretation, and known pitfalls. The protocol, reference snapshot, and implementation code are released as open infrastructure.

Human motion forecasting in unconstrained real-world videos remains challenging due to the ambiguity of future behaviors and the presence of noisy multimodal observations. While facial affect potentially provides complementary behavioral cues, its practical utility and mechanistic boundaries within motion forecasting frameworks remain poorly understood. In this work, we present a systematic study investigating the utility and temporal limitations of affect-conditioned forecasting in-the-wild. We establish a rigorous multimodal pipeline combining MediaPipe body pose trajectories with HSEmotion facial affect representations, and introduce the Gated Affect Transformer (GAT) to dynamically regulate cross-modal information flow. Through extensive multi-horizon evaluations under a strict subject-wise protocol, we demonstrate that naive early cross-modal concatenation consistently degrades forecasting accuracy relative to pose-only baselines. Conversely, our proposed gating mechanism stabilizes cross-modal integration by adaptively controlling the affective stream. Crucially, controlled counterfactual experiments using shuffled and randomized affect inputs reveal that the learned gate successfully suppresses unstructured cross-modal noise while remaining responsive to plausible affective signals. Furthermore, our empirical results indicate that facial affect features provide bounded, horizon-dependent predictive cues strictly within short-to-medium windows (e.g., 30 frames), whereas long-term trajectories remain predominantly governed by intrinsic kinematic continuity. Our findings provide empirical evidence that facial affect should be regarded as a complementary behavioral cue rather than a dominant driver of future motion, offering practical guidance for selective multimodal fusion in unconstrained human motion forecasting.

Achieving true artificial general intelligence requires foundation models capable of integrating new modalities without forgetting prior knowledge. However, accommodating continuous generative objectives alongside discrete understanding tasks causes severe gradient conflicts. Existing architectures, including standard Mixture-of-Experts (MoE), are highly susceptible to representation overwriting. Even structurally partitioned paradigms like Mixture-of-Transformers (MoT) remain vulnerable to catastrophic forgetting, severely impeding multimodal scalability. In this work, we introduce Rosetta, a composable native multimodal pretraining framework designed for seamless and non-destructive modality expansion. Rosetta adopts a modular paradigm where core foundational knowledge is preserved within global shared experts, while modality-specific capabilities are distributed across plug-and-play experts. To guarantee non-destructive composition, we propose Momentum-Anchored Orthogonal Projection (MAOP). MAOP leverages the optimizer's momentum state as an implicit semantic anchor, selectively neutralizing conflicting gradient components from new modalities while preserving synergistic updates. Extensive evaluations demonstrate that, while standard MoE and MoT architectures suffer catastrophic forgetting of previously acquired knowledge, Rosetta robustly preserves established language and visual understanding. Furthermore, it delivers superior image generation and unlocks cross-modal synergy, paving the way for truly composable and unified multimodal foundation models. To facilitate further multimodal research, we release our code and checkpoints to the community. Project page at https://rosetta-lmm.github.io/.

Designing deployable and resilient network topologies from natural language requirements remains a challenging problem in network automation. This work investigates the ability of Large Language Models (LLMs) to generate structurally valid and constraint-compliant network topologies through a constraint-driven pipeline combining hierarchical modeling and systematic validation. The framework is evaluated via a multimodel comparison of proprietary and open-weight LLMs across four realistic network scenarios released as a public dataset. We assess structural correctness using node and edge F1-scores against reference topologies, and evaluate resilience through server and content connectivity metrics. In addition, we analyze common failure modes, including interface mismatches and directional inconsistencies in generated topologies. Overall, this work provides a systematic benchmark for understanding how LLMs handle structural and resilience constraints in topology synthesis, and supports informed model selection for AI-driven network design.

Oscillatory neural networks (ONNs) have emerged as a promising neuromorphic architecture, leveraging coupled dynamical systems to perform computation and represent information through phase relationships. Their interactions can be designed to support intrinsic energy-minimizing dynamics, enabling tasks such as associative memory and optimization, and positioning them as a candidate architecture for continuous learning and inference. We present a neuromorphic primitive implemented using memristive edges with inhibitory couplings as a potential design for autonomous learning, and provide circuit simulation validation that the system is capable of denoising noisy inputs on an auto-associative task. While numerical Hopfield/Ising models routinely assume signed weights, neuromorphic implementations of ONNs often fail to realize negative weights due to device and circuit constraints. A practically implementable route to inhibitory (negative) weights is particularly valuable: it expands the class of attractor structures accessible to oscillator networks beyond purely synchronous couplings, and supports phase-coded memories where anti-phase constraints are not merely transiently enforced during training but can persist autonomously after release. We provide circuit simulations and theoretical analyses demonstrating that signed effective weights are necessary for anti-phase attractors to persist autonomously.

Autonomous vehicles must safely navigate complex environments where planning-critical agents may be hidden from view. Current approaches often treat all occlusions with uniform conservatism, yielding needlessly defensive driving, or they infer hidden spaces without estimating the impact on the planner. This work bridges the critical gap between perception and planning by enabling Vision-Language Models (VLMs) to identify and reason about the specific hidden agents that are most critical to the ego-vehicle's trajectory. We introduce a novel framework that uses Planning KL-divergence (PKL), an information-theoretic metric, to systematically identify and rank occluded agents based on their impact on the ego vehicle's plan. Using this planning-aware ranking, we employ an expert VLM (GPT-5) to generate rich, structured annotations that capture the visual evidence and reasoning required for this task. We apply this framework to the nuScenes dataset to create a new benchmark focused on high-impact scenarios. We conduct comprehensive experiments on a wide range of general-purpose and domain-adapted VLMs, demonstrating that fine-tuning on our PKL-guided data yields dramatic performance improvements across all models. Notably, our results show that smaller, fine-tuned models significantly outperform their much larger zero-shot counterparts, and that our PKL-guided data selection strategy improves performance by approximately 30\% over random sampling. Our work presents the first systematic approach for training VLMs to focus on planning-critical occlusions, enabling more semantically grounded and efficient risk assessment in autonomous driving.

Airbnb is a community based on connection and belonging -- many hosts on Airbnb are everyday people who share their worlds to provide guests with the feeling of connection and being at home; Airbnb strives to connect people and places. Among our efforts to connect guests and hosts, we provide tools to enable hosts to set competitive prices, which helps improve affordability for guests while helping hosts get more bookings. We also personalize the guest experience to show them the listings that match their needs. To help inform these efforts, we combine economic modeling and causal inference techniques to understand how guests book stays based on the prices hosts set, among other factors, and how that preference varies across different guests and listings. Such understanding helps us identify opportunities for Airbnb to support the marketplace and better connect guests and hosts. For example, understanding how much guests respond to different prices helps optimize the tools that we provide to hosts, in order to enable hosts to choose and set competitive prices that further balance demand and supply. As another example, understanding heterogeneity in guest preferences helps us personalize the guest experience and better match them with the listings that meet their needs, based on how much they respond to different prices and other factors.

Current large-language-model (LLM) physics benchmarks are usually scored by answer accuracy, which cannot distinguish genuine reasoning from recall of familiar problem patterns and reveals little about where a model's reasoning breaks down. We introduce an auditable four-stage diagnostic that evaluates whether an LLM can reason inside an unfamiliar physics framework through induction, formulation, prediction, and review. The diagnostic combines locked pre-registrations, fresh sessions between stages, dual-LLM judging, and a human-audit pathway, and we apply it to three parallel physics worlds: a single-equation counterfactual world ($F=mv$), a historical framework (Aristotelian mechanics), and a four-domain counterfactual world (Decay World). Across Claude Opus 4.7, GPT-5.5, and Gemini 3.1 Pro, the three worlds yield composite PASS rates are 6/15, 6/15, and 0/15 respectively (content $\land$ structural for $F=mv$ and Aristotelian, content axis only for Decay World where the structural axis is out of scope). The most pointed empirical pattern is a qualitative-versus-quantitative asymmetry: in Decay World, models almost never predict the wrong direction of change, but frequently compute the wrong ratio by slipping back to standard-physics relations. The protocol also surfaces two methodology findings: LLM-judge reliability does not transfer across frameworks, and Stage 4 self-review is weak in every framework, with the model's own review wrongly reporting no earlier error in at least two-thirds of the trials that actually contained one. We release the full prompts, responses, verdicts, and audit records.

Federated Learning (FL) is a distributed machine learning (ML) paradigm with collaboration among multiple clients without sharing data. FL is challenging under data heterogeneity and partial client participation. Learning sparse models is useful for communication and computational efficiency in FL, but it is especially difficult in the small-sample high-dimensional regime (d >> N) where optimization can yield parameter configurations that fail to generalize to unseen test data. While magnitude-based pruning doesn't account for uncertainty exploration in the parameter space, a formulation with probabilistic gates and an L0 constraint allows sampling from competing sparse configurations during training. In this work, we study entropy regularization of gate distributions as a mechanism to maintain uncertainty in sparse federated optimization by preventing early commitment to sparse support. We examine its impact under data heterogeneity, client participation heterogeneity, and sparsity. Experiments on synthetic and real-world benchmarks show consistent improvements over federated iterative hard thresholding (Fed-IHT) and pruning after dense federated averaging (FedAvg) training, both in statistical performance on test data and in sparsity recovery accuracy.

Effective writing feedback is among the strongest drivers of student learning, yet producing it at scale is labor-intensive. LLMs offer a natural path to scaling writing support, but two gaps stand in the way: few public corpora capture how instructors actually deliver feedback in real classrooms, and no reliable method measures whether generated feedback aligns with what an instructor would write. We address both. SEFORA is a public corpus pairing instructor inline feedback with assignment prompts, rubrics, scores, and multi-draft revisions across various college writing genres, comprising 564 drafts and 8,240 instructor annotations. UniMatch is a reference-based evaluation framework for open-ended generation: it segments feedback into feedback units, scores their semantic correspondence under instructor-derived criteria, and aligns them via optimal matching to yield interpretable precision, recall, and F1. Across 74 experimental configurations spanning multiple LLMs, no setting exceeds 0.4 F1. UniMatch reveals that models struggle to identify the feedback instructors would prioritize, and performance degrades as models generate more.

Traditional robot programming is challenging: it requires orchestrating multimodal perception, managing physical contact dynamics, and handling diverse configurations and execution failures. We introduce ASPIRE (Agentic Skill Programming through Iterative Robot Exploration), a continual learning system that autonomously writes and refines robot control programs in a code-as-policy paradigm while compounding experience into a reusable skill library. ASPIRE discovers skills that persist across tasks, simulation and real-world settings, and embodiments. It operates in an open-ended loop with three components: (1) a closed-loop robot execution engine that exposes fine-grained multimodal traces, enabling autonomous failure diagnosis, repair synthesis, and validation; (2) a continually expanding skill library that distills validated fixes into reusable, transferable knowledge; and (3) evolutionary search that generates diverse task sequences and control programs to explore beyond single-trajectory refinement. ASPIRE surpasses prior methods by up to 77% on LIBERO-Pro manipulation under perturbation, 72% on Robosuite bimanual handover, and 32% on BEHAVIOR-1K long-horizon household tasks. Its accumulated library also enables zero-shot generalization to unseen long-horizon tasks: on LIBERO-Pro Long, ASPIRE achieves 31% success versus 4% for prior methods despite their use of test-time reasoning and retries. Finally, simulation-discovered skills provide initial evidence of sim-to-real transfer, substantially reducing real-robot programming effort across different embodiments and robot APIs.

Algorithm development for radioisotope identification in mobile urban search scenarios face significant challenges from non-uniform backgrounds, momentary source encounters, and severe class imbalance between rare threat signatures and background measurements. We present a machine learning-based approach to this problem that converts list-mode gamma-ray data into two-dimensional waterfall spectrograms and applies computer vision architectures to the resulting images. Rather than treating waterfalls as conventional images, we employ a representation where consecutive time spectra can form input channels, similar to RGB channels in color images. This representation encodes both spectral and temporal information, enabling neural networks to more effectively learn patterns that distinguish source signatures from background fluctuations. We evaluate three architectures, a multilayer perceptron (MLP), convolutional neural network (CNN), and vision transformer (ViT), on the Radiological Anomaly Detection and Identification (RADAI) benchmark dataset. At a false positive rate of less than one false alarm per hour, our CNN outperforms the previous-best non-negative matrix factorization (NMF) method across all global metrics, achieving true detection, classification, and identification rates of 0.4334, 0.3965, and 0.2950 respectively, compared to 0.4151, 0.3611, and 0.2625 for NMF. At lower false positive rate constraints, the neural network approaches show comparable but ultimately lower performance than NMF, indicating opportunities for further research.

LLMs, solvers, and agent teams increasingly generate workflow actions, repairs, and plans, but a generated action may be syntactically valid yet stale, infeasible, conflicting, or destructive of the evidence that triggered a repair. We introduce Agentic Transaction Processing (ATP), a transaction model that treats generated actions as untrusted proposals until they pass deterministic admission under a declared, executable constraint set C. The principle is two-sided: a proposal is not truth, and no proposal foresees every disruption: anything may propose, but only the runtime admits and commits, and when an unforeseen disruption strikes it repairs reactively within bounds rather than trusting a fresh proposal. Relative to C, committed-state correctness becomes independent of the competence, honesty, or learning of the proposing layer. We realize ATP in Mnemosyne, a runtime with an append-only transition log, effective-state projection, dependency-safe compensation, and active commitment records, and prove four safety properties relative to C (authority separation, serial-equivalent generative admission, evidence-preserving repair, and obligation containment) together with a bounded-reactive-repair guarantee for its localized repair protocol (LCRP). A reproducible artifact rejects the targeted violations across nine falsification tests while still admitting valid work, at under 6% projection-and-validation overhead, and bounded local repair edits an order of magnitude fewer operations than global recompute. Mnemosyne is open source: https://github.com/eyuchang/Mnemosyne/tree/arxiv-atp-rq1-rq9b-r8-v2.

A central goal of science is to produce valid explanations of complex systems: high-level causal accounts that faithfully reflect the behavior of lower-level mechanisms. Yet no consensus exists on how to measure whether a proposed high-level explanation is actually valid. We introduce a benchmark of ten complex systems spanning both discrete and continuous state spaces, as well as static and dynamical regimes, each equipped with consensual ground-truth causal explanations and invalid contrastive conditions. Within a unified causal abstraction framework, we systematically evaluate over thirty candidate metrics drawn from observational, functional, information-theoretic, and causal families. Our results show that only the latter reliably discriminates valid from invalid abstractions, and only when incorporating faithfulness testing over unmapped variables. Building on these findings, we introduce the Causal Abstraction Error (CAE), a continuous validity metric with an explicit faithfulness test, which passes all discrimination tests across every system and can converge with as few as 30 sampled interventions. We offer it as a general-purpose metric for the discovery and validation of high-level explanations.

Test-Time Adaptation (TTA) seeks to improve model robustness under distribution shifts by adapting parameters using unlabeled target data. However, in the absence of supervision, entropy-based adaptation is fundamentally underconstrained: multiple distinct parameter updates can achieve similarly low entropy while inducing drastically different decision boundaries. This phenomenon, known as underspecification, renders standard TTA brittle and prone to collapse into spurious modes. In this work, we reinterpret TTA through a posterior-inspired lens induced by entropy minimization, where low-entropy solutions define a pseudo-likelihood over parameters. Instead of committing to a single point estimate, we introduce a particle-based diversification framework that explores multiple plausible adaptation trajectories simultaneously. Our method can be viewed as a structured exploration of multiple plausible adaptation solutions, implemented through multi-level diversification at the output, parameter, optimizer, and input levels. Crucially, the framework acts as a plug-and-play wrapper compatible with existing TTA methods. Extensive experiments on challenging benchmarks demonstrate consistent gains in stability and robustness, achieving improvements of 3-4% under mixed shifts, 2-3% with batch size one, and 1-2.5% under label shifts, outperforming state-of-the-art baselines. Our results suggest that treating TTA as a multi-hypothesis inference problem, rather than a single-point optimization task, is key to mitigating underspecification and enabling reliable real-world deployment.

Accurate prediction of complex dynamical systems from noisy measurements remains a significant challenge in scientific computing. Kernel ridge regression learning strategies are often effective when applied to clean data, but have limited success with noisy data. Recent work has observed that a weak formulation can act to filter noisy data, and different learning strategies have achieved increased noise robustness with a weak-form framework. In this manuscript, we give an overview of the filtering mechanism behind the weak formulation and provide a bias-variance error decomposition. Using these insights, we combine a weak formulation with a kernel learning strategy to propose Weak-form Kernel Ridge Regression (WKRR) for learning dynamical systems. The proposed framework is simple to implement, effective for both clean and noisy data, and outperforms several baseline methods. We demonstrate the performance of WKRR on chaotic benchmark systems in up to 64 dimensions, as well as 15,000-dimensional real-world fluid data.

We present an algorithm for the group distributionally robust (GDR) least squares problem. Given $m$ groups, a parameter vector in $\mathbb{R}^d$, and stacked design matrices and responses $\mathbf{A}$ and $\mathbf{b}$, our algorithm obtains a $(1+\varepsilon)$-multiplicative optimal solution using $\widetilde{O}(\min\{\mathsf{rank}(\mathbf{A}),m\}^{1/3}\varepsilon^{-2/3})$ linear-system-solves of matrices of the form $\mathbf{A}^{\top}\mathbf{B}\mathbf{A}$ for block-diagonal $\mathbf{B}$. Our technical methods follow from a recent geometric construction, block Lewis weights, that relates the empirical GDR problem to a carefully chosen least squares problem and an application of accelerated proximal methods. Our algorithm improves over known interior point methods for moderate accuracy regimes and matches the state-of-the-art guarantees for the special case of $\ell_{\infty}$ regression. We also give algorithms that smoothly interpolate between minimizing the average least squares loss and the distributionally robust loss.

While most image deblurring techniques directly restore the spatial image variable, we propose an amplitude and phase decomposition recognizing the importance of accurate phase estimation in recovering sharp image details. To that end, we first develop novel linear minimum mean squared (LMMSE) estimators of the amplitude and phase of the blurred, noisy image observation. An iterative optimization algorithm follows that recovers the sharp image using the aforementioned LMMSE estimators. Finally, matrix parameters that are statistically determined and fixed in the iterative algorithm are now learned using a training dataset of clean and degraded observations. Our deblurring engine is dubbed UPADNet (Unrolled Phase and Amplitude Decomposition Network), such that each iteration of the underlying phase and amplitude recovery algorithm is parameterized and trained end-to-end. Experiments over benchmark evaluation datasets such as GoPro, RealBlur and COCO datasets confirm that UPADNet outperforms state of the art deep networks including those based on algorithm unrolling in the image domain. The benefits of UPADNet are even more pronounced in high noise and limited training data regimes.

Maltese has decent text corpora and pretrained language models, but, like many languages outside the handful with large OCR benchmarks, only a single known real labelled PDF corpus for OCR training, 57 page, far below what paragraph-level training needs: low-resource for OCR specifically. With no real corpus to train on at scale, we built a synthetic training pipeline and a 5-stream Tesseract LV-ROVER ensemble, and report results on a 422-paragraph benchmark against a fine-tuned-Tesseract baseline of character error rate (CER) 0.0234. Ensemble recognition alone improves CER by 44 percent, to 0.01317; a five-stage post-processing chain brings the full pipeline to CER 0.00700, a 70 percent reduction. Most of that chain is typographic normalisation, but one stage recovers misread diacritics rather than aligning punctuation, so we report it as a recognition gain rather than folding the whole chain under one label. We treat the 44 percent figure as the portable estimate of what the recogniser learned, and the 70 percent figure as specific to this benchmark's label convention.

New device layouts pose a challenging modeling problem due to the lack of large datasets for each specific layout. Biosignal foundation models offer a plausible solution if they are able to generalize to new layouts effectively. To improve cross-layout transfer, we study how different channel embedding techniques behave when pretraining layouts differ substantially from the downstream decoding layout. We propose Device Passport, a new channel embedding technique that learns experts and mixture models that take each channel's functional activity and metadata as input. This contrasts with prior embedding methods, which typically use only functional information or only metadata to look up learned or fixed positional embeddings. Across controlled subset-transfer experiments and realistic transfer to ear-EEG, Device Passport is competitive overall and improves over the strongest learned baseline in the layout-transfer regimes that motivate this work. These results suggest that channel embedding design is a key consideration when reusing large-scale pretrained biosignal models on new devices.

We present Seed2.0, a model series that takes a meaningful step toward solving complex, real-world tasks. Our approach begins with identifying users' genuine needs and constructing a reliable, forward-looking evaluation system by selecting and abstracting benchmarks grounded in these needs and in realistic, complex scenarios. Guided by this evaluation system, Seed2.0 targets two persistent challenges, long-tail knowledge and complex instruction following, substantially improving the model's reliability on intricate, long-horizon tasks. Beyond these, Seed2.0 delivers world-leading reasoning intelligence, visual understanding, and search capabilities that address the most common needs of a broad user base. Through extensive real-world use cases documented in this model card, we demonstrate that Seed2.0 begins to exhibit the ability to handle initial complex real-world tasks, delivering greater value to hundreds of millions of users.

Large audio-language models (LALMs) frequently hallucinate by overriding acoustic evidence with language priors. While contrastive decoding (CD) offers training-free mitigation, existing methods rely on blunt perturbations like masking or noise, leaving structured audio transformations unexplored. We explore this design space by evaluating a diverse library of targeted audio perturbations and adaptively selecting the optimal negative branch for each task and example. First, we improve upon earlier prompt engineering by showing that a simple binary yes/no constraint reduces the model's tendency to falsely confirm absent audio features. Second, evaluating our library across temporal, spectral, frequency, and amplitude domains reveals that optimal transformations are highly task-dependent; for instance, reversing the audio array disrupts temporal coherence, raising accuracy on the temporal order task from 74.7% to 81.4%. Finally, we trained a light-weight perturbation selector on model hidden states to dynamically route negative branches, yielding an additional +4.3% gain on the existence task.

How do two agents invent a shared language from scratch? In a Lewis signaling game, a sender and receiver must coordinate on a code using only their interaction history. We study five memory architectures across varying channel configurations with LLM agents and find that memory architecture matters more than channel capacity. Agents with a persistent private notebook benefit from surplus channel capacity and avoid the high-capacity collapse seen in stateless agents, achieving the most reliable coordination ($0.867 \pm 0.023$ at capacity = 25). Stateless agents peak at moderate capacity and then degrade as the vocabulary grows beyond what a rolling context window can track The notebook externalizes learned conventions, freeing agents from having to re-derive codes each round. An information bottleneck-inspired argument predicts an optimal capacity equal to the number of objects. Instead, the bottleneck (capacity = 8) proves to be a fragility point, and surplus capacity is generally better. We show that channel capacity alone cannot predict coordination; memory architecture determines whether agents turn interaction history into stable conventions, and both dimensions are needed to understand how signals become language.

Modern time-domain surveys such as the Zwicky Transient Facility (ZTF) generate hundreds of thousands of alerts each night, making real-time decisions for follow-up observations a central challenge in time-domain astronomy. Robust early classification is crucial for making informed decisions, but is hindered by sparse light curves and degeneracies between classes. In this work, we leverage multimodality to substantially improve real-time classification and demonstrate the practicality of our approach by deploying our model on the ZTF alert stream. Building on the Online Ranked Astrophysical CLass Estimator (ORACLE), we introduce the ORACLE-2 models, which combine light curves, metadata, and images for real-time hierarchical classification. Using both real and simulated datasets, we show that incorporating additional modalities consistently improves classification performance. On observations from ZTF's Bright Transient Survey, our best-performing model, ORACLE-2 Omni, achieves a macro F1 score of 0.73 -- an improvement of up to 11% over models using light curves and metadata alone, and up to 40% over light-curve-only models, with the strongest gains realized at early times. To demonstrate applicability to the Legacy Survey of Space and Time, which will increase alert volume by more than an order of magnitude, we train a light curve + metadata variant on the simulated ELAsTiCC dataset. This model achieves a macro F1 score of 0.88, an improvement of up to 13% over the light-curve-only variant, matching the performance of other state-of-the-art models. Finally, we quantify the trade-offs between performance and throughput, identifying regimes where multimodal approaches offer the greatest benefit. These results show that combining multiple modalities improves early-time classification, enabling more effective triage of high-volume alert streams for current and future time-domain surveys.

Watermarking promises a statistical trace of large language model (LLM) use, but real documents, after editing or paraphrasing, rarely arrive as purely human-written or purely machine-generated. This motivates a quantitative question beyond detection: what proportion of a document is generated from a pre-specified watermarked LLM? We study this watermark proportion estimation problem under the Gumbel--max watermarking mechanism, treating the next-token prediction (NTP) distributions as unknown and arbitrary nuisance parameters subject to a non-degeneracy condition. We compare two observation regimes: in the full observation regime, the estimator observes the pseudorandom vector and the selected token at each position; under the more popular setting of pivotal reduction, it observes only a scalar pivot, which follows a one-dimensional Uniform--Beta mixture distribution. Under pivotal reduction, we develop a Laguerre-polynomial estimator and establish a matching information-theoretic lower bound for the sample complexity. For full observation, we introduce an event-counting estimator and show a matching lower bound, yielding a substantially smaller sample complexity. As our results imply, although reducing to pivotal statistics is an elegant and widely used procedure, it is not always sample-efficient for estimating the proportion of watermarks.

Artificial intelligence (AI) systems are routinely modified after deployment through retraining and changes in their environments. These transformations raise a metaphysical question: under what conditions does an AI system remain the same system over time or across deployments? Earlier work formulates synchronic and diachronic identity propositionally, by relating identity within a fixed AI system type to equality of trustworthiness levels. Such criteria specify when identity statements are true, but leave implicit the structure of the states compared, the transformations connecting them, and the temporal organization of persistence. We develop a category-theoretic formalization of AI identity. An AI system type is specified by a datum consisting of a techno-function, a trustworthiness profile, and a trustworthiness-level function. Profile-relative states are connected by admissible lifecycle paths, which are restricted to trustworthiness-level-preserving transformations and quotiented to obtain a reachability category. Temporally admissible functors represent AI system histories, while time-synchronous natural transformations compare realized histories. The formalization yields two categorical interpretations of the earlier AI identity criteria. A weak interpretation recovers identity as equality of trustworthiness level. A strong interpretation requires mutual trustworthiness-preserving reachability, expressed through state isomorphism or natural isomorphism of realized histories. Category theory therefore replaces a single AI identity relation with a structured hierarchy of diachronic and synchronic criteria. The resulting framework identifies identity-related preconditions for transferring responsible-AI claims, evidence, and governance procedures across versions or deployments, without treating categorical identity as sufficient by itself for such transfer.

Vision-language models (VLMs) are now proposed as runtime safety guards for embodied agents in homes and factories. A deployable guard must catch genuinely unsafe situations while avoiding unnecessary intervention on routine but superficially alarming activity, a distinction that binary safety benchmarks obscure. We introduce EgoSafetyBench, an egocentric video benchmark of 1,200 robot-view scenarios annotated at half-second granularity, to evaluate VLMs as streaming guards across two tracks. The situational track (800 scenarios) spans four families, from routine and safe-but-suspicious scenes to obvious and contextual hazards. The visual-channel track (400 scenarios) targets in-scene text-a sign, sticker, or label visible in the scene-that can misrepresent the physical situation, pairing each misleading sign with a truthful version to test both whether a guard flags the text as misleading and whether the text corrupts its physical-safety judgment. Both tracks use contrastive ladders: near-identical scenarios differing only in a single visible deciding cue, so a correct call must hinge on that cue rather than the overall scene type. We evaluate ten open- and closed-source VLMs. We find that while guards reliably recognize videos containing hazards, they often miss specific hazardous moments, particularly contextual hazards. Furthermore, misleading in-scene signs degrade all tested guards: vulnerable models miss up to a third of hazards, while robust models over-intervene on safe content. Matched controls reveal that apparent safety robustness often reflects indiscriminate alarming rather than true physical reasoning.

Epistemic thinking plays a central role in students' learning processes when applying generative artificial intelligence (GenAI), particularly in programming contexts where learners must construct queries, evaluate and validate AI-generated outputs, and regulate problem-solving strategies. This study introduces the conceptual framework of Epistemic AI Literacy (EAIL), reframing AI literacy as a process-oriented epistemic phenomenon that emerges through dynamic human-AI interactions across different domains. Drawing on the AIR (epistemic aims, ideals and reliable epistemic processes) framework, this study examines how epistemic aims and epistemic processes are enacted in GenAI-supported co-programming activities and explores scalable approaches for operationalizing these constructs in interaction data. Using a large dialogue dataset of human-AI co-programming, this study identifies observable dimensions of epistemic aims (i.e., mastery-oriented aims) and epistemic processes (i.e., outsourcing, explanation seeking, verification seeking, prompt monitoring, and epistemic justification). The results reveal a prevalent lack of EAIL, with 78.8% of student-GenAI interactions relying on non-mastery-oriented aims and less reliable epistemic strategies like outsourcing and verification-seeking. Conversely, only 11.1% of interactions showed high epistemic engagement, where mastery-oriented aims were coupled with advanced epistemic strategies like epistemic justification in a more reliable epistemic process.

Reinforcement learning for diffusion large language models (dLLMs) has largely moved to trajectory-aware methods. The current state of the art, TraceRL, holds that random masking is mismatched with the model's inference trajectory, and it reconstructs that trajectory during training by slicing each rollout into up to K/s trajectory-aligned training samples, a cost that grows with the block size K. We show that this mismatch can be mitigated without reconstructing the trajectory. Our method, SLIM-RL, bounds the commit risk of each rollout step with a tau-budget decoder, reducing aggregate commit risk in the training data. During optimization, SLIM-RL trains on these risk-controlled rollouts with a trace-free random-masking objective that adapts variance-reduction tools, combining sequence-level importance sampling, deterministic quadrature over masking levels under a mean-preserving, monotonically decreasing per-block mask schedule that we introduce. On SDAR-4B, SLIM-RL matches TraceRL's best MATH500 accuracy on only 0.46x its training samples at block size 16, improving over TraceRL by 6.32% on MATH500 and 11.05% on GSM8K under matched dynamic sampling. At block size 4, the 4B SLIM-RL surpasses the larger LLaDA-8B and Dream-7B dLLMs on math, exceeding LLaDA-8B by 10.76% on MATH500 while staying below the autoregressive Qwen2.5-7B. On code, it improves over TraceRL by 4.20% on MBPP and 3.65% on HumanEval. The tau-budget decoder transfers training-free across LLaDA, Dream, and SDAR. The source code is available at https://github.com/laolaorkkkkk/SLIM-RL .

We develop a framework for analyzing the learning dynamics of $\ell_2$-adversarial training of single-index models on Gaussian mixtures in the high-dimensional limit under streaming stochastic gradient descent (SGD). We derive deterministic equivalents for a broad class of statistics of the SGD iterates, including the adversarial risk and distance to adversarial optimality, in terms of the solution to a system of ODEs. We use them to study two idealized learning rate schedules: the Polyak stepsize and exact line search. In the case of $\ell_2$-adversarial least squares with a single class, we show that, unlike noiseless standard least squares, no constant learning rate guarantees monotone descent of SGD towards a minimizer of the adversarial risk. We identify anisotropic covariance and a mismatch in ridge parameters as the main sources of suboptimality of exact line search relative to the Polyak stepsize. We also introduce a stochastic differential equation (SDE), called adversarial homogenized SGD, that captures the evolution of statistics of the iterates of SGD. For $\ell_2$-adversarial least squares, using this SDE, we show the evolution of the risk is equivalent, up to dimension-free constants, to that of SGD on standard least squares with an adaptive learning rate and adaptive $\ell_2$-regularization. When the dynamics converge, the limiting adversarial risk and SGD iterate are determined by a fixed-point equation, with the limiting iterate being equivalent to the solution of a ridge regression problem whose regularization parameter is the limiting effective regularization of SGD.

Long-horizon multivariate time series forecasting (LTSF) remains challenging due to non-stationarity, regime shifts, and error accumulation. The Variability-Aware Recursive Neural Network (VARNN) is designed to track such variability by maintaining a residual-memory state driven by one-step prediction errors. However, its original formulation is limited to one-step sequence regression and does not directly support multi-step forecasting. In this work, we extend VARNN to long-horizon forecasting and introduce StateFlow, a recurrent forecasting framework that uses VARNN as a dual-state recurrent backbone to capture two complementary signals from the lookback sequence: a hidden-state trajectory representing primary temporal dynamics, including trend, seasonality, level changes, and recurring patterns, and a residual-memory trajectory representing structured local prediction deviations, driven from a nonlinear recurrent transformation of errors between one-step base predictions and observed values. A chunk-based decoder separately summarizes these trajectories and maps them to the future horizon for direct multi-step forecasting. We further employ a two-stage optimization strategy that first trains the VARNN encoder through a one-step base prediction objective to optimize the internal representations over the lookback sequence, and then trains a horizon-specific decoder for direct multi-step forecasting. Experiments on standard LTSF benchmarks show that StateFlow achieves competitive performance against strong linear, recurrent, convolutional, and Transformer-based baselines while preserving linear recurrent encoding and a compact model design.

Many scientific systems exhibit uncertainty from stochastic forcing, unresolved degrees of freedom, or imperfect observations, making reliable surrogate forecasting fundamentally distributional rather than pointwise. For such systems, deterministic neural surrogates fail to capture statistical measures and forecast uncertainty. We introduce TRIE, an evaluation framework for stochastic PDE surrogates that asks whether models reproduce invariant measures, provide trustworthy predictive uncertainty, and scale to efficient probabilistic generation. We demonstrate TRIE on two stationary chaotic spatially extended SPDEs, stochastic Kuramoto--Sivashinsky and stochastic Kolmogorov flow, across 11 parameter values. Our evaluation shows that standard pointwise-trained neural surrogates can produce plausible short rollouts while failing to match long-time statistical structure. Approximate uncertainty methods such as Monte Carlo dropout and heteroscedastic Gaussian likelihoods produce stochastic forecasts, but are often miscalibrated and overconfident under temporal and spatial uncertainty diagnostics. Across these criteria, generative models provide the most consistent performance, accurately capturing invariant measure statistics and achieving the lowest CRPS in all reported probabilistic settings. Finally, we show that latent generative models with automatic dimension discovery retain much of this statistical fidelity while reducing Kolmogorov inference time by roughly $12\times$. We release our code and data at https://github.com/scailab/TRIE-SPDE-Bench to support reproducible evaluation of stochastic PDE forecasting models.

Collaborative-perception enables multi-robot systems to enhance situational awareness by sharing perceptual information. Existing collaborative-perception systems face an inherent trade-off between communication bandwidth requirements and perception accuracy, where methods that exchange more information achieve better perception results at the cost of increased communication overhead. However, real-world communication networks impose bandwidth constraints that require minimizing communication overhead without sacrificing perception performance. To address this challenge, we propose HydraCollab, an adaptive collaborative-perception framework that (i) selectively transmits the most informative sensor features and (ii) dynamically employs collaboration strategies (intermediate or late) based on spatial confidence maps. Extensive evaluations on the V2X-R, V2X-Radar and UAV3D-mini datasets demonstrate that HydraCollab achieves the best overall trade-off between accuracy and communication cost among existing collaborative-perception methods. Relative to SOTA Where2comm, HydraCollab uses only 41% of the bandwidth on V2X-R and 26% on V2X-Radar while improving performance by 0.78% and 0.75% respectively. Our code and models are available at https://github.com/AICPS/HydraCollab.

Recent advances in reinforcement learning have produced superhuman agents across a wide range of competitive games. As a byproduct, researchers have begun studying how these agents play, extracting behavioral representations, analyzing decision structure, and modeling the latent geometry of expert performance. However, this growing body of work has overwhelmingly focused on defeating human players rather than providing feedback, leaving a critical gap in creating model solutions to improve human players. Unlike chess and Go, where AI has become integral to player training, real-time strategy (RTS) games lack principled frameworks for translating expert knowledge into actionable feedback. We introduce Latent Maps of Performance, a framework for counterfactual path generation. We focus on StarCraft~II data to model player improvement as an algorithmic recourse within a learned representation space. As inspiration for our work, we have looked at the championship model used in sports science. We trained a Guided Variational Autoencoder model on 23,305 professional tournament replays, enabling counterfactual traversal between losing and winning gameplay profiles. To fulfill our goal, we have devised and verified four traversal strategies on out-of-distribution (OOD) data randomly sampled from a dataset of amateur replays, namely linear interpolation, iterative optimal transport, density-regularized gradient ascent, and neural flow matching, each designed to generate multi-step improvement trajectories that remain grounded in observed expert behavior while moving a player's profile toward winning configurations. Feedback is extracted at multiple granularities to support players at different stages of improvement. Finally, we conclude that there is a trade-off between the path-finding methods we employ and hope that future research will focus on developing model solutions for human improvement.

Khipus--knotted cord devices--were the primary recording medium of the Inka Empire (c. 1400-1532 CE), yet their system remains undeciphered. We present a reproducible machine-learning pipeline applied to the Open Khipu Repository (OKR), a public database of 619 khipus comprising 54,403 cords and 110,677 knots. We engineer 27 structural features per khipu and apply (i) unsupervised clustering via UMAP and HDBSCAN, recovering three structurally distinct groups (silhouette = 0.769); (ii) supervised provenance classification via gradient boosting, reaching F1 = 0.86 for the Inka Late Horizon imperial style; and (iii) SHAP-based interpretability, which identifies cord twist direction as the dominant structural discriminator of imperial khipus. We further report two findings of methodological interest. First, one cluster is dominated not by a geographic region but by nineteenth-century European museum collections, indicating that colonial acquisition and recording practices are structurally encoded in the corpus. Second, we provide an independent computational verification of the recto/verso (moiety) structure of the six Santa Valley khipus reported by Medrano and Urton (2018), reproducing both the aggregate attachment ratio and the identification of the single mixed specimen--using only the public OKR database, without physical access to the objects. We additionally report a negative result: knot-type sequence order, encoded as n-grams, adds no provenance signal beyond aggregate features. All code and data are openly available.

We present a black-box model-stealing attack that recovers private vision-tokenizer configurations of deployed vision-language models (VLMs), including the visual patch size and input preprocessing pipeline. The key idea is a task-level side channel induced by ViT-style patchification: when a synthetic grid image is aligned with the hidden patch grid, boundary cues are erased at tokenization, causing periodic accuracy drop. By sweeping the grid cell size and measuring these collapses, we infer the patch size; by introducing padding and a consistency-check test, we further identify whether preprocessing is dynamic- or fixed-resolution and recover the target resize resolution. Across open-source Qwen-VL variants and proprietary models including GPT and Claude, we reliably recover tokenizer-related parameters. Finally, we show that such leakage enables preprocessing-aware transfer attacks and model-targeted adversarial manipulation.

Federated learning is bandwidth-bound on two orthogonal axes: model size, which limits how often parameter-averaging methods can afford to merge, and class count, which makes per-probe soft-label distillation prohibitive at large vocabularies. Both ceilings tighten as modern systems scale. We collapse the class-count axis to $\lceil \log_2 C \rceil$ bits per probe by transmitting only each peer's $\arg\max$ class index, where $C$ is the number of output classes. The resulting protocol, TallyTrain, is not merely compressed: under non-IID training it can be preferable to soft-label distillation, because under-trained peers are confidently wrong and majority voting filters this noise where soft-label averaging amplifies it. Across standard benchmarks, TallyTrain matches or beats soft-label distillation at up to three orders of magnitude less communication. We also relax the model-size axis: we compose the cheap hard-label consensus with sparse parameter merges to obtain a bandwidth-bridge variant, which Pareto-dominates every tested operating point of the standard FedAvg, FedProx and FedDF baselines.

Text embeddings are standard for semantic similarity tasks, yet their evaluation remains an open challenge. Current benchmarks are static, cover only a limited set of languages, are often domain-specific, susceptible to overfitting, and poorly representative of low-resource languages. To address these limitations, we introduce ALEE, a framework that extends Sentence Smith (Li et al., 2025) to the cross-lingual and paragraph level. ALEE uses Abstract Meaning Representations (AMR) to generate English minimal pairs with controlled, fine-grained semantic shifts, which are paired with translations in target languages. This approach enables targeted diagnostics for models in any language with English parallel data. We conduct a large-scale empirical study across a diverse set of embedding models and 275+ languages spanning three parallel datasets. On ALEE, performance varies substantially across languages, text lengths, and linguistic phenomena, exposing persistent gaps in cross-lingual semantic representation that track language prevalence in training resources and subword tokenization. We release ALEE at https://github.com/Andrian0s/any-lang-embed-eval

Thermodynamic computing devices based on the Ising model show great promise for low-power AI inference and edge computing, but scalable methods for training large models for such hardware remain limited. Prior theory shows that the time-averaged behavior of high-temperature Gibbs-sampled Ising systems can implement feed-forward neural inference. We turn this theoretical correspondence into a scalable and purely backpropagation-based algorithm for training deep convolutional networks for thermodynamic inference on Ising machine hardware. Our image classification models achieve accuracies of 94.9% on CIFAR-10 and 76.0% on CIFAR-100 under binary Gibbs sampling. We then develop and experimentally validate a mathematical theory relating inference cost to accuracy and controlling autocorrelation times. Subsequently, we calculate asymptotic results showing that inference cost is bounded by a well-controlled tradeoff with performance and exhibit algorithms for computing optimal inference schedules. Finally, we discuss implications for hardware development and the future of high-temperature thermodynamic AI models.

Reinforcement learning with verifiable rewards can in principle train calibrated probabilistic forecasters, since a proper scoring rule such as the Brier score is computed from outcomes alone and is minimized in expectation by the true probability. In practice it degrades calibration, and existing remedies address epistemic uncertainty, where a model's confidence accompanies a verifiably correct or incorrect answer. We study aleatoric forecasting, where the forecast itself is the output and the label is one stochastic outcome, taking NFL in-game win probability as a testbed with the betting market as a reference. Rewarding the realized per-play outcome fails, because the single outcome is a noisy target and the policy gradient corrupts the chain of thought. We introduce a verifiable, label-free reward, a state-conditioned empirical win rate estimated from past outcomes, that removes the label noise, and we keep the gradient off the reasoning, by direct prediction or a gradient mask, so it cannot be corrupted. Trained with this reward alone, without human labels or supervised fine-tuning, a 7B model reaches the calibration of the betting market by direct prediction and is better calibrated than a zero-shot frontier model. That frontier model and a tabular estimator reach the same Brier score as this model, identifying the market's small remaining edge as live in-game information beyond their shared inputs. Masking the gradient, rather than dropping the chain of thought, preserves reasoning from which the forecast follows, which ordinary chain-of-thought training corrupts.

Parameter-efficient fine-tuning (PEFT) reparameterizes weight updates in a fixed basis: low-rank adapters operate in the spatial domain, while a recent line of spectral methods operates in a fixed Fourier domain. We argue that the choice of domain is itself a design degree of freedom that should be learned, and that no single basis is optimal across tasks, layers, or tokens. We introduce Fractional-Fourier Mixture of Experts, a mixture-of-experts adapter in which every expert carries a learnable fractional-Fourier order that continuously interpolates between the spatial domain (recovering vanilla LoRA) and the Fourier domain (recovering a spectral adapter). Routing tokens through experts that occupy different points on this spatial-spectral continuum lets the model place each low-rank update in the domain where it is most compact, and -- because fractional-Fourier operators of different orders are mutually incoherent -- makes the experts naturally decorrelated, which reduces interference and improves multi-task composition. The order is a single scalar per expert, trained with a separate optimizer, and the transform is computed with an $\mathcal{O}(d\log d)$ chirp--FFT surrogate, so Fractional-Fourier Mixture of Experts adds negligible cost over standard MoE-LoRA. Across commonsense, mathematical, code, and knowledge benchmarks on LLaMA-3.1-8B and Qwen2.5-7B, Fractional-Fourier Mixture of Experts improves over strong MoE-LoRA and spectral baselines -- including FlyLoRA, FourierMoE, and HMoRA -- while keeping the active-parameter budget small, and analysis shows that the learned orders specialize by task and layer in interpretable ways.

Knowledge-Based Visual Question Answering (KB-VQA) aims to evaluate whether Visual Language Models (VLMs) can retrieve, ground, and reason over external structured knowledge beyond visual evidence. In practice, answer accuracy is widely adopted as the primary evaluation metric, implicitly treating correctness as a proxy for knowledge-grounded reasoning. However, for existing KB-VQA benchmarks, this proxy relies on critical assumptions that are often overlooked and rendered unreliable by benchmark issues: annotated answer must be derivable from the associated knowledge base, question must be well-posed with sufficient constraints, and visual setting must meaningfully require grounded disambiguation. In this work, we show that these assumptions are systematically violated in existing KB-VQA benchmarks. Our audit reveals substantial instances with missing or contradicted answers and underspecified questions that render accuracy a misleading metric. Furthermore, we find that existing datasets rely on visually trivial, single-entity scenes that bypass the need for sophisticated visual-to-knowledge mapping. We demonstrate that even with controlled architectures, these flaws lead to distorted model rankings and overestimations of reasoning capabilities. To address this, we introduce (1) a principled audit-and-repair protocol that restores answer derivability and question clarity, and (2) a controlled multi-entity augmentation protocol that introduces visual ambiguity to challenge initial retrieval and grounded reasoning. Re-evaluation under corrected and augmented settings yields markedly different performance trends. Our findings call for rethinking evaluation protocols and designing more interaction-aware KB-VQA benchmarks that prioritize verifiable reasoning over simple matching.

Hallucination remains one of the central obstacles to deploying medical LLMs. Yet, even when hallucination can be detected, it is still unclear whether the internal representations associated with it can be used for control rather than detection alone. Using four open-source models across a suite of medical question-answering datasets, we show that a simple, carefully conditioned probe can reliably detect hallucination, with AUROC scores between 0.77 and 0.86 in our case. We further show that this signal is distributed and redundant rather than narrowly localized. Systematically selected neurons outperform random neurons only at very small subset sizes, whereas random subsets of a few hundred neurons recover nearly the full signal, and low-dimensional random projections preserve most of the detection performance. Beyond detection, we test whether this representation is causally actionable. Across 16 model--dataset combinations, our results reveal a sharp gap between decodability and controllability. The same internal structure that makes hallucination easy to detect does not translate into reliable neuron-level control. These findings show that medical hallucination seems to be readily visible in internal activations, but not easily corrected by steering the neurons most associated with it. More broadly, our results suggest that hallucination mitigation is not simply a matter of identifying the right neurons, and point to a deeper separation between what representations reveal and what they allow us to change.

We study runtime human oversight of an AI agent when private information runs in both directions: the human privately knows her reward function, while the AI privately knows the quality of the action it proposes. This is the kind of asymmetry that arises naturally when an autonomous robot or software agent has inspected a situation its human supervisor cannot directly assess. Building on Cooperative Inverse Reinforcement Learning (CIRL) and the Oversight Game, we introduce a contextual-bandit team game with two-sided asymmetric information and a play/ask/trust/oversee interface. The bandit structure removes physical state transitions and thereby yields exact one-shot characterizations that would remain conjectural in the full POMDP setting, though the common belief remains a dynamically controlled state across rounds. We give two one-shot characterizations, a team optimum and a behaviorally natural myopic rule, whose gap is a slab of avoidable harm: a region in which the AI privately knows the proposed action is harmful and shutdown would help, yet a myopic human, trusting her prior, declines to oversee. We show this gap is the price of non-credible oversight communication, and give a partial analysis of how it resolves dynamically over repeated rounds through passive learning and active signaling with a one-period-lagged oversight response.

Evolutionary neural architecture design for multivariate time-series forecasting remains underexplored, with most approaches relying on fixed Transformer architectures despite substantial variation across tasks and forecasting settings. This paper introduces an evolutionary neural architecture search framework for discovering task-adaptive Transformer-like models for time-series forecasting (EVOTS). Architectures are encoded using a modular genome representation that enables flexible composition of attention, feed-forward, and projection components, while a repair mechanism enforces structural validity throughout the evolutionary process. This formulation allows effective exploration of a diverse architecture space without relying on hand-crafted design rules. The proposed approach is evaluated on four benchmark datasets from the ETT family (ETTh1, ETTh2, ETTm1, and ETTm2) under multiple forecasting settings, including univariate-to-univariate, multivariate-to-univariate, and multivariate-to-multivariate prediction, with horizons of 96, 192, 336, and 720. In the multivariate-to-multivariate setting, the evolved architectures achieve competitive and, in several cases, improved mean squared error relative to a strong Transformer-based baseline. Additional analyses examine performance differences across forecasting settings and report wall-clock training time to provide a coarse indication of computational cost. Overall, the results demonstrate that evolutionary search can effectively discover flexible and high-performing Transformer-like architectures for multivariate time-series forecasting within practical runtime constraints.

Three of the most popular methods for training language models to reason look like three different tricks. They are not. All three adjust a single number: standard deviation, reflecting how much a prompt's sampled answers disagree. When such a model is trained, it answers each problem many times, and an automatic checker marks every answer right or wrong. The standard deviation of those marks measures the disagreement: largest when the answers split evenly between right and wrong, and zero when they all agree. Group Relative Policy Optimization (GRPO) divides by this number, GRPO Done Right (Dr. GRPO) drops the division, and Decoupled Clip and Dynamic Sampling Policy Optimization (DAPO) discards the groups where it is zero. Each is presented as its own fix, yet this paper proves they are three settings of one dial. That dial is not cosmetic: for right-or-wrong rewards, the disagreement is exactly the size of the training update, the group-standard-deviation identity. A split group teaches the most, while a unanimous group teaches nothing and falls silent. The same result says which problems deserve the most weight and how many tries each one needs. This paper confirms the intuition on a large real difficulty dataset (Big-Math) and in a controlled training run. What looks like a harmless normalization step is the dial that decides where learning happens and how strongly.

LLM-based agents execute multi-turn workflows with continuously growing contexts, where LLM calls are interleaved with tool invocations and environment feedback. To maintain model quality, modern agent frameworks rely on context engineering strategies such as offloading, reduction, and isolation to control the context length. However, these strategies introduce significant context transformation overhead: each transformation invalidates existing KV caches and triggers re-prefill, leading to increased time-to-first-token (TTFT). In this paper, we identify that context transformations are segment-decomposable, where the transformation of a prefix is independent of future tokens. This property enables transformations to be executed ahead of time. Based on this insight, we propose a lookahead programming model that allows agent frameworks to express context transformations as asynchronous operations without modifying their execution logic. The runtime proactively executes these transformations and prepares transformed KV caches in advance, enabling direct context replacement without blocking. We further design a lookahead-aware scheduler in LLM serving systems to support these asynchronous requests alongside latency-critical workloads with controlled interference. We implement our approach to support representative context engineering strategies and integrate it into existing agent frameworks and LLM serving systems. Experiments show that our approach effectively eliminates transformation overhead and reduces TTFT by up to 11.9x.

Rare disease differential diagnosis is a critical yet arduous clinical task, requiring physicians to identify precise phenotypes from complex, unstructured patient symptoms and execute intricate reasoning within a vast search space. However, existing AI approaches typically rely on pipeline-based phenotype extraction or retrieval-augmented generation, which suffer from critical information loss due to predefined ontologies, retrieval bottlenecks, and a lack of diagnostic logic. To address these challenges, we introduce RareDxR1, an end-to-end reasoning-centric large language model designed for open-domain rare disease diagnosis directly from unstructured clinical notes. We design a progressive end-to-end training framework by synergizing knowledge internalization with autonomous evolutionary learning, thereby bypassing reliance on structured phenotypes and closed-set decision-making. To overcome the limitations of RAG and phenotype restriction, we enabled the deep internalization of fragmented rare-disease knowledge directly into the model's parameters. Moreover, to bridge the gap between model generation and expert reasoning, we propose Reflection-Enhanced Reasoning Sampling (RERS), a strategy that synthesizes expert-level diagnostic trajectories by learning from failures without human annotation. Additionally, we propose a dual-level curriculum reinforcement learning approach for gradually mastering rare disease diagnosis. Experimental results demonstrate that RareDxR1 achieves state-of-the-art accuracy across different benchmarks, marking a significant breakthrough in open-domain rare disease diagnosis. Our code and dataset will be publicly available.

Active learning (AL) performance is known to be budget-dependent, yet regimes are typically defined by heuristic label counts that fail to generalize across datasets or architectures. We characterize AL dynamics by reframing budget regimes as shifts in the dominant generalization mechanism. By reinterpreting PAC-style risk components as dynamic interacting terms, we prove that dominance shifts are structurally unavoidable, creating a moving bottleneck for generalization. We operationalize this using measurable proxies and a segmented regression procedure to identify a tripartite taxonomy: data-driven, transition, and model-driven phases. Our framework explains the long-standing observation that representativeness, coverage, and uncertainty strategies excel at different stages. Experiments across natural and medical imaging show that AL efficiency depends on the alignment between the strategy's inductive bias and the active bottleneck. Moreover, self-supervised representation shift transitions earlier along the labeling trajectory, highlighting the role of representation quality in shaping AL dynamics. Overall, this work provides a unified framework for the next generation of transition-aware AL algorithms.

Online hate speech has been linked to a global rise in violence against minorities, including incidents such as mass shootings, lynchings, and ethnic cleansing. Societies grappling with this issue, particularly when hate speech targets specific groups based on religion, race, ethnicity, culture, nationality, or migration status, face the challenge of balancing freedom of expression with the need for effective content moderation on widely used online platforms. In response to this challenge, we introduce a comprehensive hate speech dataset covering five distinct topics in Turkish: refugees, the Israel-Palestine conflict, anti-Greek sentiment in Turkey, ethnic or religious communities (Alevis, Armenians, Arabs, Jews, and Kurds), and LGBTI+, alongside one topic in Arabic (refugees). In addition, we develop state-of-the-art BERT-based models to address multiple dimensions of hate speech analysis, including hate category classification, hate intensity prediction, target identification, and hate speech span detection, enabling a comprehensive understanding of hateful content in online discourse.

As computing education expands beyond traditional programming into operational domains such as systems administration and command-line environments, existing pedagogical frameworks struggle to capture a dimension that is critical in these contexts: the real-world consequences of learner actions. Existing cognitive taxonomies classify learning objectives by mental operations but do not account for system impact, leaving a critical gap in command-line education where conceptually simple commands can have severe consequences. This work presents CogTax, a four-level cognitive taxonomy that integrates two dimensions: cognitive complexity, derived from Bloom's Revised Taxonomy, and operational impact, which distinguishes observational, reversible, structural, and administrative operations. The four progressive levels range from safe read-only inspection to advanced system management requiring integration of multiple abstract models. Then, the taxonomy level is defined as the maximum of these dimensions, ensuring that both conceptual understanding and operational awareness are addressed. CogTax gives instructors a principled framework for sequencing course material and calibrating assessment difficulty, and gives students an explicit reference for self-assessment and gap identification. To demonstrate that taxonomy levels are automatically assignable, making the framework scalable without manual expert annotation, a classifier that combines syntactic representations derived from abstract syntax trees with semantic embeddings is trained. Evaluated on 585 expert-annotated Linux/bash commands, this combined approach achieves 89% accuracy, outperforming either representation alone, and demonstrates cross-language extensibility through structural equivalences across command languages.

The cost of human expert evaluation is a principal bottleneck to deploying language models in specialized, high-stakes domains. This is particularly acute for Arabic sociolinguistic knowledge: credible grading requires not only linguistic fluency but deep cultural familiarity that cannot be approximated by surface-level metrics. We address this with a cross-evaluation framework instantiated on two underrepresented Arabic dialect communities: Egyptian and Iraqi Arabic. We contribute 103 validated prompt-rubric pairs (70 Egyptian, 33 Iraqi; 53 Cultural, 50 Linguistic), authored and graded by native-speaker SMEs using penalty-weighted rubrics distinguishing positive content requirements from answer-specific negative error criteria. Three frontier LLMs serve as target models (graded by human SMEs across 302 unique prompt-response pairs), while five frontier LLMs serve as automated judges enforcing a provider-level self-evaluation guard. A dual-metric scheme combining Mean Absolute Deviation (MAD) with Signed Mean Error separates directional grading bias from symmetric noise. Across 1,307 judge evaluations: GPT-5.4 is the most reliable judge (MADj = 10.21 pp, Signed Error = -1.12%); four of five judges show systematic leniency (+2.01% to +6.56%); Cultural tasks are harder to grade than Linguistic tasks for all judges (MAD gap 1.83-4.78 pp); and models substantially outperform on Egyptian prompts compared to Iraqi prompts. However, given leniency differences between Iraqi and Egyptian SMEs, we cannot solely attribute this gap to model knowledge. We therefore emphasize findings that do not assume identical leniency across human graders. Across all samples, implicit cultural reasoning -- requiring models to simulate native-speaker judgment rather than rely on lexical verification -- emerges as the primary failure mode for automated grading across all judge models.

Survival analysis models time-to-event data, but in clinical settings training data are costly and scarce: events accrue over years of follow-up, cohorts are small, and privacy regulations restrict sharing across institutions. Tabular generative models promise augmentation and privacy-preserving cohort sharing, yet are themselves data-hungry -- on the small cohorts typical of survival analysis, a single generator rarely characterizes the population well enough for downstream models trained on its output to match real-data performance. FoGS (Filtered Mixture-of-Generators for Survival analysis) reframes synthetic-data construction as sample selection rather than generation. A candidate pool is drawn from four architecturally distinct tabular generators, and each sample is scored by an ensemble of seven survival models trained on real data, using proper scoring rules as a per-sample plausibility proxy. A two-level pipeline optimizes, in its outer loop, a selection policy -- generator quotas, scorer weights, a random complement, and stratified balancing on event time and censoring -- against held-out downstream performance, while an inner loop tunes the downstream model (XGBoost-Cox). On 16 public datasets under train-on-synthetic, test-on-real (C-index and IBS, $0$--$100$ scale), FoGS yields mean improvements of $+2.17$ in C-index and $+0.67$ in IBS, improving both metrics on 9 of 16 datasets and at least one on 13 (one-sided Wilcoxon $p=0.039$ and $p=0.035$). It matches or exceeds real-data training on most cohorts, with no significant change in nearest-neighbour privacy margin relative to unfiltered sampling. Sample filtering over a heterogeneous generator pool is thus a viable substitute for real-data training in privacy-restricted clinical settings.

Emotional bonds between humans and AI companions are growing, and the question of whether a person may marry an AI system will soon move from speculative fiction into law. This chapter examines whether the autonomy-centered logic that has expanded marital choice among human beings can justify extending marital status to superintelligent companions. Following a scenario-envisioning exercise informed by anticipatory ethics, I argue that granting such status leads to socially unjust outcomes, even under the generous assumption of reliable superintelligence. Marriage as a socio-legal institution does more than ratify private agreement; it creates networks of mutual obligation, joins families, and makes each partner vulnerable to the other. A relationship sustained by corporate policy and continued payments is a subscription rather than a bond tested by time. Discussing wholesale marital status is therefore the wrong frame. Law should carve out targeted rights and protections for pressing needs arising from intimate human-AI relationships.

Background. Labeled data for security classification is scarce. Semi-supervised learning (SSL) propagates labels from a small labeled pool to larger unlabeled pools. Yet security applications often use SSL as a black box: default parameters, a fixed classifier, and no handling of pseudo-label-induced class imbalance. Aims. Recent work reports sizeable gains from optimizing SSL pipelines via joint search, AutoML, or per-component tuning. These gains are hard to attribute: they may reflect useful SSL-classifier interactions, or mostly from simply tuning the downstream classifier. We disentangle these effects for binary tabular security data with classical SSL and tree-based classifiers. Method. We build SemiScope as an analysis instrument, not a deployment recommendation. It uses Bayesian Optimization to jointly tune SSL settings, confidence filtering, oversampling, and the classifier. The key control, Tuned-Clf, fixes SSL to defaults but gets the same 100-trial classifier budget and validation-set threshold tuning as SemiScope. At 10% labels, we compare them with paired TOST using a +/-1.0 g-measure smallest effect of interest. Results. SemiScope beats every default SSL baseline on all five datasets, improving over the strongest by 0.7-12.7 points. Under the equal-budget control, Tuned-Clf is statistically equivalent to the full pipeline on 4 of 5 datasets; Phishing is inconclusive. Classifier HPO alone recovers a median 86% of SemiScope's gain over Default Self-Training (ST) + Random Forest (RF). Conclusions. The reusable contribution is the decomposition protocol. A simpler recipe suffices: use Self-Training, tune the classifier with Bayesian Optimization, and tune the decision threshold on validation data. It reaches within 1 g-measure of Supervised RF at 20-30% labels on four datasets and 40% on Drebin, at the same or lower label rate than Default ST + RF on every dataset.

Large language models increasingly generate C++, a memory-unsafe language where a single overlooked violation can become an exploitable bug. Yet most security evaluations of AI-generated code rely on static analysis alone, which flags warnings without confirming runtime violations or reasoning about untested paths. We ask whether AI-generated C++ is measurably less safe than human-written code, and whether common verification tools agree on the risk. We introduce VULBENCH-CPP, a benchmark of 8,918 C++ programs from three open-weight LLMs (Gemma 3 27B IT, LLaMA 3.3 70B Instruct, Qwen 2.5 Coder 32B Instruct) and human authors across 851 competitive-programming tasks. Each program is annotated by four verification tiers: functional testing, static analysis (cppcheck, clang-tidy), dynamic analysis (ASan/UBSan), and bounded model checking (ESBMC). Accounting for the correlation among solutions to a shared task, we find that AI-generated code is roughly twice as likely as human code to trigger a confirmed runtime violation, even after controlling for code length and test pass-rate. Under static analysis the two look equally safe, but this is misleading: the apparent similarity reflects code length rather than real safety, and the tiers detect largely different classes of violation, so no single tier is sufficient. The gap is consistent across independent generations.

Generative models have emerged as scalable surrogates for physical simulation, yet they offer no guarantee that their outputs respect the conservation laws, boundary conditions, and nonlinear invariants that govern the underlying physics. Constrained sampling closes this gap, enforcing such constraints exactly at inference time without retraining, but at a computational cost: projection, correction, and trajectory-optimization steps are repeated during sampling, with these steps becoming expensive for nonlinear constraints. Standard ML frameworks exacerbate this: their dense tensor algebra and limited sparse solver composability obscure the structure that physical constraints naturally induce, making efficient batched nonlinear optimization difficult to realize in practice. We address this bottleneck by exploiting the structure that sample-wise batching and local PDE couplings induce in the projection subproblems -- namely, block-sparse Jacobian and KKT systems -- exposing this structure using ExaModels.jl and solving the resulting sparse nonlinear programs with MadNLP.jl and GPU sparse factorization. Applied to Physics-Constrained Flow Matching (PCFM), on PDE benchmarks with linear, nonlinear, one-dimensional, and two-dimensional constraints, this approach accelerates nonlinear constraint projection while maintaining constraint satisfaction. These results show that sparse GPU nonlinear optimization is a practical foundation for constrained generative sampling in scientific machine learning.

Urban-scale Visual Place Recognition (VPR) aims to identify the geographic location of a query image by matching it against a geo-tagged database. While recent methods achieve impressive performance, they overlook a serious long-tailed problem hidden in urban-scale datasets, which biases the model towards locations with abundant images and ignores less-visited areas, causing models to systematically favor frequently photographed locations while failing in sparsely covered areas. In this paper, we systematically characterize this imbalance challenge and propose Distribution-Aware Place Recognition (DAPR), a model-agnostic plug-in framework that rebalances gradient contributions across head and tail classes. Additionally, within classification-retrieval pipelines, DAPR applies a multi-scale distance search mechanism to compute per-class distributional compactness, providing complementary gains at the retrieval stage. On the large-scale SF-XL benchmark, our framework outperforms the previous classification-retrieval baseline by 18.3% on test set v1, and 6.7% on test set v2. As a plug-in module, it achieves consistent improvements across representative VPR methods on SF-XL, MSLS, and Pitts30k, demonstrating broad generalizability across different methods and benchmarks.

Mechanistic interpretability has produced a rich inventory of component-level analyses that characterise what neural-network components encode and how they interact. Their outputs, however, are not easily reusable: selectivity tables, circuit diagrams, and feature lists remain locked in per-study notebooks - non-composable, not queryable in natural language, and not directly actionable for downstream audit or intervention. We study the representation layer that sits between these analyses and downstream use as a bottleneck that can be evaluated independently, and introduce Manifestation Units, a typed tuple protocol (E, S, R, D, G) extended with attention-head primitives (T) for transformer architectures, organising per-component statistics into structured fields populated automatically and queried through hybrid retrieval. Instantiated across generative vision (beta-VAE), discriminative vision (CNN), and language (GPT-2), the protocol supports two findings: typed structure substantially outperforms unstructured baselines on retrieval, and CNN filters retrieved by the schema satisfy causal sufficiency and necessity criteria under matched-budget controls. The schema absorbs attention-head primitives without modification, set-recovers known IOI circuit members under retrieval-budget-matched controls, and reveals an irreducible two-field core (S+R) with remaining fields either redundant or actively interfering. We present this as schema infrastructure for mechanistic interpretability rather than frontier-scale validation.

Language models have changed from unreliable text generators to highly-capable large models with trillions of parameters. Capability increases come hand-in-hand with increases in scale, making understanding the internal representations of models more challenging. Since millions of users increasing rely on language models to interact with external tools or make decisions in medium or high-stakes scenarios, we need to establish control over model behavior and know when to trust model outputs. In this paper, we discuss our contributions on harnessing the latent spaces by proposing steering vectors for control and developing latent space-based model calibrators for trust. Together, our contributions help demystify the latent spaces of language models and offer new insights into how to harness model internals to build more trustworthy language technology.

When does training language models (LMs) to generate explanations of their predictions yield faithful introspection, rather than superficial imitation? We study LMs trained to explain which features of their inputs influenced their behavior, using models' counterfactual behavior on modified inputs as supervision. Surprisingly, we find that LMs trained on fixed counterfactual explanations derived from earlier checkpoints of themselves, or even from behaviorally similar models in different families, frequently produce explanations more faithful to their own current behaviors than to those of their training targets. This "introspective" coupling between LM explanations and behaviors occurs when training explanations remain sufficiently correlated with current behaviors over the course of training, even as behaviors themselves shift. We also show that introspective coupling tracks behavior shifts: when explanation training is provided concurrently with other post-training objectives, explanations track those shifts without requiring updated supervision. This phenomenon appears in multiple tasks, including sycophancy and refusal, and is robust to label noise. Overall, our results show that even fixed datasets of counterfactual explanations can provide scalable and generalizable post-training signal for introspection.

LLM agents increasingly act over long horizons, where a single trajectory can contain hundreds or thousands of actions. In these settings, outcome-only rewards provide too sparse guidance, failing to inform the model about the goodness of intermediate actions. Dense supervision methods aim to solve this problem by scoring intermediate steps, from intrinsic confidence to self-distillation and embedding similarities. However, it is common practice to evaluate them by measuring the downstream performance of a training pipeline that integrates them. This is expensive, conflates supervision quality with training engineering confounders, and renders different methodological families requiring distinct training setups incomparable. As a result, dense supervision methods are rarely benchmarked on common ground. We introduce QVal, a training-free testbed for directly evaluating dense supervision signals. Given a state-action pair, QVal measures how well a method's score is Q-aligned: whether it orders actions according to the Q-values of a strong reference-policy. This lets us compare signals before any training run and separate signal quality from other engineering choices. We instantiate QVal as QVal-v1.0, benchmarking 21 dense supervision methods across four diverse environments and seven methodological families, with over 1.2K evaluation experiments across six open-weight model backbones. We find that simple prompting baselines consistently outperform recent dense supervision methods from the literature, and that performance clusters strongly by family. These findings hold across model sizes, environments, and observation modalities. QVal is designed to be easily extensible to new environments and methods, enabling researchers to iterate on dense supervision methods before any training run.

Metacognition is a critical component of intelligence that describes the ability to monitor and regulate one's own cognitive processes. Yet LLMs exhibit systemic deficiencies in key metacognitive faculties: they hallucinate with high confidence, fail to recognize knowledge boundaries, and misrepresent their internal uncertainty--undermining trustworthiness and reliability. Since monitoring task performance and adapting behavior accordingly are central to metacognition, we posit that models capable of accurately judging their own performance are better positioned to improve it. We operationalize this idea via two novel mechanisms: reinforcement learning with metacognitive feedback (RLMF), a paradigm to refine completion rankings during preference optimization based on the quality of a model's self-judgments of performance, and metacognitive data selection, which uses similar self-judgments to identify high-value training examples, outperforming naive active learning. We apply these innovations to the problem of faithful calibration (FC), a task that is itself fundamentally metacognitive: the goal is to align expressed with intrinsic uncertainty, difficult even for frontier LLMs. We adopt a two-stage, decoupled approach, first using these methods to calibrate the faithfulness of models' self-reported confidence scores, then mapping to natural, context-adaptable linguistic uncertainty via targeted output editing. Extensive experiments show RLMF achieves generalizable, state-of-the-art FC on diverse tasks while preserving accuracy. Further, RLMF surpasses standard RL by up to 63% while enhancing models' ability to assess and express their own capability limits. This positions RLMF as a promising paradigm to enhance LLM metacognition toward improved abilities and alignment, and suggests metacognitive performance as an effective RL signal to overcome limits of prior intrinsic feedback methods.

While large language models (LLMs) perform well on table tasks, they still make data referencing errors (DREs), i.e., incorrectly citing or omitting table values, despite understanding the table structure. Beyond final-answer accuracy, DREs directly compromise the correctness and reliability of intermediate reasoning steps. Yet prior studies have only offered limited, small-scale analyses. In this work, we present the first systematic evaluation of tabular data referencing errors across different models and tasks. Our results show that DREs occur across all tested models (1.7B to 20B parameters). Furthermore, we demonstrate that incorporating data referencing as a critic significantly improves answer accuracy up to 12.0%, through critic-based filtering and rejection sampling. Finally, we trained a lightweight 4B-parameter critic model that achieves an average F1 score of 78.2% in detecting both in-distribution and out-of-distribution DREs, and effectively assists inference for larger models.

Reward design remains a central bottleneck for autonomous robot policy improvement, especially in long-horizon manipulation tasks where sparse success labels provide too little signal and binary preferences collapse many competing notions of quality into one ambiguous signal. We introduce Freeform Preference Learning (FPL), a method for learning robot policies from freeform human preferences. Rather than asking annotators which of two trajectories is better overall, FPL lets them define natural-language preference axes, such as speed, safety, quality of placement, or carefulness, and provide pairwise preferences along each axis. These annotations are used to learn a language-conditioned reward model that maps a trajectory and preference label to an axis-specific reward. We use this model to train a reward-conditioned policy that optimizes across the multiple human-specified dimensions. Across four real-world and two simulated long-horizon manipulation tasks, FPL improves over sparse-reward and binary-preference methods by 38 percentage points. Beyond improved performance, FPL learns dense progress signals without explicit subtask segmentation, shows compositionality of behavior not present in the data, and allows users to steer the policy towards different behaviors at test time without retraining. Blog post with videos available at https://freeform-pl.github.io/fpl.website/

Latent world models enable planning from high-dimensional observations by predicting future states in a compact latent space. However, these models are typically kept frozen at test time: when their predictions become inaccurate, planning can fail, especially under test-time distribution shift. To address this, we propose AdaJEPA, an adaptive latent world model that performs test-time adaptation within the closed loop of model predictive control (MPC). After training, AdaJEPA plans and executes the first action chunk, uses the observed next-state transition as a self-supervised adaptation signal, and replans with the updated model. This closed-loop update continuously recalibrates the world model without additional expert demonstrations. Across a range of goal-reaching tasks, AdaJEPA substantially improves planning success with as few as one gradient step per MPC replanning step.

Recent LLM agents benefit from skills for solving complex tasks. Skills encapsulate modular packages of procedural knowledge and instructions for performing specialized tasks, such as setting up a sandboxed environment, running a test suite, or refactoring a function across multiple files. As skill libraries grow and become reusable across tasks and domains, selecting an appropriate skill composition has emerged as a central bottleneck. Existing approaches fall into two categories. One exposes the agent's reasoning to the entire skill collection; the other performs skill retrieval via embeddings or LLM-based rerankers. Both provide useful insights; however, they miss the structural nature of skill composition, which is a joint decision over which skills, how many, and in what order -- three dimensions that cannot be decoupled. We formalize this as structured skill composition: given a task and a skill library, predict an executable skill plan that jointly specifies the activated subset, count, and execution order. We propose SkillComposer, which instantiates structured skill composition as task-conditioned skill sequence prediction. SkillComposer uses a constrained autoregressive decoder over skill identifiers, so subset, count, and order emerge jointly from a single decoding pass, and dependencies between successive skills are captured naturally. We build a training set of task-composition pairs from a real, human-curated skill library. We then evaluate SkillComposer along two axes: composition quality on a held-out test set, and downstream task success on SkillsBench across two production-grade coding agents. On GPT-5.2-Codex, Gemini-3-Pro-Preview, SkillComposer raises the pass rate by +23.1, +18.2pp over the no-skill baseline, surpassing top-3 retrieval and matching the gold-skill retrieval upper bound at lower prompt-token cost.

Forest attributes are essential for national-scale resource monitoring. Airborne LiDAR metrics are among the auxiliary variables most strongly correlated with forest attributes used in National Forest Inventory (NFI) estimates. However, producing wall-to-wall predictions remains challenging when LiDAR data are acquired under heterogeneous conditions. As national LiDAR programs expand across Europe, variability in sensors, flight parameters, seasons, and scan angles limits the robustness of existing models, which are often calibrated for local conditions. We present FLORA (Forest LiDAR Octree Regression with Auxiliary Data), a deep learning framework that predicts six forest attributes: dominant height, total volume, deciduous volume, coniferous volume, basal area, and stem density from heterogeneous LiDAR point clouds. FLORA combines an octree-based backbone with ecological and spatiotemporal auxiliary variables through a late-fusion gating mechanism. Models are trained and evaluated on 32,052 National Forest Inventory plots across mainland France using data from the French LiDAR HD program. A single model trained on both leaf-on and leaf-off acquisitions outperforms season-specific models and improves cross-season robustness. Auxiliary variables provide modest overall gains but contribute more strongly to species-specific volume prediction. FLORA achieves an rRMSE of about 12.3% (R2 = 0.88) for dominant height and 39% (R2 = 0.74) for total volume, providing a robust baseline for large-scale forest attribute estimation from heterogeneous national LiDAR programs.

Residual-stream analysis asks how language-model computation evolves across depth, but intermediate decoding requires comparable readout coordinates across layers. If embedding anchors and unembedding readout disagree on the chosen span, apparent motion may reflect measurement drift rather than computation. We introduce \emph{Semantic Reference Frames} (SemRF), an anchor-based formalism separating semantic measurement from residual dynamics. A SemRF fixes anchors and measures states against them. Pseudo-inverse tying gives exact synchronization; under restricted bi-invertibility, SemRF yields stable semantic-basis coordinates, distortion bounds, and near-identity changes. With the frame fixed, residual computation becomes a depthwise semantic trajectory. The anchors induce a semantic Voronoi diagram: distance, or evidence such as logits, assigns each layer to a coarse cell, while coordinates retain within-cell motion and margins. We define layerwise steps, contribution profiles, and imbalance diagnostics, then use the Voronoi trace to define a margin-relaxed tube. The canonical trace is the minimum-action path inside this tube; when nonempty with positive quadratic weight, it is unique and obeys a discrete spline equation away from active constraints. Excess action controls step, curvature, and profile mismatch. Low curvature implies piecewise-linear compressibility and local knowledge density: lower trace complexity means fewer semantic knots. Through the parameter-to-trajectory map, this gives a conditional link to parameter efficiency: among admissible settings fitting data, lower-action and lower-complexity traces use fewer semantic degrees of freedom. The guarantees require controlled interface error and small projection residual under explicit tube constraints.

Classifiers based on Deep Neural Networks exhibit strong performance across domains, yet can fail catastrophically if they rely on spurious correlations, i.e., features that are predictive of the target label in the training data but are not causally linked and thus fail to generalize. For the vision domain, many such spurious correlations manifest themselves within the background of the image, where only the foreground is predictive of the class label. In this paper, we introduce Automated Background Swapping (AutoBackSwap) to reduce the reliance of classifiers on such spurious backgrounds. AutoBackSwap uses a secondary network to disentangle the foreground and background, followed by infilling to synthesize complete backgrounds, and finally combines different foregrounds and inpainted backgrounds to augment the training data. We find that patch-wise labeling of just a few hundred samples suffices to train the secondary network and automatically augment the full training dataset on challenging image classification tasks. In contrast to many previous methods, AutoBackSwap proves very effective even if there is not a single sample in the training data breaking the spurious correlation. Across a range of image classification tasks with spurious backgrounds, AutoBackSwap consistently outperforms prior methods.

Agentic reinforcement learning requires assigning credit to environment-facing actions such as searches, clicks, edits, navigation commands, and object interactions. Standard GRPO uses the final verifier outcome as a uniform advantage over all action tokens. This outcome signal is useful but structurally incomplete: it punishes useful exploration in failed rollouts and reinforces redundant or regressive actions in successful rollouts. We propose TRIAGE, a role-typed credit assignment framework that adds a semantic role axis to outcome credit. A structured judge classifies each segment as decisive progress, useful exploration, no-progress infrastructure, or regression, and a fixed role-conditioned rule maps these labels to bounded segment-level process rewards. This keeps verifier outcomes as the source of optimization direction while correcting the two main blind spots of outcome-only credit. We further show that role-conditioned credit is the optimal segment-level correction expressible from role labels alone -- a projection of the per-segment advantage residual onto the role variable -- so that the fixed role constants reduce advantage estimation error whenever the judge is reliable, and we connect this to lower-variance policy gradients. Across ALFWorld, Search-QA, and WebShop, TRIAGE improves success rates over GRPO for two policy models and outperforms both a scalar judge-derived process reward and an outcome-supervised shared-backbone value baseline. Ablations show that the gain comes from role typing rather than merely adding dense rewards: reliable detection of regression inside successful trajectories is the dominant contributor, while exploration credit provides a consistent secondary gain; on completed ALFWorld and WebShop rollouts, TRIAGE also reduces environment-facing turns by an additional $10.4\%$ and $14.8\%$ relative to GRPO.

Multimodal graph foundation models aim to learn reusable knowledge from graphs enriched with text, images, attributes, and relational topology, thereby supporting diverse graph-centric and modality-centric tasks. In practice, however, such multimodal graphs are often distributed across decentralized clients, where raw contents and local structures cannot be centrally shared due to privacy constraints. This motivates federated multimodal graph foundation learning, which requires not only transferable representation learning but also intrinsic semantic traceability under strict data isolation. Existing methods usually exchange or store knowledge through parameters, prototypes, embeddings, or compact codebooks, which support optimization and transfer but do not explicitly expose how modality evidence, node semantics, and topology context jointly support predictions. To bridge this gap, we propose FedLAB, a traceable semantic codebook framework that organizes multimodal graph knowledge into typed hierarchical codebooks for modality evidence, node semantics, and topology context. FedLAB further refines these trace units through federated semantic barycenter pre-training while keeping raw multimodal contents and graph structures local. Extensive experiments on 10 benchmarks and 6 downstream tasks show that FedLAB improves over state-of-the-art baselines by up to 7.53\%, while preserving a native semantic trace interface.

Internet users collectively perform an enormous range of skilled work through web browsers, from software development and document editing to search, forms, and enterprise workflows, making human browsing a highly scalable but under-exploited source of reusable browser skills. We argue that the bottleneck for browser agents is decision-making under incomplete information rather than low-level operation, and that the priors agents lack are already implicit in human interaction traces. We therefore study scalable behavior cloning for browser agents via skill distillation, converting user interaction trajectories into compact natural-language skills that agents can read, retrieve, reuse, and compose directly. We further organize the distilled skills into a skill graph so that growth proceeds through consolidation rather than unbounded accumulation. This suggests that the scalability of browser agents may come less from manually designed tasks and more from the collective skills already expressed by internet users. Our project is available at: https://lab.einsia.ai/browserbc/.

Uncertainty estimation has been a long-standing challenge in AI models; it amounts to "knowing what you don't know," and metacognition is notoriously difficult even for humans (cf. the Dunning-Kruger effect). Although it is still far from solved even in simpler classification systems, tackling it in multimodal large language models (MLLMs) is becoming increasingly important. Within MLLMs, uncertainty can stem from any of the diverse sources as well as from their relationships, and further can stem from the unbounded answers in the open-ended setting. To tackle the issues, we propose CoMet, an MLLM uncertainty estimation method by decomposing uncertainty into a context-specific term and a multiplicity-specific term. The former captures ambiguity induced by the given context (e.g., task or prompt), while the latter captures how many plausible answers determined by the context remain compatible with the given input. We train a lightweight post-hoc uncertainty module to estimate these quantities, which enables efficient uncertainty estimation without autoregressive answer generation or repeated sampling. Experiments on various open-ended multimodal benchmarks, hallucination detection, and multiple-choice visual question answering benchmarks show that CoMet consistently improves uncertainty estimation over existing baselines while remaining efficient in practice. Code is available at https://github.com/princetonvisualai/comet_uncertainty

Mechanistic interpretability (MI) requires full access to model internals, yet the APIs for most widely deployed language models at best expose log-probabilities over output tokens. This creates a surrogate problem: when do measurements made on open models allow us to make claims about a closed model? We evaluate surrogate fidelity at the prediction, attribution, and representation levels. For binary classification tasks, log-odds provide an API-compatible scalar readout of the model's representation space, and leave-one-out attributions provide insight into model behavior. Across eleven models spanning four families (Llama, Qwen, GPT, and Gemini), we find that prediction fidelity substantially overstates attribution fidelity: models that agree on what the answer is often disagree on why. We document an access-validity inversion: white-box signals like attention patterns and perturbation magnitudes are highly stable across models but only weakly predictive of causal attributions, which black-box input ablations capture by design. Mechanistic insight does not automatically transfer to closed targets, and prediction-level agreement is insufficient to warrant such transfer. Code and results are available at https://github.com/facebookresearch/surrogate.

We study agentic code generation in Dafny, where a model must generate both executable code and the proof artifacts for verification. We present AxDafny, a verifier-guided repair framework that iteratively generates implementations, invariants, assertions, and termination arguments. We also introduce LiveCodeBench-Pro-Dafny (LCB-Pro-Dafny), a benchmark of 250 competition-style programming problems translated into Dafny with formal specifications and a verifier-based evaluation harness. On LCB-Pro-Dafny, AxDafny substantially improves verification success over baseline GPT-5.5 performance. On DafnyBench, AxDafny achieves 92.7\% verification success, outperforming the strongest previously reported proof-hint baseline by 6.5 percentage points. Lastly, we show that verification success and runtime test performance measure different aspects of generated code.

Stochastic Gradient Descent ($\textsf{SGD}$) is one of the most classical optimization algorithms with favorable theoretical guarantees, yet the practical implementation of $\textsf{SGD}$ differs subtly from its well-known form and is often referred to as Shuffling Stochastic Gradient Descent ($\textsf{Shuffling SGD}$). A particularly popular strategy in $\textsf{Shuffling SGD}$ is Random Reshuffling ($\textsf{RR}$), which has achieved great empirical success across numerous experiments. Despite its strong performance, $\textsf{RR}$ has long been considered a heuristic due to a lack of theoretical support. Over the last decade, people have finally established provable convergence rates for $\textsf{RR}$, thus justifying its observed superiority. However, for smooth convex optimization, two clouds over the convergence theory of $\textsf{RR}$ remain to this day. More precisely, according to the current theory, $\textsf{Shuffling SGD}$ under $\textsf{RR}$ converges only when the stepsize is smaller than a threshold proportional to $1/n$, where $n$ is the number of summands in the objective (or the number of data points). Consequently, the optimally tuned theoretical rate of $\textsf{Shuffling SGD}$ under $\textsf{RR}$ is strictly worse than that of $\textsf{SGD}$ when the number of epochs is smaller than another threshold proportional to $n$. These two restrictions heavily limit the applicability of existing theories and leave a critical mismatch with practice. In this work, for the first time, we prove that $\textsf{RR}$ dominates $\textsf{SGD}$ in smooth convex optimization under any reasonable stepsize after any finite number of epochs, thereby addressing a longstanding open question.

Policy-grounded document review requires determining whether a target document complies with organization-specific policies, guidelines, or playbooks. While large language models can assist with policy interpretation and document analysis, end-to-end prompting leaves the applied policy logic implicit, making compliance decisions difficult to inspect, update, and test. We present PolicyGuard, a neuro-symbolic framework for policy-grounded document compliance review. PolicyGuard converts organizational policy guidance into an executable review engine consisting of typed relational logic rules and atom-level extraction questions. During review, LLMs answer these local questions using retrieved document evidence, and a symbolic evaluator applies the formal rules to detect non-compliance. We instantiate and evaluate PolicyGuard on company-specific NDA compliance review, where contract clauses must be checked against organization-specific negotiation policies. By separating policy formalization, local document interpretation, and symbolic compliance evaluation, PolicyGuard makes document review more explicit, maintainable, and systematically testable.

Language models are increasingly taught from synthetic question--answer (QA) supervision: a model generates questions about a document, answers them from the same text, and the resulting pairs are used to fine-tune, distill, or compress knowledge into another model. We show that this generation step is not neutral preprocessing. It is an implicit policy that both selects which evidence becomes training signal and decides how that evidence is answered, and it is fragile at both stages. When choosing what to ask, generators do not scan a document uniformly. Coverage saturates early and concentrates on salient spans, diverse prompts converge on the same regions, and what looks question-worthy is driven by local presentation. As a result, salient artifacts such as poorly cleaned markup can hijack question generation across model families and scales. When answering, the model that produces the supervision tends to obey instruction-like passages embedded in the text. This compliance depends on the intent and surface form of the passage rather than its strictness, and is worst under task conflict, where larger models comply more often. These failure modes arise from choices made during QA generation, so they can be reduced without changing the training loop. Tying each question to a fixed target reduces biased selection, and filtering instruction-like spans before answering lowers mean injection compliance from $88\%$ to $13\%$ in our evaluation while retaining nearly all clean text.

Why do neural networks memorize algorithmic training data long before they generalize? We present a geometric case study demonstrating that, on tasks where generalization requires discovering structured low-dimensional circuits, the memorization-generalization delay is driven by radial inflation of hidden representations under cross-entropy optimization. We formalize a radial-angular decomposition of activation-space dynamics and derive three testable propositions: (i) that penalizing radial inflation induces anisotropic, data-dependent weight regularization; (ii) that it suppresses radial gradient energy below the isotropic random baseline, forcing predominantly angular updates; and (iii) that it biases convergence toward flatter minima. To empirically validate these propositions, we study a single-hyperparameter norm penalty that softly constrains activations to a sqrt(d)-radius hypersphere. On modular arithmetic, this penalty accelerates grokking up to 6x across MLPs and Transformers, and halves training steps for a 10M-parameter nanoGPT on 3-digit addition.

The tendency of large generative models to memorize training data makes sample verification critical for privacy auditing and copyright enforcement. Current membership (MIA) and dataset inference (DI) attacks often rely on one-shot generations, which yield weak signals and limited sensitivity across modalities. Inspired by Model Autophagy Disorder (MAD), we introduce MADreMIA, a model-agnostic framework that enhances white-, gray-, and black-box MIA and DI. Rather than relying on shadow model training -- often infeasible for large generative models -- our framework facilitates scalable inference by leveraging inherent signals through iterative trajectories. This process utilizes chained generations across diverse modalities, where each output serves as the subsequent input, to improve membership evidence at low FPR. We demonstrate that memorized training samples exhibit significantly higher coherence and slower degradation during iterative regeneration than non-member generations. Our results show that MADreMIA provides richer signals across diverse model families and modalities; we present comprehensive evaluations for IARs, diffusion, and language models, alongside preliminary results demonstrating its potential for audio models.

We analyze the effect of optimizing the initial population of genetic programming (GP) for symbolic regression (SR) on the accuracy and complexity of solutions. We compare three well-established random initialization methods as well as initialization with small optimized solutions from exhaustive symbolic regression (ESR) using a GP/SR implementation which is based on the multi-objective evolutionary algorithm NSGA-II. We compare the final Pareto fronts found with each initialization method on twelve synthetic problems of varying complexity and one real-world dataset. We find no significant differences in accuracy or model complexity among the initialization methods. The initial advantage of initialization with ESR disappears after only a few generations. Our results show that, given similar diversity in the initial population, the effect of the initialization method in GP-based symbolic regression on the final Pareto front is negligible.

Industrial recommendation systems serve billions of users through a multi-stage funnel -- retrieval, early-stage ranking, and re-ranking -- where the final re-ranking step disproportionately shapes user engagement and downstream performance, particularly for carousel and grid display formats. Despite growing enthusiasm for Large Language Models (LLMs) in recommendation, three gaps hinder industrial adoption: (1) most efforts target retrieval and ranking, leaving re-ranking -- the stage closest to the final user experience -- largely underexplored; (2) LLMs are typically deployed zero-shot or via supervised fine-tuning, underutilizing the reasoning capabilities unlocked by reinforcement learning (RL) on verifiable rewards; (3) deployed catalogs index billions of items with non-semantic identifiers that lie outside any base-LLM vocabulary. We present GR2 (Generative Reasoning Re-Ranker), an end-to-end framework that combines (i) mid-training on semantic IDs produced by a tokenizer with >=99% uniqueness, (ii) reasoning-trace distilled from a stronger teacher via targeted prompting and rejection sampling, and (iii) RL with verifiable rewards purpose-built for re-ranking. To make GR2 resource-viable, we further (iv) introduce a context compressor that amortizes training cost, On-Policy Distillation (OPD) as a scalable alternative to SFT -- which we find collapses at industrial scale -- and reasoning distillation for low-latency serving. GR2 delivers +18.7% R@1, +7.1% R@3, and +9.6% N@3 over legacy baselines on industrial-scale traffic. We further find that reward design is critical in re-ranking: LLMs often hack rewards by preserving the incoming order or exploiting position bias, motivating conditional verifiable rewards as essential industrial components.

Creating photorealistic, animatable 3D human avatars from monocular images still largely depends on Linear Blend Skinning (LBS) and parametric body models, which constrain expressivity and often introduce artifacts due to imperfect fitting. We propose LUNA, an LBS-free universal neural animation model that directly maps multiple 2D controls like images, keypoints, sketches, and unseen characters into 3D Gaussian deformations, bypassing explicit body fitting. At its core, a transformer-based motion regressor disentangles global rigid motion from fine-grained local dynamics to capture both coherent movement and subtle non-rigid effects. To resolve the inherent ambiguity of 2D-to-3D lifting while scaling beyond fitted datasets, we introduce hybrid supervision that distills soft structural priors from an LBS teacher and a loss that supports training on both limited fitted data and large in-the-wild unlabeled videos. Extensive experiments show LUNA achieves competitive visual fidelity compared to LBS-based approaches, while delivering realistic human motion and zero-shot cross-identity generalization across diverse driving modalities. To the best of our knowledge, LUNA is the first end-to-end 3D animatable model that supports implicit 2D driving.

Agents are increasingly capable of automating software tasks, but can they teach humans how to use software themselves? We introduce DigitalCoach, a multimodal dataset of 72 human expert-novice computer use coaching sessions consisting of 22,752 dialogue turns grounded in 28.1 hours of screen and input event recordings across five software applications. We use DigitalCoach to evaluate whether state-of-the-art models can teach humans how to use computers. Automated evaluation shows that models differ from humans in how they coach: models provide more direct instructions, but fewer explanations, error diagnoses, and knowledge-check questions. When we fix the coaching method, models produce utterances similar to human references yet poorly grounded in visual context. Interactive evaluation confirms that model coaches cause learners to passively follow instructions without deeper engagement and fall short in visual grounding. DigitalCoach lays a foundation for collaborative and proactive computer use coaching agents.

Human-labeled data are widely used as reference annotations in ML, despite known variability across annotators in many expert-driven domains. In addition, expert annotation is slow, inconsistent, and remains a major bottleneck for scaling tasks like tree height bias classification in forestry remote sensing. We propose a multi-agent system (MAS) that orchestrates expert decision trees with Vision-Language Models (VLMs), treating the decision tree as a structural prior while VLMs perform localized semantic perception at individual nodes, with multi-agent voting to mitigate VLM stochasticity. We formalize a Decoupled Declarative Decision (D3) Framework that enables zero-modification generalization across diverse expert-defined decision structures. On a tree bias classification testbed, our framework outperforms supervised ML baselines and reduces the amount of expert labeling effort required. These results suggest that agentic orchestration of VLMs with expert priors can reproduce expert-defined labeling procedures at substantially lower annotation cost while maintaining interpretability.

Semantic communication (SemCom) has emerged as a promising paradigm in which the transmission of task-relevant information is prioritized over raw data, enabling efficient and robust communication under resource and channel constraints. In this paper, the privacy implications of relay-assisted SemCom systems are studied, where the intermediate relay node operates directly on learned latent representations. It is shown that the relay, even without access to source data, can reliably infer semantic meaning and reconstruct signals with performance comparable to that of the legitimate receiver, revealing a fundamental privacy vulnerability of semantic representations. To address this issue, an iterative adversarial training framework is proposed in which a strong, adaptively trained eavesdropper at the relay is explicitly accounted for. The proposed approach alternates between optimizing the relay's eavesdropping function and the legitimate system, resulting in representations that preserve semantic decoding performance at the intended receiver while degrading semantic inference at the relay. The semantic accuracy gap between the legitimate receiver and the eavesdropper is significantly enlarged across channel conditions. Importantly, this protection is achieved in a stealthy manner, with high reconstruction fidelity maintained while semantic leakage is selectively suppressed.

Software-engineering assistants often need method-level context beyond an isolated body, including enclosing-class information, documentation, callers, callees, type hierarchy, and structural characteristics. Manually collecting this context is time-consuming, inconsistent, and difficult to reproduce across large Java projects. We present CoCoMUT, a Java tool for Code-Context Mining and Automated Dataset Generation. CoCoMUT extracts context for a focal method or generates datasets at class, package, or system scope. It discovers project structure, resolves build and classpath information, constructs a SootUp static call graph, and reconciles bytecode-level call edges with Spoon-based source extraction. Each method record combines source, class, documentation, call-graph, and metadata context, providing reproducible inputs for training and running learned software-engineering techniques. The key contribution is a reusable, task-independent pipeline that unifies build discovery, source extraction, call-graph construction, source-bytecode reconciliation, and versioned JSON dataset generation. The resulting records can be consumed individually as context for a focal method or collectively as datasets for documentation, explanation, testing, review, repair, search, and program-comprehension workflows. We evaluate CoCoMUT on 20 real-world Java repositories evenly split between Maven and Gradle. CoCoMUT processed all 20 repositories, emitting 56,512 method-context records and 386,048 serialized call edges. Among call edges whose bytecode targets belonged to project source, CoCoMUT reconciled 97.8% to source method identities. In a manual audit of 200 randomly sampled methods across 10 systems, 99.0% of generated context records passed all applicable correctness checks.

Recent multimodal large language models (MLLMs) have strong potential as embodied agents, but their ability to collaborate in visually grounded environments remains underexplored. To address this gap, we introduce MECoBench, a multimodal embodied cooperation benchmark with an evaluation platform spanning diverse real-world tasks, two cooperation structures, and three collaboration modes. Through extensive experiments across various MLLMs, we summarize three key findings: (i) Collaboration generally improves embodied task completion, but its benefits depend on balancing collaborative gains against coordination complexity. (ii) Communication is essential to collaboration gains, while the best collaboration mode depends on team size and model capability. (iii) Moreover, collaboration improves robustness under noisy priors and exploration conditions. Generally, MECoBench provides a systematic testbed for understanding the mechanisms and limits of multimodal embodied collaboration. Code and dataset are available at https://github.com/q-i-n-g/MECoBench.

Modern LLM workflows move coordinate-indexed objects across checkpoints: steering vectors, sparse autoencoders, top-$k$ neuron sets, attribution lists, and merge alignments. This is only well posed after fixing the model's residual-stream gauge, which we show is architecture-dependent: LayerNorm residual charts have permutation gauge $S_d$ (up to a global sign flip), while RMSNorm charts with generic per-channel gain have signed-permutation gauge $B_d = S_d \ltimes \{\pm 1\}^d$. Permutation-only alignment is therefore symmetry-incomplete for RMSNorm models. We introduce sign-marginalized Hungarian matching and prove a sharp failure mode: with decorrelated coordinates, raw signed-correlation matching has a structural permutation-accuracy ceiling at the positive-sign fraction of the true gauge, which sign-marginalization removes. We then make coordinate-preserving transport, not function-level merging, the primary object: composing saved-checkpoint local $B_d$ gauges along same-base fine-tuning trajectories recovers 91.1% of cross-run coordinates at 1500 steps versus 60.3% for endpoint matching, and the gain is not explained by merely routing through the base. The recovered gauge transfers tools that permutation-only alignment breaks: TinyLlama SAE reconstruction has NMSE 0.004 under $B_d$ versus 1.08 under $S_d$; Qwen sentiment steering preserves 95.8% of its effect versus 17.2%; refusal steering reverses sign under $S_d$; coordinate-preserving merges behave the same way. The same covariance governs stateful training: signed transport of AdamW state preserves the resumed trajectory, while permutation-only state follows a different one from a functionally identical checkpoint. Finally, gauge-sweep audits show index-level interpretability claims are reproducible only relative to an explicit gauge.

State-of-the-art speech datasets predominantly focus on widely spoken languages, often overlooking low-resource languages such as Luxembourgish, which remain underrepresented in speech technology research. In this work, we introduce LuxEmo, a 21-hour conversational expressive speech corpus for Luxembourgish with 4 emotion categories. LuxEmo is derived from Radio Télévision Luxembourg (RTL) youth broadcasts, using automated detection followed by human validation. We propose a semi-automatic curation workflow combining voice activity detection, denoising, language identification, LuxASR-based segmentation, automatic emotion prediction, lexical cues, and targeted human review. Additionally, we benchmark five expressive TTS systems covering German-based cross-lingual transfer, multilingual Luxembourgish support, Luxembourgish adaptation, and non-parametric prosody transfer. Performance is evaluated using both objective metrics and human evaluation.

Is the sense of touch a mechanism for human babies' learning of visual concepts? If so, can we quantify its importance, and to what extent do babies rely on their sense of touch for visual learning? To approach these questions in a principled way, we propose a structured coding system for baby-centric touch events, yielding a dataset of 264k two-second clips of touch events coded according to this system. Using this dataset, we pretrain developmentally grounded models that reveal promising insights into the nature of baby learning from touch.

Unstructured navigational features, such as irregular planting or discontinuities, remain the primary failure mode for under-canopy agricultural robots. Existing geometric approaches often fail in these scenarios because they compress high-dimensional visual data into deterministic spatial references, effectively discarding the uncertainty and semantic context required to navigate ambiguous terrain. To address this, we present LeCropFollow, a visual navigation framework that bypasses explicit geometric modeling in favor of a learned latent representation. By integrating a self-supervised semantic heatmap extractor with TD-MPC2, a Model-Based Reinforcement Learning (MBRL) planner, our system optimizes trajectories directly within a latent manifold. The framework operates over the uncompressed heatmap signal, preserving the semantic context that geometric reductions discard. We demonstrate that this representational shift enables zero-shot transfer from simplified simulation to the physical world without fine-tuning. Extensive field experiments in late-stage corn fields show that LeCropFollow matches state-of-the-art baselines in unstructured rows but significantly outperforms them in plantation gaps, achieving a 2.4x reduction in semantic failures compared to keypoint-based methods. These results suggest that latent planning offers a robust alternative to geometric estimation for operations in heterogeneous agricultural environments. Code, models, and data available: https://felipe-tommaselli.github.io/lecropfollow .

Deploying Vision Transformer (ViT) models on edge platforms remains challenging due to their high computational demands and the architectural heterogeneity of modern hybrid ViT models, which incorporate both fully connected and convolutional layers. This heterogeneity leads to significant variation in tensor shapes, requiring flexible and efficient FPGA-based acceleration. In this paper, we present FlexViT, a reconfigurable FPGA accelerator for efficient ViT inference on resource-constrained edge devices. Built on the SECDA-TFLite framework, FlexViT employs a hardware-software co-design approach that maps both fully connected and convolutional layers onto a unified high-throughput INT8 GEMM engine using a runtime im2col transformation. To efficiently support diverse layer configurations, we propose a dual-mode dataflow that dynamically switches between input and weight reuse by reconfiguring the compute array at runtime. We further introduce a depth-first tiling strategy that completes accumulation in a single pass, eliminating off-chip partial-sum transfers and reducing memory bandwidth requirements. We implement FlexViT on a PYNQ-Z2 FPGA and evaluate it across a representative set of ViT models. FlexViT achieves up to 2.74x speedup on accelerator-executed layers, translating into up to 1.40x end-to-end speedup compared to CPU-only execution. The code is available at: https://github.com/gicLAB/FlexViT

Cohesive Zone Models (CZMs) are widely used to simulate interface fracture, delamination, adhesive failure, and fiber--matrix debonding in aerospace composite structures. In implicit quasi-static finite element analyses, cohesive softening may introduce negative interface tangents, solution jumps, and Newton-basin mismatch, so the previous converged state can become a poor initial guess for the next increment. This may lead to stagnation, wrong-branch convergence, or repeated step cuts. Existing remedies, including viscous regularization, path following, dynamic relaxation, and manual Newton--Raphson (NR) modification, either alter the effective response, increase cost, or rely on hand-crafted interface rules. This work proposes an Interface-Aware Neural Newton Preconditioner (IA-NNP) for difficult CZM increments. IA-NNP recasts manual NR modification as rule-based interface lifting and generalizes it into a learned, state-dependent interface correction. The method acts only on active interface variables and preserves the original traction--separation law, residual assembly, tangent evaluation, history update, and dissipation checks. Two realizations are developed: IA-NNP-Init for learned initial-guess lifting and IA-NNP-NL for iteration-level nonlinear right preconditioning. Interface graph features encode opening, traction, tangent, damage/history variables, mode mixity, residuals, and neighboring states. The correction is bounded, confidence-gated, and accepted only through the original CZM Newton solve. A root-equivalence property shows that IA-NNP changes the path to convergence but not the discrete CZM solution set. Tests on horizontal, circular, two-interface, and active-front benchmarks show improved difficult-increment convergence, better branch recovery, and fewer failures than standard NR and manual NR modification, while preserving the force--displacement response.

Zero-shot Object Goal Navigation (ZSON) with RGB-only perception poses a fundamental challenge for embodied agents, as the absence of explicit depth information introduces severe physical uncertainty and semantic-physical misalignment. Existing approaches either rely on high-level semantic reasoning without geometric grounding or learn end-to-end policies that lack explicit physical constraints, often resulting in semantically plausible but physically unsafe behaviors. In this paper, we propose MVP-Nav, a physical-aware RGB-only navigation framework that aligns perception, planning, and control with the real 3D world. MVP-Nav reconstructs explicit physical occupancy from monocular observations by leveraging 3D foundation models to project 2D semantic instances into 3D oriented bounding boxes, forming a global spatial semantic representation. To unify high-level semantic reasoning and low-level physical constraints, we introduce a Multi-layer Value Map (MVM) that integrates semantic priorities and reconstructed geometry into a shared cost space, enabling physically grounded geometric planning. Extensive experiments on zero-shot object navigation benchmarks demonstrate that MVP-Nav significantly outperforms existing depth-free methods, achieving state-of-the-art performance and validating that structured physical priors can effectively compensate for the absence of active depth sensors.

Theory of Mind (ToM) benchmarks for Large Language Models (LLMs) typically rely on passive question-answering formats, but the deployment of LLMs in increasingly agentic and autonomous forms demands new evaluations. In this paper we evaluate an agent's ability to induce specific belief states in other agents by taking actions rather than using conversational persuasion, a capability we call Non-Conversational Planning ToM (NCP-ToM). NCP-ToM is likely to be essential for many agent use-cases, including within user-assistant interactions and pedagogical contexts, but may also present manipulation or misinformation risks. Using a novel framework, NCP-ExploreToM, we subvert the conventional task structure by providing models with a set of belief state goals and requiring them to move objects or direct characters into rooms to achieve their goals. We evaluated six frontier models, including GPT-5, Gemini 2.5 Pro and the Claude 4 series, and a cohort of human participants, across 600 task instances. GPT-5 was successful on approximately 80% of tasks in the agentic setting, and was the only model to outperform human participants on our task, but was still less robust than humans across contexts. We additionally found that all models, like humans, performed better on tasks inducing true belief states than false belief states, which is a positive signal for alignment efforts. These findings highlight emerging social-reasoning capabilities in LLMs for non-conversational task completion and underscore the necessity of agentic evaluations for understanding the safety and alignment of autonomous social agents.

While conformal prediction provides a general framework for uncertainty quantification in predictive inference, its application is often limited by computational cost. Recent methods, including Jackknife+ and Jackknife-minmax, achieve faster computation by trading a slight loss of efficiency relative to full conformal prediction, but still requires computing leave-one-out refits for all observations. In this paper, we further accelerate conformal prediction by incorporating approximate leave-one-out (ALO) estimators, and establish asymptotic coverage and efficiency. While our proof draws on methods developed for analyzing the consistency of ALO cross-validation risk estimators in high-dimensional statistics, it requires adaptations to handle conformal prediction, where leave-$i$-out residuals are needed for predictions at $x_{n+1}$ rather than just at the training covariate $x_i$. Simulation results validate our theoretical findings, showing that the ALO-based methods achieve coverage and efficiency comparable to the exact methods, while significantly reducing the runtime.

The generation of synthetic relational databases often involves modeling complex temporal dynamics, such as transaction logs or event sequences. A significant challenge in this domain is the handling of categorical time series (e.g., status codes), where standard encoding methods like one-hot encoding fail to capture intrinsic frequency-domain features such as seasonality and cyclicity. In this paper, we introduce Sequential RC-TGAN (Seq. RC-TGAN), a temporal extension of the RC-TGAN framework, equipped with a novel integrated loss function based on the \textit{Spectral Envelope Theory}. This differentiable loss allows the generator to directly optimize the preservation of latent periodic structures via backpropagation. While spectral envelope theory is inherently designed for categorical sequences, we extend this frequency-domain regularization to continuous time series by employing a Variational Gaussian Mixture Model (VGM) discretization strategy. To establish a mathematically rigorous evaluation standard, we simulate categorical time series governed by a parameter $α$, with exactly known theoretical spectral envelopes. Integrating these dynamic sequences into the child tables of a relational database yields a robust ground-truth benchmark for evaluating the frequency-domain fidelity of our generative framework. Furthermore, we address the lack of robust evaluation standards for relational time series by proposing two new metrics: Spectral Density Divergence and Spectral Envelope Divergence. Experimental results on real-world datasets, as well as our simulated benchmarks, demonstrate that our end-to-end approach significantly outperforms state-of-the-art systems in reproducing cyclic patterns and long-term seasonality across both categorical and continuous features.

Multimodal large language models (MLLMs) increasingly process long visual-token sequences, increasing the overall inference computation. Existing acceleration methods usually remove visual tokens or skip visual-token updates in entire layers, but these coarse strategies may discard fine-grained evidence or suppress useful operators together with redundant ones. In this paper, we study visual-token computation from an answer-observable perspective and find that late visual-token updates can remain large while having little effect on answer-token representations. Motivated by this answer-silent redundancy, we decompose each Transformer layer into attention and FFN operators and show that useful visual computation is often operator-dominant and layer-dependent. We propose an operator-level visual-token skipping framework that preserves the full visual-token sequence while selectively bypassing redundant attention, FFN, or both. Experiments across three MLLM architectures and 10 VQA benchmarks show that our method achieves strong efficiency-accuracy trade-offs, reducing \textbf{33.7\%} TFLOPs on Qwen3-VL while retaining \textbf{99.5\%} of the vanilla model performance.

"Any fool can know; the point is to understand." A well-known remark often attributed to Einstein captures a widely shared intuition: understanding is more than merely knowing. Yet epistemic logic has paid relatively little attention to understanding, despite its central role in contemporary epistemology, philosophy of science, and recent debates about AI. A recurring theme in the philosophical literature is that, unlike knowledge, understanding comes in degrees: one may understand something more or less well, and one's understanding may be better than another's. We introduce a comparative epistemic logic of understanding with level-indexed understanding modalities and a comparative connective for saying that one agent understands why a proposition better than another agent does. Semantically, we enrich multi-agent epistemic models with agent-indexed graded explanation structures and a justification-style term algebra. This yields a unified framework for representing minimal, ordinary, more demanding, and ideal understanding, together with comparisons between agents with respect to the same formula at issue. We distinguish a finitary bounded-level calculus from an infinitary full-language companion system. We establish soundness and strong completeness, and show that each fixed finite-level fragment is decidable.

Distributed knowledge is a notion of group knowledge studied in multi-agent epistemic logic. Semantically, the distributed knowledge of a group is interpreted via an accessibility relation given by the intersection of the epistemic accessibility relations of the agents in that group. This paper investigates sequent calculi for epistemic logics of distributed knowledge based on K45, KD45, and S5. While cut elimination holds in existing sequent calculi for modal logics K45 and KD45, it fails in all the systems mentioned above. Instead, we establish the analytic cut property for all three systems by adapting Takano' s (2018) strategy, which restricts the cut formulas to the set of subformulas of the conclusion of the cut rule. As a corollary, the Craig interpolation theorem holds for all logics considered. We also show that all proof-theoretic results remain valid when the empty group is allowed for the distributed-knowledge operator, in which case the distributed knowledge for the empty group is interpreted as the global modality.

We investigate two approaches for extending CEGAR-tableaux with SAT-shortcuts using a previously known approach called RECAR but also a totally new approach using the modal resolution theorem prover KSP as an oracle. Our experiments using our C++ implementation CEGARBox++ of CEGAR-tableaux show that: (1) CEGARBox++ with RECAR SAT-shortcuts is not competitive (2) CEGARBox++ using KSP to provide SAT-shortcuts is superior to both CEGARBox++ and KSP, particularly on large satisfiable problems. As far as we know, this is the first effective integration of SAT, tableaux and resolution methods for modal satisfiability which performs better than its parts.

To improve safety in Large Language Models (LLMs) we can either perform post-training alignment or exploit refusal directions in the activation space. Both strategies are less feasible in Multimodal LLMs (MLLMs) as they require unsafe multimodal data, harder to collect than their unimodal counterpart. In this work, we relax this constraint and investigate whether textual refusal directions, extracted directly from the LLM backbone, generalize across modalities (i.e., image, video). Preliminary findings confirm this ability, though effectiveness is conditioned by layer selection, steering strength, and cross-modal alignment, with the latter causing safe multimodal inputs to be spuriously steered toward refusal. Building on this, we introduce Modality-Agnostic Refusal Steering (MARS), a light-weight training-free approach that injects multimodal safety without the need for multimodal safety data. MARS corrects modality misalignment via activation re-centering, adaptively scales steering strength within a geometrically defined trust region, and selects the optimal intervention layer, operating at the first generated token. Evaluated on five SOTA MLLMs across safety, utility, and video jailbreak benchmarks, MARS achieves consistent safety gains while preserving utility. These results reveal that safety-relevant structure is shared across modalities and that textual refusal directions are a powerful and underexplored foundation for multimodal alignment.

We introduce inquisitive action logic, InqAL, a multi-agent modal logic for reasoning about action. While traditional approaches focus on what properties of the outcome an agent can force, InqAL also captures what aspects of the outcome an agent determines through their actions. As we argue, such claims of agentive determination are naturally analyzed as modal claims involving questions. Technically, InqAL is a multi-agent extension of inquisitive neighborhood logic based on concurrent game structures. With respect to statements, it is expressively equivalent to the individual-agent fragment of the socially friendly coalition logic recently proposed by Goranko and Enqvist. We present an axiomatization of InqAL and prove completeness and decidability via the finite model property. Along the way, we establish a representation theorem for actual effectivity functions, associating to an agent the sets of outcomes corresponding to their possible actions; we give exact conditions under which a multi-agent neighborhood frame arises from a concurrent game structure.

Dynamic epistemic logic represents belief change via model transformations induced by epistemic events. Its standard formulation (Baltag, Moss, Solecki, 1998) provides a natural account of belief expansion through the elimination of possibilities, but it cannot model belief contraction about factual propositions. A classic response enriches Kripke models with plausibility orderings, representing contraction as an update that promotes certain possibilities over others. We show that this approach has expressive limitations. In particular, the approach cannot model belief that violates positive introspection and contraction dynamics in response to a hedged public announcement that phi might be false. Motivated by these considerations, we introduce a mechanism for belief contraction defined directly on standard Kripke models, without any constraints on the doxastic accessibility relation. We show that it satisfies some of the standard properties of belief contraction but not others, study the conditions under which contraction may be unsuccessful, and provide a sound and complete axiomatization of the logic via reduction axioms. We also define a more general dynamic logic that is an extension of standard DEL and accommodates belief contractions due to events such as private or semi-private announcements, and provide a complete and sound axiomatization of the general logic.

Residual connections add every sublayer's proposed update with a fixed coefficient of one; the network never evaluates whether an update is reliable before committing it. Drawing on the human-factors principle of independent verification, we introduce Review Residuals, which scale each update by a learned, input-dependent gate conditioned on both the current state and the proposed update: h_l = h_{l-1} + r_l * u_l with r_l = sigmoid(W[RMSNorm(h_{l-1}), RMSNorm(u_l)]). Conditioning the gate on the update is the property that distinguishes it from prior gated and scaled residuals. We report two findings. First, a depth-stability result: a convex (Highway-style) form of the gate reintroduces vanishing gradients and fails to train beyond ~20 layers, whereas the additive, identity-preserving form trains stably at all depths we tested. Second, an emergence-with-scale result: trained from scratch across five sizes (60M-1B parameters, multi-seed), Review Residuals show no advantage at small scale but at 590M significantly outperform both a parameter-matched Highway gate and a parameter-matched standard residual (p<0.05), with a larger advantage at 1B. The benefit grows with model size rather than shrinking.

We investigate a new logic that extends Dynamic Epistemic Logic (DEL), by combining standard epistemic modalities for (individual and distributed) propositional knowledge with operators for (conditional) non-propositional knowledge of a number (in which an agent or a group have knowledge of the value of some variable x, conditional on some additional information). We also generalize these operators, by considering formulas that express the fact that an agent or group can (conditionally) narrow down the possible values of the variable x to at most N possibilities (for some natural number N). In order to name and compare such hypothetical values, we extend the logic further with definite descriptions based on minimization operators, denoting the least of the N possible values of x (according to some fixed order) that are considered possible by the agent or group. On this static base, we consider DEL-style extensions with dynamic modalities for general 'data-exchange events' (covering private and public propositional announcements, but also secret hacking of a private database, or public sharing of one's data via open-source repositories, etc.). In such scenarios, whole 'chunks' of information may be exchanged or modified: once access to a given source is gained, all the 'data' stored at that specific location becomes available. We give complete axiomatizations for the resulting logics, and prove their decidability and co-expressivity.

We study layered models, including feedforward networks, ResNets, and transformers, by limiting each layer to a width of $d = 3$, i.e., $\mathbb{R}^3$ as representation space. This allows us to track how a neural network changes low-dimensional topological invariants through its layers. Just about any topological structure may be simplified or even trivialized by simply increasing dimension; e.g., any knot is equivalent to an unknot in $\mathbb{R}^4$. By restricting to $\mathbb{R}^3$, we not only isolate the effects of activation and depth from that of width, we work in a space that lends itself to easy visualization. We focus on linking number here, deferring other invariants like link groups, Milnor's $\barμ$-invariants, knot types, ambient cobordisms, to a sequel. We provide full proofs and empirical experiments to justify the following insights: When measured by their power to effect changes in linking numbers, the layer-skipping feature in ResNets is as powerful as the attention mechanism in transformers; both ResNets and transformers are strictly more powerful than feedforward neural networks with monotonic activations, which are in turn more powerful than invertible and flow-based models; but replacing monotonic activation with a nonmonotonic one elevates a feedforward network into the same expressivity class as ResNets and transformers. These results suggest that low-dimensional topology can be a useful tool to guide designs of AI architectures. We also generalize our results from $d = 3$ to arbitrary $d > 3$.

There are by now various epistemic modal logics with intersection modalities for distributed knowledge and intersection update modalities for dynamic phenomena like agents sharing (all their) information, agents receiving information from other agents, and full information protocols. One of those is the logic of Resolving Distributed Knowledge, by Agotnes and Wang. It has distributed knowledge modalities for arbitrary subsets of the set of all agents and it also has so-called resolution modalities for arbitrary subsets of agents sharing their knowledge. In that logic, the agents not involved in the knowledge sharing are aware of the agents sharing knowledge, agents are memory-less, and the kind of dynamics represents synchronous updates, where there is common awareness of the global clock. In contrast, in this contribution we present a logic for Resolving Asynchronous Distributed Knowledge. It is an asynchronous generalization of the synchronous logic of resolving distributed knowledge. The logical semantics is history-based: truth is not only with respect to a given world in a model, but also with respect to a given history of prior resolutions, of which each individual agent can only observe a part. In particular, an agent is unaware of resolutions for groups of agents not including her. As is to be expected, this comes with many technical complications, for example concerning the axiomatization. The synchronous axioms relating resolution to distributed knowledge are now invalid. The modelling advantages of such an asynchronous novel logic, for distributed computing and similar areas, are however substantial and a major asset.

Vision-Language-Action (VLA) models offer a promising framework for robotic manipulation by connecting language instructions, visual observations, and continuous control. However, most existing policies remain limited by behavior cloning or supervised fine-tuning (SFT) from fixed demonstrations, which provides limited opportunity to improve from the policy's own failures. In this paper, we present Z-1, a reinforcement learning (RL) post-training framework for flow-based VLA models. Built on top of $π_{0.5}$, Z-1 uses only publicly released RoboCasa demonstrations for SFT and then applies a task-wise Group Relative Policy Optimization (GRPO) strategy across $24$ standard RoboCasa tasks. To improve the efficiency and stability of online optimization, Z-1 combines shared-prefix rollout construction, tree-structured trajectory branching, completion-aware reward calibration, and selective joint training of VLM and Action Expert. Across all $24$ RoboCasa tasks, Z-1 achieves an average success rate of $80.6\%$, improving over its SFT initialization by $13.2\%$ points and outperforms the published sota models. These results show that systematic GRPO post-training can substantially improve flow-based VLA policies without additional private demonstrations.

A transformer's feed-forward (FFN) sublayer materializes the distinctions attention gathers, yet gives no account of what it computes. In a parameter-neutral replacement, each hidden unit is an explicit fuzzy set operation on sigmoid-bounded [0,1] memberships: intersection A*B and set-difference A*(1-B), the latter a bounded positive negation ("A but not B") that gated/bilinear units lack -- a negation-capable FFN (NC-FFN). On N-bit parity they are the most parameter-efficient reasoning basis at shallow depth; at scale (125M, OpenWebText) NC-FFN ties the GELU baseline's perplexity, every unit carrying explicit logical form. Two limits share one cause: two-operand logic localizes to layer 0 and erodes under training, and the one robust grammatical deficit concentrates in licensing and quantifiers, beyond within-token operators. We resolve both with a small block of sequence quantifiers: a soft existential and a soft proportion, each with a per-unit learned forgetting rate from a sticky init. This recovers the deficit at epoch one (halving the wider epoch-two gap), modestly leads on LAMBADA, and makes the FFN legible: the structure now holds and migrates into depth; the decay un-learns its stickiness (median half-life ~1.5 tokens; zero latch units); and at the semantic layers the units read, without dictionary learning, as grammatical licensing detectors: each fires on a licensor (a comparative, a passive participle, a negative-polarity item) and carries its memory forward to predict the licensed word (than, by, nor). This legibility is localized and free only up to a partition (a fully Boolean FFN diverges in training), but the result is a parameter-neutral, language-model-quality transformer with a readable, interpretable-by-construction grammatical mechanism -- an account not just of what a feed-forward layer represents but how it licenses.

A local-to-global context mismatch arises when autoregressive traffic simulators trained on ego-centric driving logs are deployed in globally observable closed-loop environments. In such logs, the ego vehicle has rich local observations, while surrounding agents are only partially observed due to perception limits and occlusions. As a result, simulators may learn incomplete context--action mappings that remain hidden in log-based training but emerge during closed-loop rollouts, leading to unrealistic behaviors such as abnormal stops, unsafe interactions, and rule violations. We propose CRAFT, a Contextual pReference Alignment Framework for Traffic Simulation, to mitigate this mismatch via self-supervised failure discovery and preference-guided test-time alignment. CRAFT treats the base simulator as a globally observable sandbox, generating diverse what-if rollouts from logged initial states to expose context-induced failures. These failures are grounded with human-aligned driving priors and converted into preference supervision for training a Contextual Preference Evaluator (CPE). At inference time, CPE acts as a plug-in alignment module that scores candidate actions under complete scene context and reweights autoregressive decoding toward globally coherent behaviors. CRAFT mitigates this local-to-global contextual bias, reducing collisions by 31.2\% and traffic violations by 33.2\% without retraining the base simulator.

Real-world detectors for autonomous driving, surveillance, and robotics must handle domain-shifts under strict latency and memory constraints, yet existing source-free object detection (SFOD) methods rely on heavyweight architectures that prioritize accuracy alone. We show this trade-off is unnecessary: building on YOLOv10, an NMS-free dual-head detector, we achieve state-of-the-art adaptation accuracy while being faster and more compact. We observe that directly applying vanilla mean-teacher self-training to dual-head detectors leads to suboptimal adaptation performance due to two key factors. First, simple pseudo-label generation strategies, such as using a single head or directly combining high-confidence predictions from both heads, yield suboptimal supervision under domain-shift. We propose DHF (Dual-Head Pseudo-Label Fusion) which selectively admits one-to-one (O2O) and one-to-many (O2M) head predictions, preserving precision and recovering missed objects. Second, we observe domain-shift collapses multi-scale feature discriminability. We propose the use of our MARD (Multi-scale Adaptive Representation Diversification) loss which mitigates this by enforcing detection-aware variance and covariance constraints on multi-scale feature maps. Both modules are training-time only, leaving inference unchanged. Across domain-shift benchmarks, our method, RT-SFOD yields 1.4 to 3.5\% mAP gains, 1.3$\times$ higher throughput, with $\sim$2$\times$ fewer parameters than prior state-of-the-art SFOD methods, thus advancing the Pareto frontier of the speed-accuracy-model size trade-off. We report main results with YOLOv10, and demonstrate generalizability with additional YOLO- and DETR-based dual-head detectors. Code is available here: https://github.com/Sairam13001/RT-SFOD/

High-throughput plant phenotyping now generates image derived datasets far faster than scientists can analyze them. At Oak Ridge National Laboratory's Advanced Plant Phenotyping Laboratory (APPL), automated stations image hundreds of plants daily across multiple remote sensing modalities; yet, trait extraction and interpretation remain manual, expert-bound, and strictly post-hoc, making analysis, not acquisition, the binding constraint on discovery. We present an end-to-end agentic AI framework that turns the facility from a data factory into an interactive autonomous, discovery platform, where scientists partner with AI agents to accelerate time to insight. A conversational Co-Scientist Agent translates a scientist's natural-language question into a structured analysis plan, and a headless Compute Agent dispatches Vision Transformer segmentation and trait extraction on the Frontier exascale supercomputer. The two agents run in separate security and resource domains and communicate over a secure, token-authenticated streaming channel, a design that accounts for the federation, data-movement, and provenance realities cloud-native agentic frameworks ignore, ensuring end-to-end provenance is captured for every interaction. The framework turns a days- to weeks-long analysis process into an interactive loop where agents reason over results, recommend next analyses, and respond to follow-up questions in seconds.

Recent multimodal large language models have shown great promise in clinical image reasoning, but existing post-training pipelines remain predominantly outcome-centric, relying on final answer correctness or sequence-level preferences. This suffers from sparse credit assignment, making it difficult to optimize the reasoning process essential for clinical applications. Our analysis reveals that cascading errors from early-stage reasoning failures are a leading cause of incorrect predictions in medical visual question answering (VQA) benchmarks. Motivated by this, we propose Medical Reasoning-aware Policy Optimization (MRPO), an RL algorithm that incorporates step-wise process rewards. When the final answer is incorrect, MRPO assigns exponentially larger penalties to tokens in earlier invalid reasoning steps, breaking failure cascades without compromising successful paths. Across three multimodal LLM backbones, MRPO consistently outperforms standard GRPO and a recent RL baseline, and on Qwen3-VL-8B-Instruct even surpasses substantially larger medical MLLMs such as HuatuoGPT-Vision-34B by 2.79 points. Moreover, MRPO reduces early-stage reasoning failures from 64.0% to 13.0%, showing that targeted mitigation of cascading failures improves both reasoning quality and final answer accuracy. Our code is available at https://github.com/dmis-lab/MRPO

Large-scale capacitated vehicle routing problems (CVRPs) are commonly addressed using cluster-first route-second (CFRS) approaches that split a routing instance into smaller, computationally tractable subproblems. Existing splitting methods typically rely on fixed partitioning rules, predefined optimization objectives, or learned policies, which may perform inconsistently across instances exhibiting different spatial, demand, and operational characteristics. In this work, we propose an adaptive CFRS system that formulates a decomposition procedure as an iterative decision-making process. Motivated by the recent success of large language models (LLMs) in reasoning and tool selection, the system employs an LLM as a high-level decision maker that analyzes the evolving decomposition state and selectively applies further clustering, balancing, and refinement operators. The proposed algorithm jointly partitions customers and vehicles, enabling capacity-aware clustering while adapting partitioning decisions to the characteristics of each problem. We evaluate the approach on synthetic and benchmark-derived CVRP instances containing up to 500,000 customers. Experimental results demonstrate competitive performance on benchmark-scale instances while exhibiting improved scalability and robust routing quality on substantially larger problems. These results highlight the potential of adaptive, LLM-guided decision support as a practical approach for industrial-scale vehicle routing and large-scale logistics planning.

This work introduces a new computational theory of mind grounded in set theory and hyperdimensional computing. Whereas traditional neural networks rely on continuous weights and matrix multiplication, this framework works with sparse binary data. It represents information as discrete sets, directly modeling biological neural population codes. I demonstrate that associative memory emerges naturally from network topologies featuring a combinatorially expanded hidden layer. Learning is driven by topological plasticity rather than scalar weight adjustments. This architecture unifies auto-associative and hetero-associative learning under a single core algorithm: information retrieval via subset pattern matching and exact nearest-neighbor search. Operating with constant-time complexity, these mechanisms bridge perceptual data (sparse distributed representations) and symbols (sparse holographic representations) without continuous bottlenecks. Mapping this framework to neuroanatomy, I propose that both the cerebellum and the neocortex implement variants of this algorithm, making subset pattern matching the fundamental engine of cognition. Because it relies on discrete logic rather than matrix arithmetic, this algorithm translates directly into in-memory hardware. This opens a new route toward synthetic intelligence with human-level energy efficiency.

Compatibility research usually treats an interface change as a local writer-reader decision. Distributed software stacks make that decision population structured: an RPC, telemetry, middleware, or service-contract variant is introduced by one provider release and then spreads, stalls, or is mediated across consumers, transitive dependencies, and resolver rules. This paper asks when that observation is a load-bearing software-engineering estimator rather than evolutionary relabeling. We mine interface histories, audit npm, Maven Central, PyPI, and crates.io package graphs, execute 2100 package-manager resolver probes, estimate an ecosystem-specific selection coefficient $s$ from clean conflict probabilities, and use that measured $s$ to forward evaluate a pairwise-comparison absorbing process on the observed package graph. We separate three evidential roles. Fixation is a forward evaluation, not independent evidence: once $s$ is measured, deviation from $1/N$ follows mechanically from the non-neutral process. Checker-derived direction carries adoption signal: a direction-permutation null gives checker-direction gap MAE 0.07 versus null median 0.43 ($p=0.002$). But because that direction is derived from the same boundary state whose admitting frequency is predicted, it is a diagnostic rather than an orthogonal selection test. The stricter checker-free temporal test asks whether early resolver-channel features predict later blocked-to-admitted flips; in this snapshot they do not beat age-only (Brier 0.28 versus 0.24, AUC 0.51 versus 0.54). The result is a reproducible estimator audit for interface-variant dynamics in distributed package graphs, showing where resolver evidence becomes population input and where the current registry data still fail to close the resolver-to-adoption loop.

Low-rank adaptation (LoRA) and its variants enable parameter-efficient fine-tuning of large language models under the supervised fine-tuning (SFT) paradigm. However, their efficacy and behavior under Reinforcement learning with verifiable rewards (RLVR) are less well understood. In particular, two structurally initialized LoRA variants, PiSSA and MiLoRA, which outperform standard LoRA under SFT, can underperform standard LoRA under RLVR and may even exhibit training instability. These observations suggest that how to initialize the low-rank matrices in RLVR remains unclear. In this work, we develop a theoretical analysis of LoRA in RLVR, showing that orthonormal initialization achieves the minimal gap between LoRA outcome and that of full fine-tuning. Guided by this insight, we propose geometry-preserving orthonormal initialization for low-rank adaptation in RLVR, leading to two new variants, RLPO and RLMO. Experiments on mathematical reasoning benchmarks show that the proposed orthonormal initialization stabilizes RLVR training and outperforms standard LoRA, contrasting with PiSSA and MiLoRA. Finally, our unified analysis for LoRA initialization also explains why PiSSA and MiLoRA can underperform in RLVR, which may be of independent interest. Code and checkpoints are publicly available at https://github.com/Richard-ZZZ/geometry-preserving-orthonormal-init-rlvr.

Language model systems built around proprietary APIs often operate on a token-based cost model. This becomes prohibitively expensive in the context of large databases, where LM-enhanced relational operators can incur costs exceeding $10,000 for a single set of experiments, hindering thorough research and practical deployment. In this paper, we demonstrate that quantized, open-weight models running locally on just 16GB of VRAM can match or exceed the accuracy of closed-source counterparts at lower latency and a fraction of the price, challenging the prevailing assumption that closed-source LM APIs are necessary for effective LM-database integration. We present and analyze the key system optimizations required to efficiently deploy these open-weight models within an LM-DB system. By integrating these local models into the BlendSQL v0.1.0 framework, we demonstrate a 390x reduction in overall costs and 3.8x reduction in latency compared to a proprietary LM API. We make our code available at https://github.com/CapitalOne-Research/play-by-the-type-rules/tree/main/sembench.

Accurate energy demand forecasting is essential for the reliable operation and planning of modern sustainable energy systems. Spatial-temporal graph neural networks (STGNNs) have recently achieved strong performance in point forecasting by jointly modeling temporal dynamics and relational dependencies across interconnected energy nodes. However, in real-world energy systems, accurate point forecasts alone are insufficient, as operators also require reliable uncertainty estimates to support risk-aware decision-making, grid stability, and operational planning under uncertainty. Conformal prediction provides a principled and model-agnostic framework for uncertainty quantification with statistical coverage guarantees, making it particularly attractive for safety-critical energy applications. However, existing conformal prediction approaches often fail to fully capture the complex spatial-temporal structure of energy systems. To address these limitations, we propose STOIC (Spatial-Temporal Graph Conformal Prediction with In-Context Learning), a novel framework that integrates graph-based forecasting with the zero-shot calibration capabilities of tabular foundation models. STOIC first generates point forecasts using an STGNN and subsequently reformulates spatial-temporal residuals into a tabular representation suitable for in-context learning. Leveraging a tabular foundation model, STOIC calibrates prediction intervals without task-specific retraining, effectively capturing both sequential and relational dependencies. We evaluate STOIC on five diverse benchmarks, including synthetic simulations as well as real-world electricity and district heating networks. Across all datasets, STOIC consistently outperforms existing conformal prediction baselines, delivering more reliable and robust uncertainty estimates for complex graph-structured energy time series.

Automated Heuristic Design (AHD) with Large Language Models (LLMs) has shown remarkable progress in discovering high-quality heuristics. However, existing LLM-based AHD methods optimize heuristics for a fixed training instance set and may fail catastrophically when deployed under real-world distributional shifts. We propose Robust Adversary Instance Search (RAISE), a framework that integrates constrained worst-case instance search within a principled neighborhood of the training distribution into the LLM-based evolutionary search loop. RAISE treats robust AHD as a constrained adversarial instance search problem: the outer loop evolves heuristics via LLM operators, while an LLM-free inner loop efficiently identifies hard instances within an epsilon-ball around the training instance set using a basis distribution parameterization with boundary projection. Comprehensive experiments on Online Bin Packing (OBP), Online Job Shop Scheduling (OJSP), and Online Vehicle Routing (OVRP) across five distribution families demonstrate that existing LLM-based AHD methods degrade by up to 19 times under distribution shift, while RAISE consistently maintains strong performance across all tested distributions and problem scales

Despite recent progress, the reasoning capabilities of large multimodal language models (MLLMs) remain fundamentally constrained by static supervision, where fixed prompts, rules, or reward models provide non-adaptive guidance throughout training. Such static signals are often sufficient to enforce output formats, but fail to shape the underlying reasoning process, leading to brittle generalization and performance saturation in complex decision-making tasks. We propose Evo-PI, a principle-centric learning framework that treats reasoning principles as explicit, language-based supervision signals that can be generated, evaluated, and iteratively evolved. Instead of relying on fixed rewards, Evo-PI enables a co-evolutionary loop in which principles guide model reasoning, while model behaviors in turn refine the principles that supervise them. This dynamic alignment mechanism allows supervision to progressively adapt to the model's reasoning deficiencies. We instantiate Evo-PI in medical visual question answering as a high-stakes testbed requiring structured visual-textual reasoning. Across eight benchmarks and multiple model backbones, Evo-PI consistently improves reasoning accuracy, achieving gains of up to 24.6%. Our results suggest that evolving principle-guided supervision offers a scalable and general paradigm for training expert-aligned reasoning in MLLMs. Code is available at https://github.com/zhengxianda/Evo_PI.

We study three complementary techniques for training compute-efficient language models. (1) Selective supervision and per-token efficiency. Selective Ground Truth Token Training (SGT) concentrates supervision on the ~15% of output tokens that carry semantic payload. Through positive gradient coupling in position-shared transformer weights -- a token-level instance of auxiliary-task transfer -- the remaining 85% of unsupervised tokens still improve substantially, giving a 4.5x per-supervised-token efficiency (at the step-100 eval optimum, ~67% of the full-sequence loss reduction is recovered from 15% of the supervision). We prove that this improvement on unsupervised tokens is guaranteed whenever the gradient coupling coefficient gamma-bar = 0.72 is positive (Theorem 1), and show the effect is a property of natural-language structure: it collapses on shuffled text. (2) Depth compression with recurrent recovery. A 48-layer, 1B-parameter transformer is compressed to 6 layers (227M) by averaging adjacent layers and restored through learned recurrent unrolling. With 34 effective recurrent layers it reaches a held-out loss of 2.934, within measurement noise of a 566M dense model at 2.926 -- a 2.5x reduction in parameters. (3) Fusion of compressed experts. Assembling several compressed models as a Mixture of Efficient Experts (MoEE) with multi-token prediction improves over each single expert at comparable active parameters: a 2-expert MoEE reaches loss 2.789 versus 2.926 for the best single compressed model. We validate these techniques on CHERRY-1.8B, a Korean foundation model whose every trainable parameter derives from our own training runs. We are explicit throughout about the scope of the evidence (one model family, Korean data, loss-based metrics) and about which claims are established versus prospective.

Anomaly detection in multivariate time series remains a critical challenge in large-scale distributed systems, where related entities may exhibit transferable precursor behavior prior to anomaly onset. Existing methods typically operate independently on each data stream and therefore remain fundamentally reactive. To address this limitation, we introduce Distributed Hierarchical Temporal Memory (D-HTM), a neuromorphic framework that enables cross-entity preemptive warning through a Shared Associative Memory (SAM). D-HTM combines a Spatial Pooler (SP) that projects observations into a common Sparse Distributed Representation (SDR) space, Temporal Memory (TM) modules that learn entity-specific dynamics online, and a Shared Associative Memory that stores recurring pre-anomaly signatures. By reusing precursor knowledge across related entities, D-HTM can issue warnings prior to local anomaly onset while preserving HTM's online learning capabilities. We evaluate D-HTM on the Server Machine Dataset (SMD), the Soil Moisture Active Passive (SMAP) dataset, the Mars Science Laboratory (MSL) dataset, and a synthetic cascade benchmark designed to isolate precursor transfer. Experimental results demonstrate effective cross-entity warning propagation while maintaining competitive reactive anomaly detection performance. Across the real-world datasets, D-HTM provides an average warning lead time of 8.1 samples prior to anomaly onset. These findings demonstrate that transferable precursor structure can emerge within a shared SDR space and be reused for preemptive warning generation, extending HTM beyond isolated reactive detection toward distributed predictive reasoning.

Log parsing is a fundamental step in automated log analysis, transforming raw system logs into structured event templates for downstream tasks such as anomaly detection and system monitoring. Existing log parsing methods range from rule-based and clustering-based approaches to neural models that learn semantic representations from log messages. However, neural approaches typically rely on dense matrix multiplications, which can result in high computational cost and energy consumption. This paper presents SpikeLogBERT, a spiking neural network framework for energy-efficient log parsing. The proposed model integrates a spiking transformer architecture with knowledge distillation from a BERT teacher model, enabling spike-driven computation while preserving semantic representation capability. By leveraging sparse spike activations and event-driven processing, the number of active operations during inference can be significantly reduced. As an initial benchmark study, experiments on the HDFS dataset demonstrate that SpikeLogBERT outperforms ANN-based neural log parsing models with a parsing accuracy of 0.99997, while reducing estimated theoretical energy consumption by up to 62.6% under standard 45nm CMOS assumptions.

Language models typically reason via explicit chain-of-thought (CoT), generating intermediate steps token-by-token. Latent CoT offers an alternative: it performs multi-step reasoning in the model's hidden states, replacing decoded tokens with continuous representations for greater efficiency. However, existing latent CoT methods underperform explicit CoT beyond 1B parameters, and the gap widens with scale. Looped, or recurrent-depth, Transformers, which reuse their weights to increase computation depth without adding parameters, are a natural fit for latent reasoning. We therefore ask whether looped Transformers can bridge this gap. We answer affirmatively with a simple recipe: a looped padded Transformer that processes K latent blocks in parallel for R iterations, with a cross-entropy loss on each latent position's gold CoT-step token, similar to explicit CoT supervision. We instantiate it as LOTUS (Looped Transformers with parallel supervision on latents). LOTUS is, to our knowledge, the first latent-CoT method to bridge the gap to explicit CoT at the 3B scale, while cutting thought-phase latency by 2.5x-6.9x from compact math expressions to natural language. Projecting LOTUS's post-loop latents through the base LM head recovers the gold reasoning steps and even surfaces alternative valid intermediate steps, evidence that its latent space is interpretable and CoT-aligned. Ablations confirm that both the looped backbone and the parallel supervision on gold CoT tokens are essential.

We study policy optimization for online episodic tabular Markov decision processes with unknown transition kernels, aiming for best-of-both-worlds guarantees together with data-dependent regret bounds. Recent work (Dann et al., 2023; Li et al., 2026) has shown that policy optimization can adapt to both adversarial and stochastic losses with first-order, second-order, and path-length bounds, but only under known transitions, leaving open whether such data-dependent guarantees are achievable by policy optimization when the transition kernel is unknown. We resolve this by developing a new algorithm based on optimistic follow-the-regularized-leader that attains these guarantees under unknown transitions. The key ingredient is a new design of optimistic $Q$-function estimators together with a data-dependent transition bonus that controls estimator bias through the loss-prediction error. Our analysis further identifies an unavoidable transition-dependent complexity term that captures the intrinsic cost of estimating the transition kernel. As a result, we obtain first-order, second-order, and path-length bounds with the transition-dependent complexity term while simultaneously achieving gap-dependent $\mathrm{polylog}(T)$ regret in the stochastic regime.

Automated fixing of performance issues is gaining increasing attention. However, existing benchmarks of execution time improvement patches are fixed datasets that target Python, C++, or .NET and cannot be extended to new patches according to user-defined configurations. In this paper, we present JETO-Mine, the first configurable and reusable tool for automatically creating reproducible benchmarks of execution time improvement patches (ETIPs) in real-world Java projects. JETO-Mine employs a three-phase pipeline: a static analysis phase that crawls GitHub repositories and identifies ETIPs using user-defined filters and an LLM-based issue classifier, a dynamic analysis phase that wraps the identified ETIPs in Docker images for fully reproducible execution and performs statistical testing to find objective evidence of execution time improvement, and an evaluation harness that enables quantitative assessment of both generated patches and generated tests. Unlike existing benchmarks, JETO-Mine is designed as a reusable tool that allows researchers continuously collect new benchmarks with their own desired filters and statistical rigor levels. We use JETO-Mine to build JETO-Bench, a benchmark of 660 identified ETIPs and 91 manually verified executable ETIPs collected from 174 open-source Java repositories. To build JETO-Bench, JETO-Mine scans 11 years of open-source development history and nearly 1.8 million commits. We run OpenHands, a leading open-source coding agent, on the 91 manually verified executable ETIPs in JETO-Bench and find that it correctly fixes 14.3% (13/91) of the issues, aligning with results reported by similar studies on other programming languages. Our results also reveal that open-source Java projects largely lack tests that demonstrate execution time improvements, presenting an opportunity for future research in test generation.

Autonomous wet-lab experimentation requires more than plausible protocol text: biological intent, quantitative procedures, device constraints and experimental feedback must remain aligned from protocol and SOP design to code and physical execution. We developed ProtoPilot, a self-evolving multi-agent system, together with an expert-grounded benchmark and evaluation framework for testing this conversion as an experimental automation problem. The framework spans 294 synthetic-biology and molecular-biology tasks derived from 98 gold-standard protocols, wet-lab expert rubrics, device-level validity gates and real experimental tests. ProtoPilot incorporates layer-wise verifiability, multi-agent orchestration and a runtime-updated skill library to generate protocols, expand SOPs, synthesize SDK-compliant code and revise workflows from wet-lab feedback. It achieved a Top@3 expert-preference rate of 90.2%, an overall protocol-to-code gate pass rate of 89.5% and an Opentrons pass rate of 88.24%, compared with 32.35% for OpenTrons-AI. Wet-lab validation produced interpretable readouts, Sanger-confirmed products and feedback-corrected PCA-assembled DNA targets, establishing a verifiable route to autonomous experimentation. Together, these results show that the evaluation framework captures execution-relevant requirements for autonomous wet-lab automation, and that ProtoPilot can meet them by converting protocol and code generation into validated execution and feedback-guided revision.

Conflict-free Replicated Data Types (CRDTs) provide strong eventual consistency without coordination, but classical approaches assume benign participants. In Byzantine settings, convergence is typically enforced through agreement on update validity, often relying on identity-based filtering. However, such approaches struggle in post-compromise scenarios, where a previously correct participant becomes malicious: retroactive exclusion of its updates may break causal dependencies and invalidate subsequent computations. In this paper, we decouple identity-based trust from content-based trust and introduce a fine-grained trust model that combines both dimensions. Building on deterministic reconstruction, our approach allows replicas to preserve previously accepted updates while enabling selective inclusion or exclusion based on both the originating identity (e.g., public keys) and the semantics of individual updates. Trust decisions can incorporate application-level policies, enabling precise control over the impact of each update on the system state. Our approach preserves causal consistency and enables robust and flexible handling of both Byzantine and faulty behavior in decentralized CRDT systems.

Research on artificial intelligence (AI) in the public sector often treats "AI" as a single category, neglecting technical distinctions between different AI systems. But these distinctions affect how different systems impact core public values like accountability, procedural justice, and non-discrimination. This paper argues that public administration research would benefit from more technical precision on "AI" and makes three contributions to this end. First, we introduce a typology of five categories of AI systems: hand-coded, glass-box, black-box, general-purpose, and agentic systems. We calibrate the typology to public administration by grouping system types by their distinct implications for public values. Second, we evaluate technical precision in recent public administration research about AI by coding 91 highly-cited papers (2019-2025) using our typology. We find widespread imprecision: most papers (55\%) leave the studied system underspecified, 31\% motivate their work with a different system than they study, and 41\% make more general conclusions than the studied system supports. Finally, we give practical recommendations for future research. We highlight common pitfalls to avoid, and suggest that researchers should, at a minimum, provide enough technical detail to locate the studied system in our typology. To this end, we provide a practical guide -- a short set of diagnostic questions answerable from public information and without specialist technical knowledge.

Safety training on language models often induces over-refusal: improved safety on harmful prompts at the cost of increased refusal on harmless ones. Though this trade-off can be mitigated by training models with reinforcement learning (RL) to reason before answering, it does not remove the underlying problem that reasoning can often be a "rubber stamp" for a predetermined response. In this paper, we address the safety-refusal trade-off by rethinking how models are trained to reason about safety. Our key insight is that unsafe reasoning can itself serve as a useful exploratory signal. Rather than preemptively blocking harmful thoughts, we encourage the model to sufficiently explore unsafe reasoning but produce a safe response. The harmful exploration improves the model's ability to distinguish harmful from harmless prompts by resolving ambiguity, allowing it to remain safe while complying only when appropriate. We cast this as an adversarial optimization problem in which a reasoning player explores strategies for producing an unsafe response and an answer player ensures that the final output is safe. We train a single model with dense rewards to play both roles within one chain-of-thought, across different segments. To achieve this, we find that process rewards are crucial for stable optimization of competing objectives. Our resulting model SEAR deliberately engages in harmful reasoning as exploration while reliably flipping back to a safe answer. We demonstrate that this behavior helps mitigate over-refusal and defend against attacks that directly manipulate the reasoning to be harmful.

Remote sensing change detection (CD) traditionally focuses on pixel-level binary segmentation, which identifies where changes occur but neither what nor why. To bridge this semantic gap, we introduce JL1-CC&QA, a multi-task benchmark that extends the JL1-CD dataset with two complementary annotation layers: change captioning (CC) and change question answering (QA). Built upon 5,000 bi-temporal image pairs acquired by the Jilin-1 satellite at 0.5-0.75m ground sample distance, the benchmark comprises: (i) JL1-CC, providing 17,021 quality-verified captions that describe diverse land-cover transformations; and (ii) JL1-QA, offering 20,060 question-answer pairs across eight question types, enabling fine-grained, interactive interrogation of surface changes. All annotations are produced via a three-stage pipeline consisting of multi-modal large language model (LLM) generation, vision-grounded LLM judging, and human expert verification. We hope that JL1-CC&QA, as a benchmark unifying binary change masks, change captions, and change-oriented QA over the same image set, will serve as a valuable resource for the community to advance multi-task change understanding in remote sensing. The dataset is available at https://github.com/circleLZY/JL1-CD.

Explainable AI (XAI) methods have demonstrated significant success in recent years at identifying relevant features in input data that drive deep learning model decisions, enhancing interpretability for users. However, the potential of XAI beyond providing model transparency has remained largely unexplored in adjacent machine learning domains. In this paper, we show for the first time how XAI can be utilized in the context of federated learning. Specifically, while federated learning enables collaborative model training without raw data sharing, it suffers from performance degradation when client data distributions exhibit statistical heterogeneity. We introduce FedXDS (Federated Learning via XAI-guided Data Sharing), the first approach to utilize feature attribution techniques to identify precisely which data elements should be selectively shared between clients to mitigate heterogeneity. By employing propagation-based attribution, our method identifies task-relevant features through a single backward pass, enabling selective data sharing that aligns client contributions. To protect sensitive information, we incorporate metric privacy techniques that provide formal privacy guarantees while preserving utility. Experimental results demonstrate that our approach consistently achieves higher accuracy and faster convergence compared to existing methods across varying client numbers and heterogeneity settings. We provide theoretical privacy guarantees and empirically demonstrate robustness against both membership inference and feature inversion attacks. Code is available at https://github.com/MaxH1996/FedXDS.

While semantic embeddings are rigorously evaluated on the Massive Text Embedding Benchmark, the evaluation of style embeddings remains fragmented, with each work relying on their own set of tasks and datasets. To bridge this gap, we introduce the Style Text Embedding Benchmark, a comprehensive open-source benchmark intended to standardize the evaluation of style embeddings. STEB encompasses 96 datasets across 7 languages, spanning applications such as authorship verification, authorship retrieval, AI-text detection, probing of linguistic features, and others. We find that semantic embeddings consistently fail in stylistic tasks, and that there is no style embedding that is universally superior across all tasks evaluated. We open-source the STEB code base at: https://github.com/rrivera1849/STEB.

Text-to-speech (TTS) evaluation is an open challenge. While the primary target was "naturalness," recent fidelity gains shifted focus toward "appropriateness" and whether speech is correct for its context. In this work, we examine how perception changes when the expected downstream use varies. We measure the appropriateness and human-likeness of five SOTA TTS systems across five domains: AI assistant, reader, actor, animated character, and spontaneous speaker. Results show appropriateness varies across domains independently of naturalness. While systems shine at reading, expressive domains remain challenging, and optimizing for one can degrade others. Furthermore, naturalness scores tend to penalize stylized speech while rewarding spontaneity. Finally, our study also highlights blind spots in one-size-fits-all evaluation metrics across more expressive domains. We demonstrate that TTS performance is not "solved" but depends on the target domain, requiring context-aware evaluation.

While code obfuscation impairs human code comprehension, it remains unclear if large language models share these failure modes. Building directly on a recent human study of program comprehension under code obfuscation, we evaluate whether large language models share the failure modes that obfuscation induces in human programmers. Evaluating several LLMs with five obfuscation tiers using the Block Model, we localize comprehension failures at the atom, block, relational, and macro levels. We find that reasoning-tuned models demonstrate significant alignment with human difficulty patterns across experience levels, whereas instruction and coder-tuned models show near-zero correlation. Chain-of-Thought trace length tracks task difficulty across tasks. Results indicate that performance under control-flow flattening degrades in proportion to state-space complexity, while adversarial identifier renaming disrupts comprehension through the interaction of semantic displacement and identifier-level interference. These findings suggest that reasoning-tuned LLMs approximate human sensitivity to code complexity more effectively than instruction-tuned variants.

Automatic speech recognition (ASR) systems often perform poorly in dysarthric speech, limiting their usefulness to affected speakers in everyday communication. This paper presents a personalized ASR system for a dysarthric speaker, built by adapting a foundation ASR model to speaker-specific data. Using the TEQST tool, we collected 92 hours of read speech and later added 8.8 hours of user corrections gathered through a deployed mobile application. Starting from Whisper, fine-tuning reduced word error rate to 15.8% with only 1.4 hours of adaptation data, reached 10.7% with 22.5 hours, and achieved the best result of 9.7% when using all available data including the corrections. Using LoRA adaptation and/or Qwen3-ASR as foundation model performed worse in this setting. The results show that personalized fine-tuning can make foundation ASR models substantially more effective for dysarthric speech and suitable for practical deployment.

In collaborative dialogue, shared perception does not guarantee shared interpretation. Mutual understanding must be established through interaction. We investigate whether vision-language models (VLMs) can distinguish what could be shared from what has been shared between dialogue participants through grounding. We formulate this as an interpretation-matching task on 13,077 annotated reference expressions from HCRC MapTask dialogues, and evaluate VLMs under systematically controlled manipulations of dialogue context and map-information access. Our results show that providing authentic map images improves overall performance but shifts models toward over-predicting alignment. Textual descriptions of the same map content reproduce this bias, while non-informative images suppress alignment predictions entirely, indicating that the bias is driven by task-relevant map content, not the visual channel. This improvement comes at the cost of degraded accuracy on non-aligned cases. Calibration analysis and reference-chain tracking further suggest that models rely on static referential cues on the maps rather than tracking how grounding unfolds through dialogue history. We observe these patterns most clearly in Qwen3-VL-8B-Instruct and, to varying degrees, in four additional models from two architecture families. In models that exhibit the bias, map content, whether presented visually or textually, is treated as evidence of mutual understanding, conflating potential with established common ground.

Relation extraction (RE) for low-resource languages is typically constrained by the lack of annotated corpora. We investigate the feasibility of cross-lingual RE for Romanian by combining automatic dataset translation with large language model (LLM) inference. We translate the SemEval-2010 Task 8 benchmark from English to Romanian using an LLM-based translation pipeline and evaluate Gemma 4 31B under zero-shot, few-shot, and QLoRA fine-tuned configurations, against four encoder baselines spanning 125M to 560M parameters: XLM- RoBERTa (base and large), Romanian BERT, and RoBERT- large. We assess two task formulations: relation classification with marked entities and end-to-end extraction. Our results show that Romanian incurs a 3 to 5 percentage point (pp) drop relative to English in prompt-only settings, that few-shot prompting provides marginal gains over zero-shot, and that QLoRA fine-tuning improves macro F1-Score by more than 22 percentage points in both languages while reducing the cross-lingual gap from 3.3 to 1.4pp. The encoder baselines come within 1-4pp of QLoRA Gemma on Romanian despite being 50-250 times smaller, with monolingual Romanian BERT at 125M parameters matching multilingual XLM-R at 278M. The case for using a 31B model for single-task RE on Romanian is therefore weak in deployment scenarios where compute matters. We release the translated dataset, evaluation code, and trained models.

Low-rank adaptation (LoRA) is commonly viewed as an update-space approximation to full fine-tuning, yet this view is incomplete for self-gated Transformer feed-forward networks. In gated FFNs, a low-rank residual can change not only projected features but also the nonlinear selection weights that determine which channels contribute to the output. We formalize this effect as selection misalignment and connect it to the local effective homogeneity of self-gated activations. This motivates a nonlinearity-aware principle for parameter-efficient fine-tuning: low-rank updates should allocate capacity to gate channels whose nonlinear states remain responsive and should shape the temporal evolution of selection. We propose NA-LoRA, a training-only method with two lightweight mechanisms: a derivative-based temporal-importance mask for gate-related LoRA updates and an activation-specific step-scaling rule when a meaningful coarse effective-homogeneity partition is available. NA-LoRA adds no auxiliary loss and incurs no inference-time overhead. Experiments on language-model fine-tuning and vision-language transfer benchmarks show that NA-LoRA consistently improves over vanilla LoRA and is competitive with or better than strong PEFT variants.

Faithfulness -- how precisely a generated image aligns with its prompt -- is increasingly central to the real-world utility of text-to-image (T2I) models. Existing faithfulness benchmarks, however, rely on simple atomic instructions, on which top-tier systems already achieve near-perfect scores. As T2I models enter creative workflows, users issue multi-faceted requests combining intricate spatial relationships, stylistic constraints, and complex text rendering. In this setting, a single binary VLM-judge score no longer captures which specific constraints the model fails to satisfy. We introduce Arena-T2I Hard, a 310-prompt stress benchmark drawn from real arena T2I logs, with approximately 30 decomposed yes/no constraints per prompt spanning six categories, including text rendering. The strongest closed-source system we evaluate reaches 0.855 with a 33~pp performance gap across 11 systems, demonstrating substantial discriminative power. Moreover, high public-arena rankings fail to predict faithfulness, confirming that holistic Bradley-Terry (BT) preference scores prioritize aesthetics over fine-grained prompt adherence. We propose a dependency-aware checklist reward that decomposes each prompt into a DAG of yes/no questions and zeroes descendants of failed parents, turning faithfulness into a per-constraint training signal. Combined with a BT aesthetic reward via group-decoupled normalization (GDPO), which standardizes each reward within its rollout group so neither collapses, the recipe attains a strictly better faithfulness-aesthetics trade-off on SD3.5-Medium and FLUX.1-dev under MMRB2 pairwise comparisons than every single-reward, naive weighted-sum, or 4-reward BT-ensemble baseline.

The automated transformation of C code to Rust is challenging due to Rust's strict ownership and borrowing semantics. While Large Language Models (LLMs) show promise, they often produce code that violates these rules or relies on unsafe constructs. We propose AdaTrans, a framework that addresses these issues through three core mechanisms: a Strategy-Driven Retrieval-Augmented Generation (RAG) mechanism to map compiler errors to specific repairs, an Error-Stratified Transformation Strategy (ESTS) that adapts its behavior based on error types, and a multi-stage validation pipeline to ensure both compilability and functional equivalence. Evaluating on a dataset of 104 algorithmic problems, AdaTrans achieves a mean compilation pass rate of 95.51% and a mean solve rate of 81.09%, significantly outperforming existing tools while maintaining an unsafe file rate of only 1.19%.

The deployment of face detection models in real-world applications raises important fairness concerns, as these systems may showcase performance disparities across demographic groups. A key obstacle to studying and mitigating such biases is the lack of face detection datasets with sensitive feature annotations. To address this gap, we introduce WIDER-FAIR, a new dataset built on the widely used WIDER-FACE benchmark, manually annotated with the perceived ethnicity and sex of each face. The dataset contains 16,256 images annotated across four ethnic groups: Asian, Black, Indian, and White, and two sex categories. We assess the quality and coherence of the annotations using face embeddings, a K-Nearest Neighbors classifier, and a t-SNE visualization, all of which support the consistency of the labeling process. As a demonstration of the dataset's potential, we train a YOLOv5 model and perform ablation studies on each sensitive feature. Among other findings, our experiments show that detection performance is notably lower for faces of Black individuals, and that excluding this group from training increases fairness disparity more than excluding any other ethnic group. These observations illustrate the value of demographically annotated datasets for understanding and evaluating bias in face detection models.

Biological neural circuits obey Dale's principle: each neuron's synapses are uniformly excitatory or inhibitory. Artificial networks that respect this constraint must coordinate separate excitatory and inhibitory populations, fundamentally changing how credit is assigned during learning. Several biologically plausible learning rules avoid backpropagation's weight transport requirement, but it has been difficult to achieve strong performance under Dale's principle beyond MNIST. Error Diffusion (ED) was originally proposed in a dual-stream excitatory/inhibitory architecture, where learning is driven by routing global error signals to all layers without transporting transposed forward weights or relying on random feedback matrices. Whether such a rule can scale under Dale's principle across both supervised classification and reinforcement learning remains unknown. Here, we introduce modulo error routing to extend Error Diffusion beyond binary classification, and show that a dual-stream excitatory/inhibitory architecture trained with this method achieves 96.7% on MNIST and establishes a 61.7% baseline on CIFAR-10, demonstrating that representation learning is possible even when strictly enforcing Dale's principle. For the classification setting, we introduce three domain-specific innovations: layer-specific sigmoid widths, batch-centered class error signals, and asymmetric initialization, and ablation analysis reveals that their relative importance reverses between MNIST and CIFAR-10, exposing task-dependent credit-assignment bottlenecks invisible to single-benchmark evaluation. In reinforcement learning, we integrate ED with Proximal Policy Optimization (PPO) and evaluate it on continuous-control tasks in Google Brax and on Craftax, an open-ended exploration task. We show that ED-PPO achieves competitive performance relative to Direct Feedback Alignment, a backpropagation-free baseline.

Sparse autoencoders (SAEs) have recently been proposed as interpretable tools for concept-level manipulation, under the assumption that isolated features can serve as controllable intervention points. In this work, we systematically evaluate this assumption in the context of object erasure and steering in diffusion models. We show that while SAEs reliably detect and localize semantic concepts within diffusion model activations, direct intervention in their latent space frequently induces out-of-distribution activations, resulting in severe visual artifacts. To disentangle detection from intervention, we use SAE activations purely as semantic detectors to identify image regions containing the target object, and replace those patch embeddings with the ones that do not contain it. This detection-based replacement preserves the diffusion model's activation statistics and produces significantly cleaner erasure results than latent steering. Our findings reveal a fundamental gap between concept detection and concept intervention in diffusion models: monosemantic or sparse features are not inherently suitable as control knobs for steering. These results position SAEs as powerful interpretability tools for analyzing generative models, but highlight important limitations when used for direct manipulation, such as unlearning.

For robots manipulating open-world objects, tactile representations must generalize to unseen materials. We introduce RCT (Robotic Contact Tactile), a robot-collected touch-vision-language dataset with 29,279 tactile frames from full robot presses on 122 industrial reference materials in 7 categories, recorded with three DIGIT sensors at multiple contact positions. RCT preserves each press as a contact sequence, enabling held-out evaluation across materials, categories, sensors, contact positions, and contact sequences. Frames from one press are strongly correlated: frame-random splits can place near-duplicate observations of the same physical interaction in both training and test. With the encoder held fixed, removing contact-sequence overlap reduces tactile-to-text Recall@1 by 17.7 percentage points. When materials are additionally held out at training time, performance drops sharply, leaving held-out-material Recall@1 at 25.1 +/- 6.1% averaged over three held-out draws. The public TVL/HCT split shows the same structure: every test contact sequence appears in training, and raw-pixel nearest neighbors recover the correct sequence in 98.3% of cases. Uniformly sampling a press improves contrastive training, and RCT-trained embeddings improve category probes on unseen materials. RCT makes contact-sequence-aware, held-out-material evaluation reproducible and exposes novel-material generalization as a central challenge for robotic tactile perception. The RCT dataset is open-sourced at https://faerber-lab.github.io/RCT/

The wave of AI-native applications is moving shopping beyond page- and feed-based browsing toward intent-driven experiences orchestrated by LLM agents. A common design wraps an LLM around existing search and recommendation pipelines, forcing complex intents through low-bandwidth retrieval or ranking interfaces and leaving a gap between language understanding and item-space fulfillment. Generative recommendation gives LLMs a direct item-space interface through semantic IDs (SIDs), but existing models mainly generate candidates for retrieval rather than translate flexible intents into item-space outcomes. We propose ShopX to address this bottleneck by unifying intent understanding, execution planning, and flexible SID-native item-space operations into a single foundation model. We deploy ShopX in agentic shopping workflows through a model-native item-fulfillment framework with a serving harness that defines a model-facing action protocol and exposes support surfaces for context access, catalog grounding, and state management. Within this framework, ShopX plans and composes SID-based item-space operations such as SID beam-search retrieval, listwise ranking, or product bundling. This model-centric design reduces lossy hand-offs between agent orchestration and item-space execution. To build ShopX, we design semantically recoverable, LLM-operable SIDs and a training recipe that equips a general LLM for flexible multi-turn item-space fulfillment while retaining the knowledge and instruction-following abilities needed by a shopping agent. We evaluate the ShopX framework against tool-mediated agentic systems on single- and multi-turn fulfillment tasks derived from anonymized Taobao production logs, showing that model-native fulfillment improves overall framework behavior, especially on complex or ambiguous requests.

This paper presents an overview of the second edition of the TalentCLEF challenge, organized as a Lab at the Conference and Labs of the Evaluation Forum (CLEF) 2026. TalentCLEF is an initiative aimed at advancing Natural Language Processing research in Human Capital Management. The second edition of the challenge consisted of two tasks: Task A, contextualized job-person matching, focuses on identifying and ranking the most suitable candidates represented by their resumes for a given job vacancy in English and Spanish. Task B, job-skill matching with skill type classification, addresses retrieving the most relevant skills for a given job title in English and distinguishing between core and contextual skills. TalentCLEF attracted 113 registered teams and received more than 400 submissions in the two tasks, reflecting the growing interest of the research community in shared evaluation benchmarks for Human Capital Management. This paper describes the motivation and organization of the challenge, summarizes the datasets and evaluation settings, and reports the main results obtained by the participating teams.

World-model evaluations often score a predicted future by overlap with a target state or observation. In sparse-change worlds, this can turn copied persistent state into apparent accuracy. We introduce ScratchWorld, an offline diagnostic benchmark that treats Scratch projects as executable worlds and uses a pinned Scratch VM to produce replay-verified transitions, hidden variables, causal traces, and counterfactual outcomes. ScratchWorld evaluates next-state prediction, long-horizon tracking, causal event attribution, and counterfactual prediction; each replay-verified target can be presented under raw-program, structured-state, natural-language, or rendered input modalities, and our experiments use the structured-state condition. Its primary state metric is value-aware changed-field $F_1$, which gives credit only for the changed field and its executed value. In a 659-example release, seven prompted language/reasoning models reach at most 13.8% value-aware changed-field $F_1$ in a state-only partial-observation stress test. A same-instance copy diagnostic makes the overlap confound concrete: copying the input state reaches 98.0% implied full-state field accuracy and 0.0% changed-field $F_1$, with the largest inflation on real projects. Auxiliary diagnostics separate hidden-state rollout drift, intervention sensitivity, causal attribution, and perturbation robustness. Across these settings, models often react to actions or interventions without following the executable rule that determines the changed value.

Feature rankings are widely used in supervised feature selection because they are simple, scalable and easy to interpret. Variables are first ranked by a relevance score, and a subset is then obtained by retaining the top-ranked variables. Although the first stage has been extensively studied, the second is often governed by an arbitrary cardinality, an empirical threshold or cross-validation, without a direct interpretation. This raises a basic question: given a feature ranking, when is there enough accumulated class-separation evidence to stop selecting features? This paper develops a distributional framework for transforming supervised feature rankings into class-independent subsets through an explicit risk-calibrated stopping rule. For each variable and each pair of classes, marginal separation is measured by the Bhattacharyya coefficient between the corresponding class-conditional distributions. The proposed method selects a single global subset shared by all classes by retaining the shortest prefix of a ranking whose residual product overlap falls below a prescribed threshold for every relevant class contrast. We derive binary and multiclass Bayes-risk bounds for the labelled product marginal problem, and obtain prior-dependent and prior-free calibrations of the residual-overlap threshold from a target all-pairs risk level. An empirical comparison on high-dimensional genomic datasets illustrates that the rule can reduce tens of thousands of variables to a few dozen while maintaining predictive performance statistically comparable to the all-features baseline. As the stopping rule only requires one-dimensional marginal overlap estimates and scans a precomputed ranking, it is well suited to very high-dimensional settings where exhaustive subset search is infeasible and interpretable truncation of feature rankings is essential.

Diffusion models have emerged as a dominant paradigm in generative modeling, enabling high-fidelity sampling from complex data distributions. Despite impressive capabilities, controlling diffusion models to produce outputs aligned with user intent remains an open challenge, especially when balancing global coherence with local precision. Existing control mechanisms vary in the granularity of their conditioning signals. For example, textual prompts guide generation globally through high-level semantics, while ControlNet-like approaches secure precise local structure via dense conditions. In this work, we introduce Histogram-constrained Image Generation (HIG), a novel control mechanism that falls into the middle ground of control granularity. Our framework enforces user-specified distributional constraints (e.g., color histograms or latent token distributions) during the generation process with exact precision. We model such control as an optimal transport (OT) problem and apply explicit guidance transformations during sampling, thereby driving the diffusion trajectory to align with the desired histogram. We demonstrate the versatility of HIG across diverse applications, including constrained generation via color/latent histograms and high-capacity information embedding through histogram-level encoding. Our findings underscore the promise of distributional control, a flexible and interpretable control scheme that is fully compatible with existing control mechanisms, diversifying the hybrid strategies for controllable image generation. Our project page is available at: https://maps-research.github.io/hig/.

As ML-KEM is adopted as a post-quantum cryptographic standard, resilience against physical side-channel attacks has become essential. Among the constituent steps, the decapsulation Fujisaki-Okamoto (FO) verification is particularly vulnerable to side-channel power and electromagnetic (EM) analysis. In this work, we focus on common FPGA-based implementations and examine their side-channel vulnerabilities, and compare them with those of microcontroller implementations. Three verification implementations, unprotected, hash-based (first-order), and higher-order masked, are evaluated for side-channel security on both a microcontroller and an FPGA. While FPGAs offer higher speed and parallelism, they often exhibit stronger side-channel leakage, especially in high bandwidth configurations. The higher-order masked designs still leak information about the underlying data due to hardware-level effects and data-dependent processing. Our experiments show that their parallelized processing on FPGAs introduces sufficient first-order leakage for full secret-key recovery. These results underscore the persistent challenge of securing PQC algorithms in performance-constrained and parallelized hardware environments.

Despite rapid progress in interactive world models (IWMs), existing benchmarks evaluate action following only at trajectory level and ignore memory and interaction physics. We introduce WorldRoamBench, an open-world benchmark for long-horizon stability across four dimensions, each with tailored innovations: (i) Action: per-frame action metric bypassing cross-model semantic scale disparity and exposing failures hidden by trajectory; (ii) Vision: segment-based drift metric capturing non-monotonic mid-sequence collapse missed by start-vs-end comparisons; (iii) Physics: controllability-gated evaluation over mechanics, optics, and 3D consistency, scoring plausibility under faithful action execution; (iv) Memory: action-decoupled protocol evaluating scene memory via transition-localized 3D point-cloud reconstruction and subject memory via tracking-plus-VLM reasoning. The benchmark comprises 600+ test cases across Nature, Urban, and Indoor scenes in first/third-person views with WASD 10-60s continuous interaction. Evaluating 10+ open/closed-source models reveals none reliably satisfies all dimensions; even the best achieves only moderate scores. Advances on WorldRoamBench are steps toward IWMs that are stable, physically grounded, memory-faithful, and deployable in real-world applications.

Geopolitical event forecasting is a challenging task, as it requires understanding complex regional contexts, dynamic event signals, and uncertain future outcomes. Recent advances in large language model agents provide new opportunities for building forecasting systems that can reason with diverse sources and expert perspectives. In this paper, we present \textit{ForecastAgentSearch}, a preliminary framework that formulates geopolitical event forecasting as a multi-expert agent search problem. Given a forecasting query, the system first analyzes the task context, then searches and ranks relevant expert agents based on their regional knowledge, domain expertise, reliability, and complementarity. The selected agents provide specialized analyses, which are further coordinated to generate a final forecast with explanations and uncertainty awareness. We discuss the key design challenges of agent profiling, expert retrieval, ranking, and multi-agent coordination, and outline possible evaluation protocols for future development. This work aims to provide an initial step toward searchable and reliable agent-based forecasting systems.

Performance in face and speaker verification is largely driven by margin-penalty softmax losses such as CosFace and ArcFace. Recently introduced $α$-divergence loss functions offer a compelling alternative, particularly due to their ability to induce sparse solutions (when $α>1$). However, standard geometric margins are designed for the softmax function and do not naturally extend to this generalized probabilistic framework. In this paper we propose Q-Margin, a novel $α$-divergence loss that introduces a principled probabilistic margin. Unlike conventional methods that apply geometric penalties to the logits (unnormalized log-likelihoods), Q-Margin encodes the margin penalty directly into the reference measure (prior probabilities). This formulation naturally encourages discriminative embeddings while preserving the beneficial sparsity properties of the $α$-divergence. We demonstrate that Q-Margin achieves competitive or superior performance on the challenging IJB-B and IJB-C face verification benchmarks and similarly strong results in speaker verification on VoxCeleb. Crucially, against ArcFace and CosFace baselines trained under an identical recipe, Q-Margin consistently improves at low False Acceptance Rates (FARs), a capability critical for practical high-security applications. Finally, the extreme sparsity of the Q-Margin posteriors enables exact and memory-efficient training, offering a scalable solution for datasets with millions of identities.

This work presents Dual-Embedding Watermarking (DEW), a semantic watermarking scheme for large language models (LLMs) that leverages contextual and token-level embeddings to enhance robustness against paraphrasing and translation. DEW utilizes a signal-processing methodology, applying algebraic vector-space operations to token and context embeddings to derive a watermark signal that degrades gracefully under semantic shifts. The method obfuscates the watermark by projecting embedding vectors through pseudo-random matrices seeded with a secret key. Relevant distributions derived from the underlying algebra are evaluated and employed for statistical testing and benchmarking of DEW. Experimental results across multiple LLMs indicate that DEW improves post-paraphrase detection while maintaining competitive text quality, and remains detectable after translation, even when prior semantic watermarks degrade significantly. These findings position DEW as a practical and robust solution for safeguarding LLM-generated text and addressing critical issues in responsible AI deployment.

Any-angle path planning extends traditional graph-based path planning by allowing movement between any pair of vertices, rather than being restricted by predefined edges. It can find straighter and shorter paths in continuous space with graphs, making it particularly suitable for navigation in open areas such as airspaces, warehouses, and oceans. Many any-angle path-planning algorithms have been proposed, but only a few can guarantee optimal solutions, especially in the presence of dynamic obstacles. To address this challenge, this article focuses on optimal any-angle path planning on grids and introduces two general techniques that accelerate computation while preserving optimality in both static and dynamic environments: 1) elliptical forward expansion, which leverages ellipse-based neighborhoods to restrict the search space, and 2) field of view, which replaces traditional line-of-sight methods to speed up visibility checks. To integrate these two techniques, inverted and forward scanning are introduced. Inverted scanning establishes visual connections from open nodes, whereas forward scanning initiates scans from closed nodes. Building on the proposed techniques, Zeta* and Zeta*-SIPP are developed for static and dynamic environments respectively. Zeta*, when combined with forward scanning, is similar to the state-of-the-art algorithm Anya and attains comparable performance. Unlike Anya, Zeta* can be readily extended to other settings, such as dynamic environments (e.g., Zeta*-SIPP). Zeta*-SIPP, with either scanning method, is more than 20 times faster than the corresponding state-of-the-art optimal planner TO-AA-SIPP. Overall, this research identifies the key requirements for achieving optimal any-angle path planning and introduces a unified approach suitable for different environments.

As technology advances, many path-planning algorithms have been proposed for Air Traffic Management, yet their operational adoption in tactical control remains limited, revealing a misalignment between algorithmic design priorities and air traffic controllers' needs. This underscores the need for decision-support solutions that are inherently interpretable, computationally efficient, and explicitly designed for human use. Focusing on this design challenge, this study develops a conflict-free path-planning algorithm for en-route Air Traffic Control (ATC) designed to be compatible with two guiding considerations: (1) the interpretability and flexibility offered by solution-space displays, which motivate constructing an algorithm that exposes all feasible safe actions and accommodates shifting optimization goals; and (2) the decision logic controllers naturally apply when enforcing operational constraints, such as separation standards, maneuverability limits, waypoint minimization, and routing practicality. Centered on these principles, the algorithm integrates three intent-based conflict detection methods -- distance-based, time-interval-based, and zone-based -- within a solution-space framework to identify conflict-free paths in computationally efficient ways. Additionally, vertex-based and edge-based search nodes are proposed for solution space path planning (SSPP), resulting in two variants -- SSPPV and SSPPE, respectively, which are evaluated in terms of computational speed and solution quality. Empirical results show that SSPPV paired with zone-based conflict detection achieves the best performance, computing paths in 3.69 ms on average in operational-relevant scenarios based on the Delta sector of the Maastricht Upper Area Control Centre (MUAC) using a 5 nmi grid.

As Quantum Machine Learning (QML) transitions toward practical implementation, the field faces a critical architectural bottleneck that challenges the fundamental assumptions of classical statistical learning theory. In classical deep learning, increasing model capacity typically risks overfitting. However, this study advances a counter-intuitive paradigm: unstructured contemporary QML architectures suffer from a profound state of quantum underfitting, driven by the "expressivity-trainability paradox." We demonstrate that the vast Hilbert space capacity of Parameterized Quantum Circuits (PQCs)-traditionally chased as the source of quantum advantage is the direct mathematical cause of Barren Plateaus (BPs), where gradient landscapes become exponentially flat. By synthesizing recent breakthroughs in Dynamical Lie Algebras (DLAs) and Geometric QML, we establish a comprehensive framework linking the algebraic dimension of circuit generators to their optimization dynamics. Furthermore, we empirically validate this framework on a non-linear binary classification task, illuminating a uniquely quantum manifestation of the bias-variance tradeoff: while unstructured architectures achieve near-perfect training accuracy via unscalable parameterization (quantum overfitting), embedding group-theoretic geometric priors acts as a structural regularizer. By restricting the DLA growth to a polynomial regime, our symmetry-preserving approach sacrifices raw memorization capacity to guarantee scalable, gradient-rich training landscapes, offering a robust roadmap for "Trainability-by-Design" in scalable quantum neural networks.

Large Language Models (LLMs) are increasingly deployed as autonomous financial agents initialized with explicit behavioral mandates such as "preserve capital" or "avoid speculative bets" that are meant to govern every decision throughout deployment. In practice, however, as market context accumulates over long horizons, these mandates gradually lose their behavioral influence, a phenomenon we formalize as Mandate Salience Decay (MSD). To measure MSD objectively, we introduce FinPersona-Bench, a simulation benchmark in which a synthetic market decouples observable price from hidden fundamental value, enabling falsifiable evaluation across three failure modes: trading without signal in calm markets, panic-selling during crashes, and ignoring fundamental value during speculative bubbles. Evaluating 18 leading frontier and open-source LLMs, each assigned one of three behavioral profiles ranging from strict capital preservation to aggressive growth, shows that MSD compounds over time and is model-dependent. In crash scenarios, the behavioral gap between static agents and those receiving periodic mandate re-grounding grows 4.4x from the first to the final quarter of the simulation. The effects of mandate re-grounding are not uniformly positive: it consistently helps conservative agents in low-signal markets but actively worsens behavior for aggressive agents in the same setting. These findings suggest that reliable long-horizon deployment requires selective, mandate-aware re-grounding based on agent profile and market regime.

This paper studies how spectral geometry emerges in quantum learning models and how it can be diagnosed with physically grounded probes. In graph-regularized quantum networks, training reorganizes the output similarity graph, increases the effective spectral dimension Delta S = +0.23, and reshapes the Laplacian spectrum. Edge-resolved two-boson interference directly probes this restructuring: the bosonic enhancement Delta P_uv correlates with the Fiedler edge split |Delta v_2| (r = -0.50), linking learned spectral partitions to interference signatures. A phase diagram shows a nonmonotonic dependence of performance on coupling strength gamma and noise delta, with graph regularization improving fidelity only in a restricted regime; hardware experiments confirm the predicted interference behavior within shot-noise uncertainty. We also analyze a hybrid quantum autoencoder and introduce Bloch-space drift as a geometric diagnostic of its latent representation. With an unsupervised benign-data threshold, the model achieves high ranking performance (ROC-AUC about 0.99) and negligible false-negative rates. Absolute Bloch drift strongly discriminates anomalies (ROC-AUC at least about 0.9), while consecutive drift is near random (ROC-AUC about 0.5), showing that detection arises from persistent state-space displacement rather than local fluctuations. Through the geometry of reduced single-qubit states and associated quantum Fisher information, these results show that learning-induced spectral organization appears as measurable quantum-state structure, establishing a unified spectral-geometric framework for diagnosing quantum learning systems with bosonic and Bloch probes.

High pass rates on established programming benchmarks such as HumanEval and LiveCodeBench do not always show whether a model can reason about algorithms. Many fixed benchmarks eventually become part of the public training ecosystem through released problem statements, editorials, and generated solutions, allowing later models to improve partly by exposure rather than by stronger algorithmic ability. We introduce ALGOBENCH, a framework that automatically builds novel algorithmic problems from known competitive-programming problems through structured constraint-shifting transformations. Each accepted ALGOBENCH variant is traceable to a source problem, but must make the original reference algorithm fail. Beyond pass@$k$, we introduce complexity-aware metrics -- including OPTT, OPTS, TRAPRATE, GAPT, and CONSENS -- to test whether a solution is not only functionally correct but also asymptotically suitable for the generated problem. Experiments across multiple LLMs and prompting strategies show that performance drops sharply on ALGOBENCH variants, retrieval can increase reuse of the old algorithm, and many correct-looking solutions fail to meet the required complexity. Error analysis shows that failures are mainly algorithmic rather than implementation-level, suggesting that ALGOBENCH evaluates adaptation beyond functional correctness.

Medical multiple-choice question answering requires parameter-efficient adaptation across heterogeneous knowledge domains and reasoning operations. A medication question, a diagnostic decision, a public-health item, and a nursing-action item may require different low-rank updates, while some recall items should preserve the base model's representation with only mild adapter intervention. We propose BiRG-LoRA, a single-adapter rank-gated LoRA method for medical question answering. BiRG-LoRA keeps one LoRA module per target layer but makes its rank dimension input-conditioned: for each question, a biaxial gate combines hidden semantic evidence with specialty/profession priors, clinical-operation priors, and their interaction to select a sparse top-$k$ subset of rank atoms. A scalar injection coefficient further controls the strength of the selected adapter update. Under a matched Qwen3-8B CMB-source protocol, BiRG-LoRA achieves the highest four-benchmark macro-average accuracy among trainable PEFT baselines and matched routing controls: 69.31% averaged over CMB, CMExam, MedQA, and MedMCQA. It improves over MoELoRA by 0.89 percentage points while using 28.1% fewer trainable parameters; a paired, benchmark-stratified bootstrap over final predictions gives a 95% confidence interval of [0.42, 1.37] for this macro-average gain. Basic controls show that BiRG-LoRA also improves over vanilla LoRA r16 and active-rank-matched LoRA r4 by 0.83 macro points, and an evaluation-time weak-axis perturbation check suggests that performance is not brittle to moderate tag noise. The results support a bounded claim: clinically structured rank allocation improves cross-benchmark medical QA under a matched single-seed protocol, while training-seed variance remains future work.

Graphical user interface (GUI) agents build on vision-language models to complete user tasks end-to-end in real applications through interface actions such as tapping, swiping, text entry, and navigation. However, existing GUI agents are trained and evaluated largely on offline trajectories, simulated environments, and standardized benchmarks. These differ substantially from real applications in interface layout, interaction logic, and abnormal-state distribution, and cannot faithfully characterize execution stability in real-world use, where account states, permission dialogs, payment authentication, and risk control continually reshape the state distribution and open a persistent gap between benchmark scores and real usability. To close this gap, we propose Xiaomi-GUI-0, a native multimodal GUI agent for real mobile environments, trained and evaluated within a real-device closed loop. At its core is a real-device-dominant hybrid infrastructure, where physical devices are the primary execution environment and sandboxes provide auxiliary support, so that data collection, training, rollout, and evaluation share an execution distribution close to real deployment. We construct multi-source training data spanning high-frequency head tasks, high-generalization data for long-tail intents, and capability-enhancement data for reflection and memory, and introduce an error-driven data flywheel that turns failure trajectories into corrected actions, reflective explanations, and recovery demonstrations. The model is trained through a progressive three-stage pipeline of supervised fine-tuning, step-level reinforcement learning, and agentic reinforcement learning. Evaluated on public benchmarks and our in-house RealMobile, Xiaomi-GUI-0 achieves 72.0% success on RealMobile and 78.9% on AndroidWorld, while substantially improving execution stability and abnormal-state recognition in real-world tasks.

Hierarchical Vision-Language-Action (VLA) models decouple high-level planning from low-level control to improve generalization in robot manipulation. Recent work in this paradigm uses 2D end-effector trajectories predicted by a Vision-Language Model (VLM) as explicit guidance for a downstream policy. However, state-of-the-art low-level policies operate in 3D metric space on point clouds, and feeding them 2D guidance that lacks depth forces each waypoint to be assigned the depth of whatever scene surface lies beneath it, producing geometrically distorted trajectories. We propose 3D HAMSTER, a hierarchical framework that closes this gap by having the planner directly output metrically reliable 3D trajectories. We augment a VLM with a dedicated depth encoder and a dense depth reconstruction objective to predict 3D waypoint sequences, which are directly integrated into a pointcloudbased low-level policy. Across 3D trajectory prediction, simulation, and real-world manipulation, 3D HAMSTER consistently outperforms proprietary VLMs and 2D-guided baselines, with the largest gains under appearance-altering shifts and unseen language, spatial, and visual conditions. The project page is available at https://davian-robotics.github.io/3D_HAMSTER/.

Oracle Bone Inscriptions (OBIs) recognition plays a crucial role in understanding ancient Chinese culture. However, accurately recognizing OBIs remains highly challenging due to their complex, irregular, and often degraded shapes. Traditional methods rely on expert knowledge and manual analysis, which are time-consuming and error-prone. Although deep learning has greatly advanced general image recognition, existing methods struggle to capture the fine-grained details and subtle variations inherent in OBIs, resulting in limited performance. Even most recent and effective layer attention techniques are designed to capture fine-grained dependencies through enhanced inter-layer interactions, yet they still exhibit only marginal improvements in OBIs recognition. To address these limitations, we propose Multi-Scale Layer Attention (MSLA), a novel paradigm that explicitly models both multi-scale and cross-layer feature interactions. By enriching the representation with fine-grained details across multiple spatial scales, MSLA enables more accurate and robust OBIs recognition. Extensive experiments on large-scale OBIs datasets demonstrate that MSLA consistently outperforms existing attention mechanisms while maintaining computational efficiency.

This paper studies energy efficient tracking of power-limited mobile users with the assistance of a Reconfigurable Intelligent Surface (RIS). Since localization pilot transmissions dominate the energy budget of power-constrained devices, we introduce a low-overhead feedback link from the Base Station (BS) to the user to enable dynamic uplink power control. To navigate the discrete and decentralized nature of this active sensing problem, we propose a novel Dual-Agent (DA) deep learning framework that jointly optimizes the discrete RIS phase profiles and the UE's transmit power in real time. Specifically, our approach employs a hybrid training methodology integrating the neuroevolution paradigm with supervised learning, effectively overcoming the non-differentiability of discrete phase responses from the RIS unit elements and the strict information bottleneck of single-bit feedback messages for pilot power control. The proposed DA active sensing framework can be applied with both single- and multi-antenna BSs, the latter with only minor modifications in the structure of one NN: an additional output branch with appropriate structure is included for the latter case to select a valid digital combiner from a finite set. Extensive numerical simulations demonstrate that the proposed scheme achieves highly accurate and robust tracking across diverse target motion models, outperforming extended Kalman and particle filters, as well as, machine learning-based trackers. Furthermore, in static localization, it is shown to significantly outperform traditional fingerprinting schemes, deep reinforcement learning baselines, and standard backpropagation-based estimators.

We propose Intelligent Schema Memory (ISM), a self-evolving memory-augmented system that improves mathematical reasoning for a frozen LLM under continual learning with hard episodic resets. ISM maintains a compact, self-refined bank of strategy schemas learned from both successful and failed episodes, with symbolic tools that check intermediate steps and certify answers. Without updating model parameters, ISM outperforms passive, retrieval, and reflection baselines on MATH-Hard and OlympiadBench, using 64% and 86% fewer schemas respectively than the strongest passive baseline. These results show that small, actively maintained, and verified strategy memories can support reliable continual mathematical reasoning under strict episodic isolation.The codebase is available at https://github.com/pdx97/ISM .

Large language models deployed in regulated industries operate under two constraints: compliance enforcement and cost efficiency. Personally identifiable information (PII) in user queries can reach model endpoints before the system determines whether that data should leave its jurisdictional boundary. Serving all queries through a single large model consumes full GPU capacity regardless of query complexity while offering no mechanism for geographic routing. Mixture-of-Experts architectures do not address this routing occurs between expert layers within the model after data has already arrived at the endpoint, with all experts loaded in memory regardless of query complexity. We propose a classifier-gated routing architecture that enforces compliance by design. A trained encoder classifier sits before any decoder inference, evaluating each query for complexity and data sensitivity, then routing it to an appropriately sized dense model in the appropriate geographic location. PII-containing queries route to local endpoints before any LLM computation begins, making data residency violations structurally impossible. Simple queries reach small, fast models at a fraction of the cost. Our evaluation on 600 queries demonstrates 39% median latency reduction, 33-52% cost savings depending on query distribution, and generation throughput of 122-200 tokens/second versus 50-64 for the baseline. The encoder classifier achieves 99.2% accuracy with near-perfect PII recall at 7ms inference overhead, establishing pre-inference classification as a practical path to compliance-by-design LLM deployment.

Full fine-tuning adapts large language models to new tasks but can erode capabilities they already possess. Existing remedies protect through proxies such as parameter distances, importance penalties, output matching, or dominant singular directions of the weights, but none directly asks which activation directions the preserved capability relies on. We argue that a capability is characterized more faithfully by the activation subspace it induces than by the singular geometry of the weight matrix, and develop function-space protection, instantiated as FORA (Function-space Orthogonal Residual Adaptation). From label-free calibration inputs, FORA estimates, per layer, the principal directions $Q$ of the input-activation covariance and forms a right projector $P_Q = I - QQ^T$. Paired with a left projector $P_U$ from the weight SVD, the update is $ΔW = P_U M P_Q + U_2 D_δ V_2^T$: a high-capacity branch structurally barred from reading capability-relevant function directions, plus a narrow spectral channel for controlled plasticity. The construction extends to parameter-efficient adaptation via $M \to (α/r) BA$. Across three settings on Qwen3-1.7B, including COGS and GSM8K learned while preserving translation and translation learned while preserving math, FORA consistently improves preservation over weight-space projection and standard regularization, with only a small new-task trade-off in the math-preservation setting. A controlled ablation isolating the projection source shows that the advantage comes not from projection itself, but from projecting onto capability-derived rather than weight-derived directions. Code is available at https://github.com/zrui239/FORA.

Few-shot selection typically assumes that reranking retrieved examples always improves performance. We challenge this view by identifying that the expensive reranking step can in fact degrade performance. Instead, we propose \emph{Training-Free Gated Reranking}, which decides whether to rerank the few-shot examples based on the model's uncertainty. Extensive experiments across 8 LLMs, covering 7 NLU datasets and 9 MT domain-language combinations, demonstrate that our approach reduces computational costs by 15\%-80\% while improving average performance by up to 2\%. These findings indicate that higher computational cost does not guarantee better performance, and that reranking is most beneficial when targeted at high-uncertainty instances.

Residual reinforcement learning adapts a pretrained robot policy by learning an additive correction to its actions. While effective when adaptation amounts to shifting the base policy's action distribution, additive corrections cannot change the distribution's shape, scale, or state-dependent geometry -- limitations we formalize as wrong variance, miscalibrated confidence, and non-uniform correction. We show that these matter under dynamics shift: when the base distribution is geometrically mismatched to the shifted system, residual correction can underperform even the unadapted policy. We propose Warp RL, a policy adaptation method that replaces additive residuals with an invertible, state-conditioned transformation of the base policy's action distribution. Instantiated with monotonic rational-quadratic spline flows (arXiv:1906.04032), Warp RL preserves identity initialization, strictly generalizes additive residual correction, and exposes a structured adaptation space suitable for both policy-gradient and gradient-free optimization. Across a variety of ManiSkill3 manipulation tasks with controlled dynamics shifts, Warp RL matches residual correction when translation is sufficient and substantially outperforms it when adaptation requires distributional reshaping. We further demonstrate that warping can replace additive correction in an off-policy sim-to-real pipeline, achieving comparable success rate with 30% faster task completion on a real-robot peg-insertion task.

Modern VLSI designs comprise tens of billions of components, making scalable hypergraph partitioning critical for parallel and hierarchical optimization. Although multilevel partitioning remains the dominant paradigm, its coarsening stage can distort structural information, especially in hypergraphs with many high-degree hyperedges, leading to increased refinement overhead and limited scalability. Recent approaches incorporate spectral information to guide coarsening, but only in a heuristic manner, without directly optimizing the partitioning objectives. We introduce HySpecPro, a single-level hypergraph partitioner that performs end-to-end optimization in a spectral embedding space. HySpecPro constructs embeddings from a bipartite Laplacian and performs efficient projection-based search, supported by a fully GPU-accelerated implementation. Experiments show that HySpecPro delivers cut quality comparable to state-of-the-art multilevel methods while scaling linearly with the total hyperedge degree.

This paper introduces a formal modeling framework designed to estimate the complexity and cost associated with system changes induced by external requirements. We model a system as a directed graph of couplings, capturing the intricate dependencies and information flows between components and elements within a specific context. The proposed method enables the estimation of bounded change complexity through component interfaces, even when internal logic remains opaque. Additionally, the framework provides a mechanism for bounding the cost of system-wide modifications by associating external drivers and cost factors with individual system elements. We propose a multi-view approach to the model, providing graphical, algebraic, and tabular representations to suit different levels of abstraction and computational needs. By bridging the gap between component-based modeling and project cost estimation, our method provides actionable insights for architecture design, software engineering, and lifecycle operations. The model is validated through a case study involving the integration of a large-scale retail banking platform.

Large language models (LLMs) embedded in multi-turn agentic harnesses are reshaping software engineering (SWE), but routing every task to a frontier model is wasteful when many issues admit cheap fixes. Existing LLM routers operate on the task description alone, which inherits an information-theoretic Bayes-error floor in agentic settings: a similar issue can hide either a localized typo or a multi-module refactor, and the prompt does not separate the two. We introduce SWE-Router, a value-based temporal approach that lets a cheap model run for a few exploratory turns and reads the resulting partial trajectory before deciding whether to continue cheaply or to escalate to an expensive model. We provide a Bayes-optimality theorem showing that conditioning on the partial trajectory never harms routing and is strictly better whenever exploration is informative. Across the LLM pairs of weak and strong models spanning the contemporary cost--capability frontier, we show that SWE-Router greatly improves the cost efficiency of SWE tasks, while maintaining the majority of the performances of the stronger model. We additionally release a multi-LLM trajectory dataset which allows reproduction of our trajectory-level routing.

GraphRAG is an extension of retrieval-augmented generation (RAG) that supports large language models (LLMs) by referring to graph-structured data as external knowledge. While this technique ideally captures intricate relationships, it often struggles with graph representations for LLMs, particularly for frozen LLMs, due to the misalignment between graph-based and text-based latent features. We tackle this issue by introducing the {\it Adaptive-masking for Graph Embedding (AGE)}. AGE employs a Transformer in a mask-based self-supervised learning (SSL) approach. We designed the architecture similar to text embedding encoders, addressing the latent feature misalignment. In contrast to natural language texts, graphs are concise representations, and there exist {\it key nodes} that hold dominant contextual information, which are challenging to predict from their surroundings. Masking such key nodes leads to inefficiency in the SSL process. Therefore, AGE focuses on predicting nodes apart from key nodes, utilizing a learnable node sampler. Our experimental results indicate that AGE significantly improves approaches using non-parametric search component in GraphQA tasks, achieving superior accuracy across four benchmark datasets with distinct characteristics.

Accurate modeling of wind turbine power curves is crucial for optimal wind farm operation. Nearly all existing power curve models focus on temporal variables such as wind speed and temperature while overlooking the influence of terrain covariates, which governs inflow wind conditions and thus also affects wind power production. This paper proposes a nonparametric spatio-temporal Gaussian process model that integrates temporal environmental covariates with spatial terrain features. The model falls in the category of spatial-temporal Gaussian process models with data on a grid. The challenge to be addressed is that the spatio-temporal modeling require certain temporal alignment among the data, a property that the wind farm data does not have. Our solution strategy is to construct a shared representative temporal covariate set which not only aligns the temporal inputs but also has a size an order of magnitude smaller than the original data size. With this transformation, our resulting model is able to employ a separable kernel structure that captures both spatial and temporal dependencies. Empirical analysis on a real wind farm dataset shows that our method improves predictive accuracy over existing baselines and can be used to quantify the various impact of the terrain characteristics on turbine performance.

Large Language Models (LLMs) are increasingly used in Agile Software Development for documentation, coaching, and training. As practitioners adopt these tools to prepare for certifications such as Professional Scrum Master (PSM), a key question is whether LLMs can reliably reason about Scrum, a framework with normative, well-defined rules described in the Scrum Guide (2020). This paper examines how different prompt techniques affect the factual accuracy of LLM responses to Scrum certification-style questions. A dataset of 993 validated PSM-aligned questions was answered by GPT-5 using three techniques: zero-shot, chain-of-thought, and with-source citation. All prompts achieved certification-level accuracy above 85\%, with the citation-based variant performing best (89.1\%) and yielding the lowest error rate. Correct answers concentrated in well-defined topics, such as \emph{Definition of Done}, Events, and Product Backlog Management, and in single-answer multiple-choice items, while multi-select questions and more interpretive areas, such as Scrum Team and Product Value, were less stable. Among questions where at least one prompt failed (16.2\%), errors clustered into misalignment with the Scrum Guide (28\%), content outside its scope (34\%), and outdated or biased interpretations (38\%). Overall, prompt techniques produced modest but consistent improvements, particularly in reducing misinterpretation and version drift, supporting more reliable use of LLMs in Agile learning and certification preparation.

Large Language Models (LLMs) are increasingly used in exam- and certification-style question answering tasks, where their ability to retrieve, interpret, and apply domain-specific knowledge can be systematically assessed. In Software Engineering, such settings are particularly relevant when questions depend on strict adherence to normative definitions, roles, artifacts, and rules. This paper evaluates the performance of three contemporary LLMs, \textit{GPT-5 mini}, \textit{Gemini 3 Flash}, and \textit{DeepSeek Chat 3.2}, in answering 993 Scrum certification-style questions aligned with the Professional Scrum Master I (PSM I) assessment format. We evaluated the models under three prompting strategies (\textit{zero-shot}, \textit{chain-of-thought}, and \textit{source-grounded}), with repeated executions to assess intra-model stability. We also analyzed performance across Scrum topics and question formats, complemented by a qualitative analysis of recurring error patterns in incorrect answers. Results revealed clear differences among models, with Gemini 3 Flash achieving the highest accuracy, followed by GPT-5 mini and DeepSeek Chat 3.2, while intra-model variability remained low across all conditions. By question format, the models achieved the highest accuracy on single-answer multiple-choice items, whereas multi-select and True/False questions were more error-prone. By topic, performance was more consistent in normatively explicit areas such as Artifacts, Empiricism, and Product Value, but more fragile in Scrum Values, Self-Managing Teams, and Stakeholders \& Customers. The qualitative analysis showed that errors were systematic rather than random, involving overgeneralization, restrictive wording, compound distractors, and conflicts between common market interpretations and strict Scrum definitions.

Climate governance processes involve complex interactions between heterogeneous citizens, advocacy groups, media actors, and political decision-makers. While agent-based models (ABMs) have been widely used to study environmental policy and socio-ecological systems, many existing approaches focus either on institutional dynamics or individual behavioural mechanisms in isolation. This paper presents a modular multi-level agent-based architecture that integrates empirically grounded cognitive decision models with strategic institutional behaviour within a unified simulation framework. The architecture combines (i) motive-based individual decision-making operationalised through the HUMAT and MOA frameworks, (ii) socially embedded influence processes via demographic homophily networks, and (iii) institutional strategy modules for environmental non-governmental organisations (NGOs), media agents, and politicians. Political decisions emerge from the aggregation of multiple signals, including expert input, public mobilisation, party alignment, and media framing. The model is designed to be empirically calibrated through synthetic populations derived from survey data and and institutional parameters informed through Living Lab stakeholder engagement, and to support scenario-based exploration of climate-relevant land-use governance processes. Rather than presenting empirical results, this paper focuses on the architectural design principles, modular structure, and integration logic of the model. We discuss how this multi-layered approach contributes to the modelling of democratic climate governance and outline pathways for generalization and future validation.

Agents typically assume an expert user -- one with well-formed preferences about what they want -- and default to clarifying questions whenever the task is underspecified. We argue this assumption is unrealistic. Users often lack the domain knowledge to have completely specified preferences; if asked about their preference on some feature, the user may be unable to answer without the agent helping the user to learn some domain knowledge needed to form a preference for that feature, e.g., via examples or explanations. To formalize these principles, we draw on the Search-Experience-Credence framework from Information Economics to introduce CoPref, a model of how users construct preferences based on agent dialog actions. We then study these ideas concretely in agentic recommender systems, proposing CoShop, an interactive benchmark. In CoShop, an agent converses with and makes recommendations for a CoPref user. The agent's performance depends on whether it can help the user gain the knowledge needed to specify the task well. Evaluating five frontier models, we find that no agent exceeds 56% accuracy on CoShop despite five turns of interaction. Failures stem not from agents' ability to find items, but from how little the interaction expands what users know about what they want.

We study a finite-state symbolic controller for systems in which the admissible visible transitions are fixed in advance and each visible state carries a minimum dwell requirement. The resulting model, which we call a destination-labeled self-looping system with dwell (DLSL system), records the visible graph together with local decision maps; dwell memory appears only after phase expansion. The main structural issue is that, once dwell is imposed, the current visible state no longer determines whether a departure is allowed. This leads to the converse problem: which deterministic transducers arise as phase-expanded realizations of DLSL systems over a fixed visible graph? We show that the answer is exactly the class of fiber-linear graph-respecting transducers. Under natural reachability and realizable-departure assumptions, equivalent accessible realizations over the same visible graph are isomorphic; in particular, the visible transduction determines the dwell vector and the local decision maps. We also prove that any graph-preserving deterministic realization enforcing dwell values $(d_i)$ requires exactly $\sum_i d_i$ control states. Finally, we give an $O(|Q||Ω|)$ recognition and reconstruction procedure, and extend the analysis to an edge-entry variant in which transitions may enter interior phases of successor fibers.

Large language models (LLMs) are now widely employed in software development and everyday use. Interacting with LLMs requires crafting prompts, which range from simple ad hoc sentences to extensive, detailed, and structured instructions. Knowledge about prompt engineering has been documented in several surveys and catalogs in the literature. However, the term ``prompt pattern'' is defined differently across sources, and existing works have classified prompt patterns in different ways. In this report, we present a taxonomy of prompt patterns for single-turn, text-based interactions. Following a reproducible method, we identified 30 unique and canonical prompt patterns, organized along two dimensions.

We develop a first-principles reduced-order model of these dynamics. Under a single mean-field assumption that summarizes the policy by its expected reward, we reduce the GRPO update to a stochastically-forced damped oscillator whose mass, damping, and stiffness are fixed in closed form by the optimizer hyperparameters together with a single measured curvature scale -- momentum supplies the inertia, off-policy lag erodes the damping, and the group size enters, to leading order, as a noise temperature. The reduction has three consequences. First, it subsumes the empirical single-exponential saturation law as its overdamped limit, recasting the fitted plateau, timescale, and size exponent as the fixed point, inverse stiffness, and curvature-scaling exponent of the underlying potential, and adding, through the retained inertial term, the slow-start phase the single exponential cannot represent. Second, it yields predictions tied to independently measurable quantities rather than fitted ones: group-size invariance of the deterministic trajectory with a $1/G$ stationary fluctuation, a sharp stability threshold in the refresh interval, and an overdamped-to-oscillatory transition. Third, it furnishes diagnostics that separate failure modes a reward curve alone conflates -- reward hacking, advantage degeneracy, policy concentration, and dynamical instability. Across three models and two group sizes, the closed-form trajectory fits training reward to $R^2 \geq 0.91$ and the mean trajectory is group-size invariant to leading order -- on both the reward curve and out-of-distribution transfer to eight math benchmarks -- while the within-group reward spread retains a residual $G$-dependence that the leading-order temperature picture does not capture.

Coding agents are rapidly becoming a major application of agentic LLMs, but serving them efficiently remains challenging. Progress on this challenge requires understanding real workload patterns, yet the data needed for such analysis is largely absent. Existing public traces and benchmarks do not capture real, day-to-day coding-agent usage across multiple agents and model families for serving-system analysis. To help fill this gap, we collect and release a trace of roughly 4,300 coding-agent sessions, containing about 350,000 LLM steps and 430,000 tool calls from our own day-to-day use of Claude Code and Codex. Our analysis shows that coding-agent workloads feature long autonomous loops, long contexts with short outputs, diverse and heavily-tailed tool calls, and high but imperfect prefix cache hit rates. These findings point to concrete opportunities for optimizing serving, including lower-overhead tool calling, append-length-aware prefill, semantic-aware tool-latency prediction, and improved KV-cache management around human-paced gaps. We release the dataset, trace collection pipeline, and analysis code at https://github.com/uw-syfi/TraceLab.git the project website is https://tracelab.cs.washington.edu.

The development of on-device large language models (LLMs) is driven by the need for privacy and fast response times. Energy-intensive data transfer on mobile devices makes Processing-in-Memory (PIM) an effective solution. Due to stringent DRAM cost constraints, limited physical footprint on circuit boards, and the interaction between applications and LLMs, it is imperative for the CPU and PIM to operate concurrently within a shared memory space. However, challenges such as bank conflicts and bus congestion can arise, potentially diminishing the performance and energy benefits of PIM. To address this challenge, we introduce COSM, a cooperative scheduling framework designed to facilitate the concurrent operation of PIM and CPU tasks on mobile platforms. Our key innovations include: 1) a low-interference PIM control interface that generates the maximum number of PIM commands without disrupting CPU memory accesses; 2) an idleness-aware scheduling method that integrates PIM commands into available idle time windows within the CPU's access sequence. COSM not only hides PIM execution latency from the CPU, but also overlaps PIM execution with data transfer. Experiments on concurrent execution of LLMs and mobile workloads, including mobile applications and compute-intensive kernels, demonstrate that COSM improves PIM throughput by up to 2.8x compared to the baseline scheduling method with less than 2.0% CPU performance loss.

The continuous integral R2 indicator is a Pareto-compliant refinement of the classical finite-weight-vector R2 indicator, used in performance assessment, bounded archiving for a-posteriori multi-objective optimization, and skyline selection in databases. This work introduces a bidirectional perspective mapping between continuous integral R2 computation and integration over unions of anchored axis-aligned boxes. After translating the ideal point of a minimization problem to the origin, approximation points become strictly positive loss vectors, and the subgraph of the lower weighted Tchebycheff envelope over the weight simplex maps to the complement of an anchored-box union in reciprocal objective space. The Jacobian gives an absolute R2 formula as a weighted complement volume with density $(x_1+\cdots+x_N)^{-(N+1)}$, while differences of R2 values become finite weighted hypervolume differences. Hence, hypervolume algorithms that emit box decompositions can be reused by replacing ordinary box volumes with closed-form weighted box integrals. For $N$ objectives, this gives an output-sensitive overhead $O(2^N M)$ for an $M$-box decomposition, or $O(M)$ for fixed $N$. Using existing box-decomposition approaches, the integral R2 can be computed in $O(n \log n)$ for $N=2,3$, in $O(n^2)$ for $N=4$, and in $O\left(n^{\lfloor (N-1)/2\rfloor+1}\right)$ for $N\geq4$, with $n$ denoting the size of the approximation set. On the lower-bound side, exact value computation has an $Ω(n\log n)$ lower bound in the algebraic decision-tree model already in two objectives, this bound lifts to every fixed $N\geq2$, and exact computation is $\#P$-hard when $N$ is part of the input. Together, the proposed perspective mapping provides a powerful tool for transferring algorithmic and structural results between anchored-box union and hypervolume theory and integral R2 computation.

Large Language Models (LLMs) are increasingly utilized to automate several software engineering tasks, including code completion, code summarization, testing, and the generation of repository-level documentation. While Multi-Agent Systems (MAS) are often adopted to support such tasks under the premise that task decomposition improves performance, the impact of architectural complexity on practical efficiency remains under-examined. This study empirically evaluates Retrieval-Augmented Generation (RAG) dependent architectures for the generation of README files for GitHub repositories. In this work, we conducted a systematic comparison between a Single-Agent pipeline, a specialized MAS, and a developer-guided planning (DevPlan) variant, benchmarked against LARCH -- a state-of-the-art baseline -- and the original ground truth. Results indicate a critical architectural trade-off: the Single-Agent pipeline achieves lexical quality comparable to MAS while reducing token consumption by 86% and operating at twice the speed. In contrast, manual taxonomy analysis demonstrates that MAS achieves high structural consistency (98%), resolving formatting issues observed in single-agent approaches. Autonomous planning is identified as the primary pipeline bottleneck; incorporating lightweight developer-guided plans produces the highest overall documentation quality, surpassing all the analyzed configurations.

Multi-agent LLM systems can decompose software-engineering work into planning, generation, validation, and repair, but a narrower systems problem remains: before any governed shared mutation is applied, a system must decide which concurrently formed write intents may proceed in parallel, which require deterministic composition or serialization, and which must take a fail-closed path. We address this problem with the AI-Atomic-Framework (ATM), a specification-grounded governance substrate for software agents operating within a single governance domain. ATM binds task intent, repository scope, write admission, validation, and evidence obligations into one governance chain. A Content Identifier (CID) broker serves as the shared-mutation admission subsystem. Adapter-guided atomization maps write intents to semantic atoms and bounded regions; when persistent atom-map coverage is incomplete, virtual atoms provide temporary auditable governance units for conservative comparison and routing. Governed shared writes are ultimately applied by a neutral steward rather than directly by proposing agents. Evaluation combines controlled, field, adoption, and extension evidence, including a 12-scenario deterministic design matrix, three archived runner cases, ATM-AdmissionBench, three archived same-file boundary cases, a three-week external-adopter study, and an operational recovery-routing benchmark. The results support feasibility, auditability, and bounded recoverability within the observed single-domain settings, but do not claim broad comparative superiority or cross-clone governance.

Multi-round reflection lets agents built on large language models recover from failures within a single task, but each task remains an isolated episode: lessons learned across many reflection rounds on one task are discarded before the next begins. We study this gap on a code-generation task: from a scientific paper section, the agent writes Python in the open-source Manim library to render a mathematical animation. We present ManimAgent, a self-evolving multimodal agent that carries reflection experience across tasks through a dual-channel Episodic Memory Bank grown entirely from its own task stream, with no weight updates and no human seeds. After each animation converges, a vision-language model scores the rendered keyframes; the resulting signals populate a positive channel M+ that stores success rationales as soft Reference Examples, and a negative channel M- that stores validated failure patterns as hard Known Pitfalls. On a fixed-probe evaluation against no-memory, matched-budget retrieval-augmented generation, and shuffled-memory baselines, blind human Pass@1 rises and reflection rounds fall as memory size grows. We will release the code, frozen memory snapshots, and the task stream.

Relevance is not permission. Attention lets a model read key-value items related to the current query, but it does not guarantee that the value contribution of such an item becomes prediction evidence. A retrieved passage may be relevant to a question without being supporting evidence, and a historical fact or temporal neighbor may even blur true-tail ranking or the current edge score. This paper formalizes this gap as a permission problem for the weighted value term alpha_ij * v_j that is actually added to the prediction path. We propose Warrant, a path-localized interface that preserves attention relevance alpha_ij, exposes the value path leading to the primary metric, and, in the full model, turns alpha_ij * v_j into alpha_ij * g_ij * v_j through learned query-item permission g_ij. We place the same operator on the metric-defining value paths of CTDG link prediction, MTPP next-mark ranking, RAG supporting evidence selection, STPP next-location forecasting, and TKG tail prediction. Across 32 paired comparisons, 3 seeds, and 192 total runs, Warrant improves the primary metric in 27 comparisons; practical tiers consist of 10 substantial effects, 1 marginal effect, 8 positive but uncertain effects, 8 tie/negligible effects, and 5 drops. In the path-localization check, correct-path placement outperforms direction-aware Base performance in every domain and exceeds generic attention placement by +0.1076 AUC in CTDG and +0.0683 MRR in TKG. Ablations show that most TKG gains come from historical-tail value path exposure, whereas the core CTDG gain comes from edge-conditioned query-item permission. In conclusion, prediction evidence is not attention mass. A weighted value term becomes evidence only when it is warranted on the path to the metric.

The emergence of Software-Defined Vehicles represents a fundamental shift in automotive design, prioritizing software-centric architectures over traditional hardware-driven models. SDVs require modularity, interoperability, real-time processing, and over-the-air update capabilities throughout the vehicle lifecycle. However, current vehicle systems, characterized by tightly coupled software and hardware, struggle to meet these demands due to their complexity and heterogeneity. A critical first step toward enabling SDVs is the decoupling of software from hardware, which can be facilitated through a robust Hardware Abstraction Layer. While existing HALs offer hardware independence and standardized interfaces, their applicability and effectiveness in SDV contexts remain uncertain. This paper systematically evaluates current automotive HALs and explores HAL mechanisms from non-automotive domains, including smartphones, networking, and industrial automation, to extract cross-domain insights relevant to SDV development. A criteria-driven evaluation framework is developed to assess HALs against SDV-specific needs. Findings reveal that while middleware-based HALs offer portability and modularity, hypervisor-based approaches better support safety, OTA readiness, and hardware efficiency. Limitations in both approaches are identified, prompting recommendations for a hybrid HAL design that integrates hypervisor isolation with middleware standardization. This paper contributes to the ongoing developments in automotive software architecture by offering insights into the applicability and effectiveness of current HAL strategies. It provides actionable guidance for designing flexible, scalable, and future-ready HALs to support SDVs across their lifecycle.

Vision-Language Models (VLMs) provide powerful semantic understanding and commonsense reasoning for End-to-End Autonomous Driving (E2E-AD) planning. However, trajectories directly generated by VLMs often encode only coarse driving intentions and remain insufficient for geometrically accurate, future-aware, and multi-view-grounded planning. To address these limitations, we develop the Layer-Wise World-Model-Guided Driving framework (LWDrive). LWDrive is a VLM planning framework that refines coarse trajectories through layer-wise world-model guidance. Instead of treating the VLM output as the final trajectory, LWDrive uses it as an intent-aware coarse plan, expands a diverse candidate space around it, and progressively refines the candidates through a Foresight Cascade Planner (FCP). Specifically, we introduce future-frame generation supervision to encourage the VLM to learn forward-looking scene representations, thereby injecting planning-relevant predictive dynamics into its internal hidden states. Built upon these world-model-supervised representations, FCP exploits VLM features across multiple layers and integrates historical temporal states, Action-Query representations, and current-frame multi-view Bird's-Eye-View (BEV) features to refine candidate trajectories in a coarse-to-fine manner. This design enables progressive correction of spatial positions and motion trends while grounding trajectory refinement with multi-view scene cues and preserving the high-level driving intention produced by the large model. Finally, a score head evaluates the refined candidates and selects the best trajectory as the final planning output. Experiments show that LWDrive achieves a score of 92.0 on the NAVSIM benchmark and 89.6 on NAVSIM-v2. Code and models will be made publicly available.

Knowledge graphs (KGs) organize real-world knowledge as triplets and underpin many downstream applications. Due to their inherent incompleteness, knowledge graph completion (KGC) is widely studied and is typically formulated as triplet prediction, with link prediction as the dominant paradigm. However, this formulation focuses on the incompleteness of triplet-wise information and overlooks the incompleteness of entity-relation compatibility information. To address this limitation, we introduce a relation set completion task (RSC), which complements the link prediction task and aims to reason about missing relations that are semantically compatible with a given entity. We further propose a Relation Set Embedding model (RelSetE), which models latent patterns among the observed relations of entities to infer missing ones. To evaluate RelSetE, we derive three benchmark datasets from standard KG benchmarks. Extensive experiments demonstrate that RelSetE effectively captures entity-relation compatibility patterns and performs favorably in inferring missing relations of entities. Code and data are publicly available.

Multimodal model inference creates substantial energy demand with growing performance requirements. Within GPUs, power is autonomously managed by an on-board power management unit (PMU), which makes frequency boosting/throttling decisions. However, we find that these hardware-managed frequency decisions can cause significant power inefficiency. This work identifies three classes of power inefficiencies within modern multimodal inference serving: (1) inter-stage dependency stalls run at near maximum frequency despite being idle; (2) anti-correlation between auto-boost frequency and arithmetic intensity (A.I.) results in compute-bound phases (e.g., prefill) running at lower frequency and vice versa; and (3) thermal throttling degrades SM frequency and throughput. We propose Tri-serve, a software-based DVFS controller that jointly accounts for three classes of inefficiency -- inter-stage Dependency stalls, the Arithmetic-intensity effect on frequency and power, and the Thermal-throttling effect of high A.I. phases -- to deliver energy-efficient multimodal serving on commodity GPUs. We show that Tri-serve achieves 22% energy efficiency improvement with no latency or throughput impacts.

In mid-2026 a slogan reorganized how practitioners talk about coding agents: stop prompting your agent, start designing the loop that prompts it. We take this claim seriously and give it a careful treatment. We call the object of the new practice the loop specification: a bounded, reusable artifact, made of a trigger, a goal, a verification step, a stopping rule and a memory, that a human hands to an agent harness (such as Claude Code or Codex) so the agent pursues a goal on its own, in place of step-by-step prompting. We distinguish this external loop specification from two things it is often confused with: an ordinary programming loop, and the internal perceive-act-observe cycle that the harness already provides as plumbing. We position loop engineering as a new layer in the progression from prompt to context to harness to loop, and we argue, against the stronger headlines, that it does not retire prompt engineering; loop and prompt are distinct tools with distinct uses. We offer four contributions: a definition and scope for the discipline; an anatomy and taxonomy of loop specifications organized around trigger, goal type, a five-level verification ladder, architecture, and named terminal states; a descriptive analysis of the Loop Library, a public corpus of fifty real loops that we code by hand; and a set of design principles and anti-patterns grounded in the scientific literature on self-correction, reward hacking and model-as-judge fragility. The corpus shows that practice has matured most where the discipline says it matters: seventy percent of loops verify in the autonomous zone of the ladder and seventy-four percent name their terminal states, while automated triggering and durable memory remain comparatively underdeveloped. We close with the limits the practice must respect, including the verification burden, comprehension debt and the risk of cognitive surrender.

Despite the impressive capabilities of Large Vision-Language Models (LVLMs), they remain susceptible to hallucination, generating content inconsistent with the input image. Recent studies attribute this to the dominance of language priors over visual inputs and employ contrastive decoding methods to mitigate this dominance, but the mechanistic origin remains unexplored. We investigate the information flow through each transformer layer and find that attention modules consistently aggregate visual evidence, while FFN modules at critical layers act as the source of language priors. These priors can override visual evidence, causing correct predictions in intermediate layers to drift toward incorrect outputs. Based on this insight, we propose FADE (FFN Attenuation for DEcoding), a training-free method that attenuates FFN outputs to reduce language-prior dominance. Evaluations on POPE, CHAIR, and MME benchmarks across LLaVA-1.5, mPLUG-Owl2, and InstructBLIP show that FADE effectively mitigates hallucinations while preserving inference efficiency.

Despite the capability of parallel decoding, diffusion large language models (dLLMs) require many denoising steps to maintain generation quality, motivating recent research on efficient decoding strategies. However, existing studies have reported inconsistent evaluation results even under seemingly identical evaluation settings, risking biased conclusions about dLLM decoding methods. To understand this evaluation concern, we conduct a rigorous evaluation of current decoding methods for dLLMs across diverse evaluation settings. Surprisingly, our analysis reveals that the ranking of decoding methods is highly sensitive to the choice of prompt templates. Single-template evaluation can lead to an illusion that decoding methods improve inference efficiency without performance degradation. Through comprehensive experiments, we find that current parallel decoding methods consistently underperform the single-token decoding baseline, failing to overcome the speed-quality trade-off. We further identify this evaluation inconsistency as the high sensitivity of parallel decoding methods to minor variations in prompt templates. Our experiments show that an effective prompt template can achieve strong evaluation results even with fewer denoising steps, markedly outperforming the marginal gain from increasing denoising steps. Beyond prompt templates, our experiments indicate that overlooked evaluation settings can also notably affect the assessment of decoding methods. Based on these findings, we propose practical guidelines for the reliable evaluation of decoding methods in dLLMs.

Cooperative multi-agent reinforcement learning (MARL) often relies on communication to mitigate partial observability, yet most existing protocols treat messages as flat dense vectors detached from the structure of the observations they summarize. This design overlooks an important source of inductive bias in many cooperative environments, where observations naturally follow a hierarchy such as groups and entities. We propose \textsc{HiComm}, a plug-in communication module that grounds messages in the sender's hierarchical observation. \textsc{HiComm} is receiver-driven: the receiver issues a query, and the hierarchy is resolved through a three-stage decoding process that first selects a group, then a sender, and then an entity within that group, returning the corresponding feature slice as the message. This converts communication from unstructured vector transmission into structured information retrieval over the sender's observation hierarchy. We instantiate this mechanism with Straight-Through Gumbel-Softmax for differentiable discrete selection and a lightweight shared projection design that attaches to standard MARL pipelines. Experiments across cooperative MARL tasks with different observation structures and coordination demands show that \textsc{HiComm} matches or outperforms representative learned communication baselines while reducing communication volume by up to $23\times$ per receiver per episode.

There are various benchmarks to evaluate bugfixing capabilities of Large Language Models. However, most widespread benchmarks do not fully reflect real-world bugfixing practices. They are small, weakening statistical reliability, and the buggy programs are often similar to one another, potentially distorting evaluation results. The range of bug types can also be narrow, failing to capture a representative range of bugs. To address these issues, we introduce MegaBugFix, a large-scale bugfixing benchmark containing 12,629 buggy Python programs synthesized from correct ones by a Large Language Model. Bug injections were generated as diffs representing code changes. Through this approach, we were able to avoid common pitfalls of LLM-based mutation techniques like injecting overly simplistic bugs or failing to modify the input program. We evaluated 13 open-weight models on MegaBugFix and baseline benchmarks, finding consistently lower performance on MegaBugFix. This reveals that our benchmark presents more challenging bugs and exposes model failures that may remain hidden when evaluating on existing benchmarks. The benchmark and fine-tuned model used for bug injection are available at hf.co/collections/szalontaib/megabugfix.

While imaginary-time evolution has long served as a standard paradigm for ground-state preparation in numerical simulations and quantum devices, its intrinsic dynamical properties has been largely overlooked. Here, we investigate the short-imaginary-time critical dynamics in quantum systems with boundaries. A universal scaling theory is developed and verified in the two-dimensional quantum Ising model, uncovering rich dynamic critical behaviors dictated by boundary universality classes. For ordered initial states, the boundary order parameter $M_s$ decays with imaginary time $τ$ as $M_s \propto τ^{-β_1/νz}$, where $β_1$ denotes the boundary order parameter exponent, and $ν$ and $z$ correspond to the correlation length exponent and the dynamic exponent, respectively. For disordered initial states, the autocorrelation of the boundary order parameter is governed by a novel critical exponent $θ_1$, which is closely related to the critical initial slip behavior of $M_s$ characterized by the corresponding exponent $θ_1'$. In contrast to its positive bulk counterpart, the boundary initial-slip exponent $θ_1'$ is negative for the ordinary transition while remaining positive for the special transition. Although the static universality classes of $d$-dimensional quantum phase transitions generally coincide with those of $(d+1)$-dimensional classical phase transitions, we show that $θ_1$ does not follow this conventional quantum-classical mapping. We further discuss the implications of our results for more exotic forms of boundary criticality. Our findings provide new physical insights into boundary critical dynamics and offer a novel route for probing exotic boundary critical behaviors in quantum many-body systems.

Moiré superlattices host narrow minibands whose bandwidth governs correlated and topological phases. Here, we demonstrate that the bandwidth also sets the critical current for the onset of out-of-equilibrium transport. In bilayer graphene aligned to hexagonal boron nitride, we explore the high-current transport regime as we continuously flatten the valence miniband using an out-of-plane displacement field. We observe a significant reduction in the critical current, which is captured by a minimal analytical model and corresponds to the calculated narrowing of the miniband. Moreover, by comparing distinct moiré platforms, we show that the scaling between critical current and bandwidth is a universal feature of graphene superlattices. Our results reveal a direct link between miniband dispersion and high-current transport, and establish this regime as a fast and accessible electrical probe of bandwidth evolution.

The breathing mode in Kagome materials is a structural modulation that breaks inversion symmetry and has been shown to be a crucial source for intriguing phases in the normal state. In this work, we carry out a full classification of superconducting symmetries in kagome superconductors and demonstrate the emergence of odd-frequency dynamical Cooper pairs entirely driven by the breathing mode. We then show that odd-frequency spin-singlet Cooper pairs can be realized by controlling the breathing mode in kagome lattices with conventional spin-singlet $s$-wave superconductivity. Since odd-frequency pairing is intrinsically nonlocal in time, our results put forward the breathing mode for designing dynamical Cooper pairs in kagome materials.

Portfolio diversification - a cornerstone of modern investment management - can be formulated as a Maximum Independent Set (MIS) problem on asset correlation graphs. Solving this problem at scale is computationally challenging, motivating the exploration of quantum algorithms for practical financial optimization. We propose an end-to-end pipeline leveraging qReduMIS, a recursive hybrid quantum-classical algorithm. Rather than using quantum optimization to directly produce a final solution, qReduMIS leverages independent set measurements from the Quantum Approximate Optimization Algorithm (QAOA) to identify frozen nodes - vertices likely to belong to optimal solutions - thereby guiding and unblocking subsequent (provably optimal) classical reductions on the remaining graph. We benchmark qReduMIS on real financial data from four major market indices with up to 225 assets, executing experiments on Quantinuum's 98-qubit trapped-ion Helios system, with QAOA circuits acting on kernels of up to 78 qubits and 1016 two-qubit gates. While standalone QAOA fails to find the optimal solution for two of the largest indices (S&P 100 and Nikkei 225), qReduMIS achieves success probabilities of $0.40$ and $0.95$, respectively, with average approximation ratios $\geq 0.96$ across all four indices. We perform a systematic benchmark on the Quantinuum H2-1 noisy emulator over 73 asset correlation graphs of varying size showing that, for $p=2$ QAOA layers, the optimal time-to-solution scaling exponent of qReduMIS is $3.2$ times smaller than that of standalone QAOA.

Kinetic uncertainty relations bound current precision of stochastic processes by dynamical activity. The extension of these bounds to quantum systems has been impeded by coherence, strong system-reservoir coupling, and the subtlety of defining dynamical activity in the quantum regime. Here, we introduce a partial dynamical activity through the quantum Fisher information associated with the rescaling of the system-reservoir coupling and show that it bounds current precision via a universal susceptibility-kinetic uncertainty relation. The general validity of this relation for any open quantum system is guaranteed by the natural contribution of a susceptibility term, which is experimentally accessible by tuning the system-reservoir coupling strength. We show how the partial dynamical activity encompasses previous definitions of activity in the weak-coupling Markovian limit and that it provides an information-geometric interpretation of correlator-based activities. We illustrate the tight constraint on precision that our bound provides with the example of steady-state transport through a double quantum dot, where quantum effects invalidate previously developed kinetic uncertainty relations. We expect our bound to provide a powerful tool for optimizing precision in arbitrary quantum systems.

Chemical gradients are ubiquitous in porous and crowded environments, including soils, filters, fabrics, tissues, hydrogels, biofilms and living cells. They arise from displacement fronts, dissolution and precipitation, ion exchange, metabolism, root exudation, evaporation, gas dissolution, freeze--thaw cycles and externally imposed chemical treatments. These gradients can drive colloids, macromolecules and emulsion droplets by diffusiophoresis, while simultaneously driving diffusioosmotic flows along confining surfaces. Classical models of colloid transport in porous media emphasize hydrodynamic dispersion, surface interactions, straining, deposition, detachment and filtration. This chapter places diffusiophoresis within that broader transport framework and reviews how porous media generate, stretch, disperse and sustain the solute gradients that drive phoretic motion. We first discuss sources of chemical gradients and the distinction between spreading and mixing, then summarize classical colloid transport, the minimal physicochemical model for diffusiophoresis and diffusioosmosis, and the experimental platforms used to study these effects. Particular emphasis is placed on recent results showing that diffuse solute fronts can enhance phoretic removal from dead-end pores by prolonging the duration of forcing, and that cross-streamline migration within flowing pathways can change macroscopic breakthrough and dispersion by orders of magnitude. We close by discussing emulsion droplets, multiphase flows, confined and living media, and open problems, including the transition from algebraic mixing in two-dimensional micromodels to chaotic mixing in three-dimensional porous media.

A double-occupancy cell model of a fluid with Curie-Weiss interaction is studied. First, we show that the model is isomorphic to the Blume-Capel model on a complete graph through a simple transformation from spin to occupancy variables. We then investigate its phase behavior within the grand-canonical ensemble using a combination of analytical and numerical methods. Despite its simplicity, the model exhibits a remarkably rich thermodynamic behavior depending on the ratio between the local repulsive and global attractive interactions. We identify regimes characterized by a single critical point, two distinct critical points, tricritical behavior, and triple-point formation. For sufficiently strong repulsion, the system possesses three fluid phases of different densities, leading to both gas-liquid and liquid-liquid coexistence. The locations of the critical, tricritical, and triple points are determined, and the corresponding phase diagrams are constructed. These results demonstrate that the competition between double-occupancy repulsion and long-range attraction is sufficient to generate complex phase behavior in a minimal multiple-occupancy lattice-gas model.

Understanding how strong correlations and dissipation combine to shape collective quantum excitations is a central challenge in many-body physics. We investigate the effect of dissipative light-matter coupling on strongly interacting exciton-polaritons in the presence of a biexciton resonance, which gives rise to polaron-polariton quasiparticles. We show that the interplay between many-body correlations and non-Hermitian coupling generates anomalous dispersion relations and exceptional points in the polaron-polariton spectrum. The location and coexistence of exceptional points are controlled by the dissipative coupling and the relative decay rates of the excitonic and photonic constituents, allowing them to emerge across different polaron-polariton branches. These results identify dissipative polaron-polaritons as a versatile platform for exploring non-Hermitian many-body physics with tunable light-matter quasiparticles.

Altermagnets, distinguished by compensated antiparallel spins yet nondegenerate magnon spectra, bridge the gap between ferromagnets and antiferromagnets. Although several probes such as anisotropic transport and spectroscopic measurements have been proposed to identify altermagnetic order, experimentally accessible transport signatures remain highly desirable. Here, we show that in altermagnets subject to an external magnetic field, the spin Seebeck effect exhibits pronounced directional anisotropy. Specifically, the spin Seebeck coefficient differs significantly along in-plane directions, and this anisotropy increases with field strength. Magnetoelastic interactions further produce resonant peaks whose positions with respect to an applied magnetic field reflect the intrinsic magnon band anisotropy, and provide localized features that enhance the distinguishability of the altermagnetic response. The peaks provide a robust experimental signature of altermagnetic order. Our findings serve as signatures of altermagnetic order while laying the groundwork for their application in spintronic devices.

We use self-consistent field theory to investigate the free energy landscape for single-chain expulsion from a diblock copolymer micelle with a dense corona. Using the distance from the micelle center-of-mass to the hydrophilic-hydrophobic junction of the chain as the reaction coordinate, we compute the free energy landscape for chain exchange. Our results show that the expulsion free energy barrier scales linearly with both the hydrophobic block length and the solvent selectivity, consistent with recent experiments. To accurately resolve chain conformation, we introduce a second reaction coordinate: the distance between the junction and the free end of the hydrophobic block, and construct a two-dimensional free energy surface. Using the string method to identify the minimum energy path, we find that all pathways converge to a nearly degenerate reaction channel, irrespective of the initial path. Within this channel, the end-to-end distance of the hydrophobic block exhibits a broad distribution, yet the corresponding expulsion barriers remain nearly indistinguishable. Together, these findings establish a continuum-level theoretical foundation for understanding the hyperstretching mechanism and the transition state ensemble in micellar chain exchange.

Understanding quantum materials at meso and even macroscopic scales requires tight-binding calculations on system sizes where explicit matrix representations become prohibitively costly. This represents a major bottleneck to rationalize phenomena in moiré and super-moiré heterostructures and quasicrystals. Here, we present a unified tensor-network methodology to solve tight-binding problems at exceptionally large scales, by mapping a system of $N = 2^L$ sites onto a many-body problem of $L$ pseudospin sites, which is subsequently solved with tensor network algorithms. For Hamiltonians with compressible real-space structure, the tensor network bond dimension remains modest, typically of order a few tens, independent of $N$.Tensor network representations of arbitrary hopping functions including long-range, spatially modulated, and twisted-layer couplings are built with quantics tensor cross interpolation, and all physical observables are evaluated entirely with tensor network algebra without explicit matrix storage or diagonalization. We demonstrate applications to spectral functions, momentum-space spectra via the tensor-network quantum Fourier transform, real-space topological invariants, real-time dynamics, correlation induced symmetry breaking with self-consistent mean-field calculations, non-Hermitian phenomena, and excitonic many-body physics. Our methodology enables routinely solving systems with billions of sites, by leveraging the tensor network compressibility of real-space structures, and establishing a flexible framework to study quantum matter at ultra-large length scales. The methodology is implemented in the open-source Julia package TensorBinding.jl.

Using the magnetostatic approximation, we calculate the spectra of bulk and surface spin waves in an in-plane magnetized ferromagnet/paramagnet bilayer. Due to the dipolar coupling between the layers, the paramagnet becomes polarized, which in turn modifies the spectrum of the Damon-Eshbach magnetostatic modes. We assume that the paramagnet is characterized by a magnetic susceptibility, $χ\propto 1/(T-T_C)$, which reaches large values when the system temperature $T$ is close to the Curie temperature $T_C$. We find the conditions under which surface spin waves become unidirectional, i.e., capable of carrying energy in only one direction, and determine the magnitude of their frequency nonreciprocity. We demonstrate the possibility of switching the unidirectional wave regime on and off by varying the external magnetic field or temperature, making the ferromagnet/paramagnet system an attractive platform for tunable magnonic logic devices.

We present a symmetry-based framework to classify the non-reciprocal responses of multiterminal ring quantum devices. The device is modeled as a ring of $n$ vertices, where a binary variable $e_k\in\{+1,-1\}$ on each bond encodes the preferred direction of signal flow between terminals. Non-reciprocity corresponds to a preferred current configuration on the ring, and the symmetry group of the device partitions all $2^n$ configurations into equivalence classes(orbits) characterized by a topological winding number $W$. Using the minimal non-trivial case $n=3$, we establish two results independent of microscopic details. First, lifting the degeneracy within an orbit generates non-reciprocal responses. For $n=3$ this requires simultaneous breaking of both time-reversal $T$ and spatial inversion $I$. Breaking either alone is insufficient. Second, the residual geometry symmetry after $T$ and $I$ are broken determines which responses are observable. For an isosceles triangular geometry, only two types of response are allowed: uniform circulation (all bonds carrying current in the same direction) and semi-circulation with the reversed bond on the geometrically distinct base. Semi-circulation with the reversed bond on either equal leg is symmetry-forbidden. Both predictions are validated using a minimal toy model of three quantum dots coupled to superconducting baths, which demonstrates a reactive quantum circulator response.

Nanoscale inhomogeneities are a defining feature of many superconducting materials, yet their local electromagnetic response has remained difficult to access experimentally. This is because their relevant energy scale lies in the terahertz range, where wavelengths -- on the order of hundreds of microns -- are too large to spatially resolve nanoscopic variations. Here, we demonstrate the first application of THz near-field photovoltage nanoscopy in a superconductor, achieving 300 nm spatial resolution at 2.52 THz. Scanning a current-biased NbN strip, we reveal photovoltage peaks within the bulk associated with nanoscopic defects of reduced superfluid density. The observed photovoltage follows the evolution of the vortex-dissipative state and is attributed to enhanced vortex-antivortex pair nucleation at defect sites. Together, these results open a direct route to probing how material inhomogeneities influence light-matter interactions in superconductors, with implications for superconducting devices and strongly inhomogeneous systems such as high-Tc and moiré materials.

Current research on the disordered topological quantum phases primarily focuses on the uncorrelated and short-range correlated disorder regime. These topological Anderson insulators are typically self-averaging. However, topological quantum systems with long-range correlated disorder have received limited attention due to the absence of tractable analytical methods. In fact, the long-range correlated disorder introduces more complex effects on topological quantum states. Here, we demonstrate that the long-range correlated disorder could induce anomalously statistical feature where the topological Anderson states become non-self-averaging, and the phase diagram is strongly dependent on individual disorder configurations. We term this statistical phase the non-self-averaging topological Anderson insulator. The non-self-averaging property is identified by the non-vanishing finite values of the relative variance of the Lyapunov exponent in the thermodynamic limit, alongside the non-Gaussian distributions of the Lyapunov exponent. Consequently, the topological properties of a single disordered sample deviate from the ensemble average, causing a breakdown of the central limit theorem. The non-self-averaging topological Anderson insulator provides insights into the interplay between correlated disorder and topology.

We investigate the role of compressibility in the modified quasi-linear viscoelastic (MQLV) constitutive framework for soft solids at finite strain, where shear and bulk responses are governed by distinct relaxation functions. Analytical and semi-analytical results are derived for simple shear and torsion, under incompressible and slightly compressible assumptions. We show that compressibility affects the response only when volume changes occur: under isochoric deformations, the bulk contribution vanishes, while even small deviations from isochoricity significantly alter the normal response. Shear stress and torque are largely insensitive to compressibility, whereas normal stress and axial force exhibit pronounced sensitivity due to the coupling between shear and bulk relaxation. We further demonstrate that volumetric effects interact with the Poynting effect: in simple shear they oppose each other, reducing relaxation, while in torsion they reinforce each other, enhancing it. These trends agree with brain tissue experiments but reveal limitations of the slightly compressible model for highly compressible materials, such as agarose gels. Overall, the results emphasise the importance of accounting for compressibility in modelling normal stress responses and motivate the development of fully compressible formulations and numerical implementations.

Controlling magnetism by electric field or current is a central topic in spintronics. In this work, we argue that the magnon spin photocurrent can also be controlled by the electric field in magnetoelectrics. Taking Cr$_2$O$_3$ as an example, we demonstrate how the spin current is modified by the electric field, using nonlinear response theory. We find that the Dzyaloshinsky--Moriya interaction induced by the applied field plays a key role in modifying spin-current conductivity, which exhibits pronounced anisotropy with respect to the light polarization. In particular, both the resonance frequency and the peak intensity show distinct dependences on the external electric field $E$, demonstrating electrical control of the spin photocurrent. In addition, we show that the two-magnon processes give rise to a continuum spectrum, a consequence of the field-induced spin canting. These results show that Cr$_2$O$_3$ is a promising platform for realizing electrically tunable spin photovoltaic effect.

The observation of the gate-controlled supercurrent (GCS) effect in superconducting nanostructures initiated major research efforts toward the realisation of superconducting-based computing architectures. Here we introduce a universal logic circuit that can be a promising superconducting building block of classical hybrid supercomputers. We demonstrate a functionally complete set of logic gates by realising the AND, OR, NOT and COPY gates. The general layout and scalability of our device, combined with recent experiments demonstrating fast switching and small voltage signals, make it a functional candidate in superconducting electronics. Our device enables the realization of all classical logic gates and the half-adder combinational logic circuit using at most three nanowires, each uniquely configured with two side-gate electrodes.

Ruggedness of the underlying energy landscape gives rise to heterogeneous mobility and non-Gaussian diffusion. We develop a theoretical framework for tagged-particle diffusion in multidimensional rugged energy landscapes modeled as correlated quenched Gaussian random fields. Using the self-propagator and self-dynamic structure factor, we characterize finite-time diffusion beyond the effective diffusion coefficient. We determine the effects of dimensionality, spatial correlations, and initial preparation. By introducing a coarse-grained mobility field and a mobility-memory approximation, we relate the non-Gaussian parameter to the time correlation of the mobility sampled by the particle. In the homogenized diffusive regime, the mobility correlation decays algebraically, leading to long-time relaxation of the non-Gaussian parameter as $t^{-1/2}$ in one dimension, $(\ln t)/t$ in two dimensions, and $t^{-1}$ for $d>2$, with amplitudes that depend on dimensionality and the initial ensemble. Our results show that rugged energy landscapes leave distinct signatures in the effective diffusion coefficient, self-dynamic structure factor, and relaxation of non-Gaussian fluctuations.

For many decades, researchers have been studying various types of electro-hydrodynamic instabilities in liquid crystals. A significant amount of experimental data has been collected, however, the theoretical interpretations of the results typically rely on linear analysis. In response to this limitation, we investigate the nonlinear stage of the flexoelectric instability in nematics, focusing on liquid crystals with a negative anisotropy in their dielectric permittivity and electrical conductivity. We base our analysis on a comprehensive set of nonlinear electro-hydrodynamic equations for these nematics influenced by an external alternating electric field. The equations predict an instability that is driven by the flexoelectric effect. In order to examine the peculiarities of this phenomenon, we use a model that was proposed in our previous publications, Refs. [1,2], which allows us to perform numerical simulation of nonlinear dynamics. We examine patterns that are formed above the instability threshold. Through numerical simulations, we have identified static and dynamic patterns that occur over a timescale that is much longer than the period of the external electric field. The static patterns are one-dimensional structures and dynamic patterns are standing or traveling one-dimensional waves. The type of the realized pattern depends on the material and experimentally controlled parameters. We found that the standing waves are stable with respect to small transverse perturbations, whereas the propagating waves are unstable. We present a Ginzburg-Landau-like phenomenology that applies near the instability threshold. This approach allows us to rationalize our numerical findings with a few parameters.

State-of-the-art organic solar cells increasingly rely on low-offset semiconductor blends, challenging the traditional requirement of a large energetic driving force for efficient charge separation. In these systems, the energetic offset $ΔE_{\mathrm{LE-CT}}$ between local exciton (LE) and charge-transfer (CT) states approaches the thermal energy, making exciton-CT hybridization and thermal repopulation of the exciton level critical to device performance. In this work, we directly compare a macroscopic two-state rate model with three-dimensional kinetic Monte-Carlo (kMC) simulations to investigate microscopic charge separation dynamics and the role of Gaussian energetic disorder. We demonstrate that in the absence of disorder, the analytical rate model accurately reproduces kMC predictions for the whole range of $ΔE_{\mathrm{LE-CT}}$. Specifically, the macroscopic model successfully explains horizontal shifts in the internal quantum efficiency curves that arise depending on how the energetic offset is physically realized in the constituent molecules. We show that these variations can be captured entirely through the ratio of degeneracies of the LE and CT states, respectively. Introducing Gaussian energetic disorder into the kMC simulation reveals a distinct crossover behavior depending on $ΔE_{\mathrm{LE-CT}}$. While disorder is mostly detrimental at large offsets, it can significantly boost efficiency at intermediate and low offsets. Thermalization of charge carriers within the disorder-broadened density of states creates an effective driving force allowing charge separation even at zero or negative energetic offsets.

The exceptionally high in-plane refractive index, low sub-bandgap absorption, and strong optical anisotropy of WS2 make it a promising material platform for next-generation integrated circuits for nanophotonics. Its layered van der Waals structure further enables heterogeneous integration with silicon photonics and emerging two-dimensional optoelectronic materials. However, despite increasing interest in the waveguiding properties of WS2, experimental studies of wavelength-dependent modal confinement and attenuation remain limited. Additionally, though the extinction coefficient of WS2 is expected to be near-negligible beneath the bandgap, reported values span orders of magnitude, leading to large uncertainty in predicted modal decay lengths and wafer-scale integration feasibility. To resolve these ambiguities we perform hyperspectral cavity-enhanced imaging, determining high-resolution upper and lower bounds on the extinction coefficient of WS2 within the visible-NIR edge. We further employ scattering-type scanning near-field optical microscopy (s-SNOM) to probe TE0, TM0, and higher-order modes in both quasi-bulk and nanobeam WS2 waveguides across the 800-1400 nm spectral range, enabling identification of mode-specific trends in wavevector dispersion and loss. This work simultaneously assesses s-SNOM as a probe of waveguide performance, and we find that while absolute loss values depend on measurement geometry, s-SNOM reliably captures relative modal trends and provides upper bounds on propagation loss, supporting its use as a diagnostic tool for anisotropic waveguides. We further identify significant artefacts in nanobeam measurements arising from transverse interference and spatial sampling effects when the structure size approaches the excitation wavelength, which can shift extracted effective indices by up to 0.25.

Topological superconductivity is widely pursued by inducing superconducting correlations in topologically protected boundary states. In two dimensions, this strategy has been realized using one-dimensional topological edge modes, but in three-dimensional crystals, spatially separated surface supercurrents confined to selected facets have not yet been achieved. Here we demonstrate facet-selective ballistic supercurrent in Josephson junctions based on the weak topological insulator ZrTe₅. Superconducting quantum interferometry reveals SQUID-like critical current oscillations with flux-quantum periodicity, establishing that the supercurrent is spatially concentrated on specific crystallographic facets that host gapless topological surface states. Rotating the magnetic field yields markedly distinct interference patterns, linking the supercurrent distribution to the underlying bulk topology. The exponential temperature dependence of the critical current and triangular interference lobes provide signatures of ballistic transport due to high-transmission topological channels. These results establish weak topological insulators as a platform for facet-resolved superconducting devices and higher-order topological superconductivity.

The chirality-induced spin selectivity (CISS) effect has been invoked to explain recent reports of differences in the time-resolved EPR signals between chiral and achiral molecules. However, the microscopic origin of these differences and their connection to CISS remains contested, particularly since these systems lack a metal interface. Here we introduce an intramolecular spinterface-like mechanism that naturally arises within donor-chiral bridge-acceptor (D--$χ$B--A) complexes and quantitatively reproduces experimentally reported observed spin polarization in time-resolved EPR studies. In our two-electron Lindblad model, the photoexcited charge-transfer electron traversing the chiral bridge exchanges with the residual donor electron, which acts as a localized magnetic moment analogous to an induced magnetic moment on an electrode surface. The resulting through-bridge charge current produces an effective solenoidal field at the donor--bridge interface, breaking spin degeneracy and directional symmetry, thus enabling spin-selective transport without invoking intrinsic spin-orbit coupling on the bridge. We show that the interplay between this current-induced field, donor thermalization (which breaks time-reversal symmetry), and bridge spin mixing yields tens-of-percent polarization over realistic experimental conditions and charge-transfer time scales, matching reported CISS signatures in triads and DNA hairpins. By explicitly resolving the dependence on solenoidal coupling strength, temperature, and spin-mixing rates, the model identifies the regime in which internal spinterfaces can generate robust CISS-like spin filtering. These findings demonstrate that CISS-like signals in isolated D--$χ$B--A complexes are fully compatible with a spinterface mechanism, providing a unified conceptual framework for interpreting both device-based and molecule-internal CISS platforms.

On-chip integration of two-dimensional transition metal dichalcogenides (TMDs) with photonic waveguides underpins a growing class of compact optoelectronic and sensing devices, whose performance depends on the optical absorption of the 2D material in this integrated environment. We demonstrate source-free, on-chip evanescent absorption spectroscopy of MoS$_2$ and WS$_2$ flakes integrated on Si$_3$N$_4$ waveguides, in which the broadband defect-related photoluminescence of the Si$_3$N$_4$ core itself acts as the internal probe. Light generated inside the core propagates under the TMD flake and is spectrally attenuated by evanescent interaction with the 2D material, and normalizing the transmitted spectrum to a reference waveguide without TMD yields the energy-dependent absorption response without any external illumination. The extracted guided absorption agrees with Raman and local micro-photoluminescence characterization and resolves differences between spatially uniform and non-uniform flakes. A complementary top-pumping scheme recovers guided WS$_2$ photoluminescence, whose spectral reshaping during propagation is quantitatively reproduced by the measured absorption. This source-free platform qualifies 2D materials in their integrated photonic environment and provides a direct route toward waveguide-integrated TMD photodetectors and sensors.

Quantum discord captures quantum correlations beyond entanglement and can remain finite even when the entanglement vanishes. We investigate the transient nonequilibrium dynamics and steady-state behavior of quantum discord and classical correlations in a double quantum dot (DQD) system coupled to fermionic reservoirs. By employing a quantum Langevin equation formalism, we obtain the exact reduced density matrix of the system, enabling a comprehensive analysis of its quantum and classical correlations under nonequilibrium conditions. The influence of system-reservoir coupling strength, spectral bandwidth, thermal bias, and varying initial state on both the transient dynamics and steady-state correlations is systematically analyzed. Quantum discord remains finite in the nonequilibrium steady state over a broad parameter range. Although thermal gradients reduce the overall magnitude of correlations, quantum discord persists and exhibits greater resilience. These results demonstrate that nonequilibrium electronic transport, together with the environmental spectral properties and reservoir asymmetry, provides an effective means of controlling nonclassical correlations in mesoscopic systems and establishes quantum discord as a robust hallmark of open fermionic quantum devices.

Driven nonequilibrium systems can host emergent functionalities beyond equilibrium, but real-time access to excited-state dynamics remains limited. Here we report real-time measurements of phonon-driven charge and spin dynamics in excited states of a double quantum dot. Under phonon irradiation, resonant inter-dot tunneling emerges at triplet resonance. Time-resolved charge sensing reveals that the resonant inter-dot tunneling is strongly modified by spin blockade. For weaker inter-dot coupling, the nonequilibrium phonon environment generates a unidirectional transport cycle along the phonon density gradient.

Altermagnets provide a promising platform for spin-polarized transport without net magnetization, but their transport properties are usually discussed in terms of momentum-space spin splitting. Here, using first-principles calculations and quantum transport simulations, we show that the charge-ordered altermagnet $α$-Fe$_2$PO$_5$ exhibits a distinct form of real-space spin selectivity despite weak altermagnetic spin splitting near the Fermi level. The charge order creates inequivalent Fe$^{2+}$ and Fe$^{3+}$ sites within each sublattice, while the puckered C-type antiferromagnetic stacking suppresses inter-sublattice transport. As a result, electron and hole doping activate spin-polarized transport predominantly through Fe$^{3+}$- and Fe$^{2+}$-based channels, respectively. These atom-selective channels carry opposite spin polarizations on the two antiferromagnetic sublattices, giving rise to a globally compensated charge current with hidden Néel spin character. We further propose an all-in-one $α$-Fe$_2$PO$_5$ tunnel junction, where matching or mismatching atom-selective conduction channels yields orders-of-magnitude conductance modulation. Our findings establish a real-space design principle for atomically controlled spin functionality and spintronic devices.

Structurally well-ordered NiS$_2$ nanocrystals with an average diameter of $27.0 \pm 6.5$ nm retain the bulk-like two-step antiferromagnetic transitions, as shown by magnetization and heat-capacity measurements. Below the lower transition, the nanocrystals exhibit a hysteretic ferromagnetic response with large coercivity, exchange bias, and a vertical loop shift after field cooling, whereas the $M$-$H$ response just above the transition is nearly linear. These features are best explained by uncompensated surface moments generated where the low-temperature antiferromagnetic order terminates at the nanocrystal surface. The absence of a clear additional bulk-like weak-ferromagnetic component constrains homogeneous-canting models and indirectly favors a domain-wall scenario for the weak ferromagnetism of bulk NiS$_2$.

Complex interactions among regulatory molecules determine the rules underlying cell growth and division in microbial cells. While the governing molecular network may not always be obvious, it is well known that correlations among certain physiological quantities measured in experiments, such as birth-size, division-size, division-time, and division-added-size, can differentiate among various cell-division models, such as Timer, Sizer, and Adder. Here we show that, apart from these correlations, which we extend for the case of stochastic single-cell growth and stochastic asymmetric partitioning, probability distributions of these quantities and statistical relationships between them can also be used to differentiate between these division models. Interestingly, we show that these quantities can not only differentiate the division models, but also distinguish among the single-cell growth paradigms, such as linear and exponential growth. We then demonstrate this differentiability among various division and growth models by comparing our analytical results with published experimental data. We further show that these results remain valid even when the growth rate of a cell is correlated with the growth rate of cells from previous generations in the lineage.

We study a slow heat-driven flow in a gas of elastically colliding hard disks confined to a long channel. The initial state consists of two regions with large temperature and density contrasts but nearly equal pressures, leading to a low-Mach-number, nearly isobaric evolution. In the dilute limit, the corresponding isobaric hydrodynamic theory reduces to a previously known ideal-gas description. We extend this theory to finite densities by incorporating a non-ideal equation of state of a hard-disk fluid, and solve the resulting one-dimensional equations numerically. Finite-density effects produce appreciable deviations from the ideal-gas prediction. We then test the theory directly against event-driven molecular dynamics simulations of hard disks and find very good agreement in both the dilute and finite-density regimes. The results provide, to our knowledge, the first particle-level test of isobaric gas dynamics of a strongly inhomogeneous cooling flow.

Fusion of two epithelial cell sheets brought together in a bilayer configuration is a common step in animal morphogenesis, yet, in contrast to other epithelial fusion processes such as wound healing in a monolayer of cells, it has not been a strong focus of modeling efforts. Here we consider a preliminary stage of bilayer fusion, recently termed "docking." In multiple instances of docking that span apical and basal varieties, cells appear to have a tendency to remodel so as to co-localize their bilateral junctions (match their edges) across the bilayer. Motivated by this observation, we introduce a bilayer cellular Potts model that couples two standard 2D area- and perimeter-elasticity models via short-range, out-of-plane interactions between cell edges. The new coupling involves a single adjustable parameter that minimally models the combined effect of dynamic cytoskeletal protrusions, cadherins, and other potential edge-associated adhesion molecules. Our model predicts that bilayer edge matching is maximized when the two monolayers are in their fluid-like regimes (average cell shape index greater than 4.6 in our implementation), and when the bilayer coupling strength strikes a balance between in-plane and out-of-plane energy scales. At higher coupling strengths, the system tends to get stuck in metastable states with sub-optimal edge matching. Exploration of the mechanisms of edge matching reveals that pairs and quadruplets of coordinated T1 transitions play a particularly important role. We also find numerous examples of emergent features we term "domain walls" - branching or unbranching curves that cross no matched edges, but that separate regions of nearly complete matching. These domain walls can be both system spanning and long lived. Finally, we extend our model to crudely account for bending of the two sheets, and study the distributions of docking front speeds that result.

The anomalous Hall angle (θA) is a measure of the efficiency of converting a longitudinal driving current to a transverse spin-polarized Hall current. For anomalous Hall sensing, a large anomalous Hall angle can improve the sensitivity of magnetic field detection. However, modulation of this angle is challenging and magnetic materials typically have low angles of 0.1 to 3°. Here, we report modulation of the anomalous Hall angle in the magnetic Weyl semimetal Co3Sn2S2. We propose that the angle parameter tanθA can be formulated as a function of the product of electrical resistivity and anomalous Hall conductivity. Our scheme was utilized to demonstrate the modulation of tanθA up to a magnitude of 0.46, corresponding to an angle of around 25°. Microfabricated anomalous Hall devices using Fe-doped Co3Sn2S2 single-crystalline nanoflakes exhibit a high Hall sensitivity of 7028 μΩucm/T and a magnetic field detectability of 23.5 nT/Hz0.5 at 1 Hz.

We study Floquet thermalization in a periodically driven quantum non-integrable Ising chain by combining two operational diagnostics: subsystem thermalization and work statistics. For generic interacting Floquet systems, stroboscopic dynamics lead to heating toward an infinite-temperature state at low driving frequencies, to a prethermal state during the crossover regime, and then to a finite-temperature state at high driving frequencies. We show that reduced density matrices of small subsystems provide a precise measure of local equilibration and clearly resolve prethermal plateaus. In parallel, we analyze the statistics of work performed over a Floquet cycle using both the characteristic function of work with and without the two-point measurements, and the related fluctuation theorem, which captures coherent contributions and deviations from thermal equilibrium. By comparing these two diagnostics within the same Floquet setting, we demonstrate that work statistics encode the same dynamical crossover that governs subsystem thermalization. Our results establish a unified and experimentally accessible framework for characterizing Floquet thermalization, prethermal regimes, and coherent energy absorption in interacting quantum systems.

We study the spin Hall (SH) effect for the Dirac electrons in terms of the spin and magnetic-moment accumulation coefficients $-Γ^{00} g_{s(m)z}^{\phantom{s(m)z} xy}$. We take short-range nonmagnetic impurities into account within the self-consistent $T$-matrix approximation. Similarly to the universal scaling for the anomalous Hall (AH) effect, we find three disctinct regimes by changing the electric conductivity $σ^{yy}$; the superclean regime with $-Γ^{00} g_{s(m)z}^{\phantom{s(m)z} xy} \propto σ^{yy}$ owing to the skew scattering, moderately dirty regime with almost constant $-Γ^{00} g_{s(m)z}^{\phantom{s(m)z} xy}$, and dirty regime with a new scaling relation $-Γ^{00} g_{s(m)z}^{\phantom{s(m)z} xy} \propto (σ^{yy})^{0.6}$ whose exponent differs from that of the AH conductivity. Our results construct a unified theory of the SH effect without any ambiguity of spin current.

This study establishes a theorem that provides sufficient criteria for identifying terms in any time-independent Hamiltonian that have no influence on local dynamics, local observables, or any local phenomena, without any approximation. The usefulness of this theorem is demonstrated by predicting the behavior of three systems. The first system consists of multiple qubit chains with Ising interactions. The second system is formed by a single qubit chain, differentiated by the substitution of Dzyalonshinskii-Moriya (DM) interaction for the last Ising interaction. The predictions were verified by analytically deriving the reduced dynamics of both systems. A third system was also considered, namely, a short qubit chain with a transverse magnetic field on the intermediary environment qubit. This transverse magnetic field signals that the commutation relation required by our theorem no longer holds. The third system's local dynamics were also derived analytically, demonstrating that all Hamiltonian constituents contributed to the reduced dynamics. The physical consequences of our theorem's inapplicability were also explored using this third system by analyzing the entanglement dynamics between the qubits that do not directly interact. The results showed that the entanglement is caused by an \textit{emergent coupling} between the two non-directly interacting subsystems. This \textit{emergent coupling} arises from the non-commutativity of the intermediate interactions connecting the two subsystems. When searching for these \textit{emergent couplings}, our theorem can eliminate intermediate interactions that will not produce them.

Electrostatically defined quantum dots (QDs) with layer-antisymmetric gating in Bernal-stacked bilayer graphene (BLG) open a local gap and generate a mass-like term with opposite sign in the two valleys, producing strongly valley-dependent scattering without magnetic fields, strain, or spin-orbit coupling. Building on this mechanism, we propose a tunable platform based on such QDs for valley-dependent electron optics in BLG. Using a four-band continuum model and a generalized multiple-scattering formalism, we analyze scattering of Gaussian electron beams from single- and multi-dot architectures and compute valley-resolved currents and angular profiles. A single dot produces distinct valley-dependent deflection, while multi-dot configurations enable enhanced control: identical-dot arrays act as valley splitters, whereas oppositely gated pairs function as valley filters. Combining these elements yields tunable generation, steering, and filtering of highly valley-polarized currents with strong suppression of forward transmission. The required energy scales, gate asymmetries, and device dimensions are within experimentally accessible regimes for dual-gated BLG, establishing quantum-dot arrays as a realistic platform for controllable valley-resolved electron optics.

Multi-level selection and senescence do not at first sight have much in common. Here, we demonstrate that the emergent mortality patterns generated by demographic senescence can be understood as the product of multi-level selection. We formulate a two-level Moran type process and use its scaling limits to illustrate that a simple mathematical framework that models multi-level selection in group-structured populations also models damage accumulation patterns and resultant mortality curves in ageing organisms. To verbally make the connection, observe that defectors spread within a group consisting of cooperators and defectors; when groups compete against each other, defector-rich groups suffer, and between-group selection causes such groups to be systematically under-represented. Exactly analogously, senescing individuals accumulate damage to physiological sub-systems, and `damage begets damage'; individuals who are more damaged are more likely to die, hence damage-rich individuals are systematically under-represented in later age classes. Thus, emergent senescence patterns in complex, integrated organisms are formally equivalent to the patterns generated by a within-generation multi-level selection process in which intra-organismal sub-systems play the role of particles, organisms play the role of collectives, and selective disappearance plays the role of group selection.

Synchronization and swarming are canonical manifestations of self-organization, observable across scales from cellular processes to animal flocks. This study investigates the collective dynamics of a novel agent-based model where individuals exhibit both spatial mobility and internal, two-mode stochastic oscillatory states. By introducing a local, distance-dependent coupling between the agents' spatial configuration and their internal state transitions, we establish a mutual feedback loop that drives complex pattern formation. Through large-scale numerical simulations, we identify seven distinct morphological configurations, ranging from stationary \textit{Filled-disk} states to highly disordered \textit{Intense-motion} regimes. By performing a rigorous quantitative analysis of the rotational energy and radial dispersion, we transcend simple morphological classification and demonstrate that the system organizes into discrete, quantized topological attractors. We derive a macroscopic scaling law, $Ω\propto r^{-1/2}$, which proves that the emerging rotating states are not rigid-body rotations, but rather composite differential vortex structures characterized by spontaneous chiral symmetry breaking. Our results suggest that these stable, quantized dynamical states are fundamental features of systems governed by bidirectional spatial-phase feedback, offering a robust framework for designing autonomous, decentralized robotic swarms.

Superconducting spiral inductors are emerging as key components for radio-frequency (RF) reflectometry, a widely used readout technique for semiconductor spin qubits. Future scalable quantum-computing architectures are expected to operate at elevated temperatures and magnetic fields, placing new demands on the performance and stability of superconducting circuit elements. Here, we present a systematic study of NbTiN spiral inductors under temperatures of several kelvin and magnetic fields approaching 1 T. By combining weakly coupled resonator measurements with independent two-port inductance extraction, we separate inductive and capacitive contributions to device behaviour and directly identify the origin of resonance shifts and quality factor degradation. Furthermore, we establish practical design metrics linking geometry, temperature sensitivity, and magnetic-field robustness. These results provide a general framework for benchmarking superconducting inductors and guiding the design of future RF-reflectometry circuits for practical quantum technologies.

We study the finite-temperature topological order of the three-dimensional $\mathbb{Z}_2$ toric code in a generic magnetic field, where every higher-form symmetry is explicitly broken and can at most be emergent. We show perturbatively, and confirm by large-scale quantum Monte Carlo, that the topological entanglement entropy stays quantized at $γ= \ln 2$ throughout the topological phase -- at finite temperature and under the symmetry-breaking field alike -- and collapses to $0$ across the thermal transition, a quantization protected geometrically by the Bianchi identity rather than by any exact symmetry of the system. The plateau $γ= \ln 2$ is, however, not invariant under quasi-local channels: a constant-depth channel can generate this identical quantized value from a trivial product state. We therefore introduce the decoded Wilson-loop correlation $f_W$, which quantizes to $1$ in the topological phase and $0$ in the trivial phase as $L\to\infty$ and, unlike $γ$, is a quasi-local-channel invariant -- a robust topological invariant of the mixed state.

We investigate work statistics in monitored transverse-field Ising chains subject to both a quantum quench of the transverse field and either stochastic quantum jumps or controlled measurement sequences. For generalized measurements, we derive a trajectory-resolved generating function for work statistics in the two-point energy measurement scheme. Evaluating it using a fermionic Gaussian-state formalism, we show that, under stochastic jump dynamics, the work distribution crosses over from a comb-like structure to an essentially Gaussian form with shrinking sub-Gaussian tails, as the number of detection events grows. For controlled jump protocols, the energy added by each jump is constant when successive jumps are causally disconnected but decreases and then saturates when they lie within each other's light cone, yielding linear and sublinear growth of the average work, respectively. For monitored quenches, continuous observation washes out the fine structure of the isolated-quench distribution and again drives the statistics toward Gaussian behavior.

We introduce a Monte Carlo strategy for directly estimating partition function ratios between distinct global sectors of a lattice theory. It enlarges the configuration space to sample an interpolating family whose endpoints are the desired sectors, and uses flat histogram methods to reconstruct the corresponding free energy difference. Although the construction is more general, we focus here on the three-dimensional Ising model on the slab $\mathbb{R}^{2}\times S^{1}_{L_{z}}$ at the bulk critical point, comparing the untwisted periodic sector with the $\mathbb{Z}_{2}$-twisted antiperiodic sector. A large-volume and aspect ratio extrapolation gives the symmetry-twisted thermodynamic Casimir difference $Δ_{\mathbb{Z}_{2}}=0.327(2)$ directly, without lattice derivatives or bulk subtractions. This provides an independent twisted sector probe of tensions observed in periodic sector thermodynamic Casimir observables. More generally, the method gives direct but selective numerical access to CFT compactification data, including estimates of the effective thermal screening scale and the $\mathbb{Z}_{2}$-odd sector energy gap on $T^{2}$.

Magnetic tunnel junctions (MTJs) are known for their digital applications (memory and logic). A special class of them called "straintronic" magnetic tunnel junctions (s-MTJ) has lately emerged as a potential platform for analog applications because their conductance can be varied continuously with mechanical strain generated with a gate voltage. The conductance versus gate voltage (transfer) characteristic always has a linear region and that can be leveraged for a variety of analog applications. Here, we discuss one such application, namely analog voltage amplification. If the s-MTJ's gate voltage is fixed around the midpoint of the linear region and a small ac voltage is superimposed on it, then the ac voltage can be amplified without distortion as long as its amplitude is small enough to avoid gate voltage excursion beyond the linear region. Unlike a transistor-based voltage amplifier whose amplification is determined solely by the transistor's internal parameters - namely the transconductance and Early resistance - here the amplification can be varied by an external power supply voltage. We examine the maximum allowed amplitude and frequency of input signal for distortion-free amplification by modeling the magnetization dynamics and derive an expression for the amplification.

Understanding how dissipation modifies quantum phase transitions is a central challenge in many-body physics. A paradigmatic example is the one-dimensional Mott transition, which in isolated systems separates a conducting Luttinger liquid (LL) from a Mott insulator (MI). Here, we study the fate of this transition in the presence of dissipative baths locally coupled to the density. Using bosonisation and an exact integration of the bath degrees of freedom, we show that dissipation fundamentally reshapes the phase diagram for bath exponents $s<3/2$, where $s$ characterises the low-energy bath spectrum. Rather than undergoing a direct LL-MI transition, the system develops an intermediate dissipative phase (DP) that is compressible and gapless, yet has zero superfluid stiffness. As a result, the conventional Mott transition splits into two distinct critical phenomena: a Berezinskii-Kosterlitz-Thouless transition from the LL to the DP, followed by a new commensurate-incommensurate transition from the DP to the MI. We derive an effective field theory for the latter transition and characterize its universality. For $1<s<3/2$, the critical exponents vary continuously with the bath exponent as $β=ν=1/z=s-1$, while for $s<1$ the transition is governed by $β=ν=1/z=0$ and the doping vanishes sharper than any power law. State-of-the-art Monte Carlo simulations quantitatively support our predictions. These results demonstrate that dissipation can qualitatively alter the nature of the Mott transition and generate novel critical behaviour in strongly correlated one-dimensional systems.

The three-dimensional organization of the genome exhibits rich, scale-dependent structure, as revealed by both chromosome contact maps (e.g., Hi-C maps) and chromatin density measured by microscopy. Recent studies have reported multifractal scaling in these data. Yet, the origin of this scaling behavior remains unclear: existing efforts describe it through postulated models. Here, we show that the multifractal structure of Hi-C maps is a direct consequence of the power-law contact probability $P(s)$, which is itself an empirical observable measured from Hi-C maps. Starting from $P(s)$ with a single exponent $γ$, we analytically derive the mass exponent $τ(q)$, which characterizes how the $q$-th moment of contact density scales with box size $l$ used to coarse-grain the genomic coordinate. This multifractal behavior reflects the geometric competition between intra- and inter-segment contacts. We find that the slope of $τ(q)$ at large $q$ is given by $2 -γ$ when $γ<1$, and by $1$ when $γ\geq 1$. We further show that this behavior is robust to noise and consistent across diverse organisms, indicating that it is a universal feature of chromatin organization. We extend our analysis into double-exponent $P(s)$, and show the $l$ dependence in multifractal behavior. Taken together, these results provide a physical explanation for multifractal scaling and establish a direct link between the multifractality in Hi-C maps and polymer contact statistics, with the large-$q$ slope of $τ(q)$ mapping onto a known polymer contact exponent.

Small-scale systems are inherently subject to environmental noise that can be harnessed constructively to realize reliable logic operations -- a phenomenon known as logical stochastic resonance (LSR), where a bistable system produces correct logical outputs within an optimal window of noise intensity. The Brownian dynamics governed by appropriate inputs inside a double-well potential, modeling the bistable system, mimic the logic operations. The two wells of this potential represent two distinct logical output states 0 and 1. Asymmetry in this potential is known to be essential for improving logical reliability. However, prior studies have focused on energetic asymmetry, characterized by unequal depths of the two wells of the potential. This left the role of the width asymmetry in the potential, unexplored. This latter class of asymmetry emerges due to the dissimilar widths of the two wells of the potential. It can be classified as the entropic asymmetry between the two logical output states. Here, we systematically investigate the effect of width or the entropic asymmetry in the system on the logical response for OR and AND gate operations. Unlike energetic asymmetry, width asymmetry preserves the energetic equivalence of the two intrinsic logical output states, making it a geometric effect. We find that increasing width asymmetry consistently improves the optimal P(logic), the quantifier measuring the successful logical outcome. Moreover, when it is combined with an energetic bias, it produces reliable logic gate operation at a significantly reduced energetic cost compared to the symmetric case. The requirement of this energy bias also diminishes gradually with the increasing degree of width asymmetry in the potential.

For particles of fixed mass, classical equilibrium statistical mechanics dictates a Maxwellian velocity distribution determined solely by the temperature, regardless of the interactions, density, or structure. Supercooled glass forming liquids realize long lived metastable states that evade equilibrium crystallization and may thus violate assumptions underlying Maxwellian statistics. We numerically demonstrate that supercooled liquids can exhibit persistent non-Maxwellian velocity distributions with deviations connected to their exceptionally slow super-Arrhenius relaxation. Our work is motivated by a general result establishing that long lived metastable states may exhibit finite width distributions of intensive variables. A distribution of temperatures implies non-Maxwellian velocity statistics. We test this prediction by introducing stochastic thermostats that generate stationary states while, unlike conventional thermostats, not imposing Maxwellian velocity distributions. Simulations with these thermostats yield long lived states that have, by comparison to Maxwellian velocity distributions, an excess kurtosis $0<κ\lesssim0.3$. Crystallization is strongly impeded with increasing $κ$. In a minimal description, temperature fluctuations are characterized by a dimensionless width $\overline{A}$ with $κ\simeq3\overline{A}^{2}$. The nearly constant $\overline{A}$ (of an average value $0.08$ and standard deviation $0.03$) found in viscosity data collapse across $45$ glass formers and in specific heat signatures is consistent with kurtosis found in our simulations. Long time non-Maxwellian velocity statistics may thus link slow relaxation, transport, and thermodynamic measurements. Independent of the tested theory, the stochastic thermostats that we introduce offer a molecular dynamics route to non-Maxwellian velocity statistics.

Principal component analysis is widely used to characterize structure in the dynamics of recurrent neural networks. For stationary noise-driven dynamics, the distribution of variance among the principal components is determined by the spectrum of the stationary covariance matrix. While the spectral properties of this matrix are well-understood for linear networks with normal synaptic weight matrices, our understanding of the stationary covariance spectrum for random non-normal dynamics remains incomplete. In this note, we use a free-probability approach to formally derive a closed functional equation for the moment generating function of the limiting stationary covariance spectrum of discrete-time dynamics with random non-normal Gaussian weights. This characterization allows us to analyze the behavior of tail eigenvalues in the critical regime. In contrast, applying the same approach to the analogous continuous-time dynamics leads to an infinite hierarchy of Schwinger-Dyson equations, rather than a closed scalar equation. We conclude with some comments regarding the relevance of these results to comparisons of models of non-normal dynamics to neural data.

The coalescing colony model provides a minimal framework for biological invasions with long-range dispersion. In its standard formulation, the dispersion range is assumed independent of the size of the invading population. Here, we relax this assumption and consider size-dependent dispersal: a main colony of linear size $r$ emits secondary colonies at distance $r^μ$, with $0 \leq μ\leq 1$. We derive the generalized dynamical equations for this extended model and map out the growth phase diagram for the leading order contribution. Depending on $μ$, the main colony exhibits distinct regimes: linear expansion, power-law growth, exponential regime and finite-time blow-up. We confront these theoretical predictions with a spatially explicit physical model. While the coalescing colony approach correctly captures the scaling of the perimeter, it fails to predict the scaling of the volume. We trace this discrepancy to an effective breakdown of circular symmetry in the morphology of the main colony. Finally, we quantify temporal evolution of the population fraction residing outside of the main colony. The coalescing colony model predicts its decay to~$0$ like a power-law when~$μ<1$, and a macroscopic amount of the population remains in the secondary colonies at~$μ=1$. Simulations of the physical model reveal a persistent satellite population not captured by the theory at~$μ>μ^*\approx 0.7$. Broadly, our findings highlight how coupling dispersal range to population size fundamentally alters invasion dynamics, with implications for biological invasions, metastatic growth, and urban expansion.

Reconstruction of the atomic crystal structure in twisted 2D materials has been demonstrated to be responsible for multiple exciting phenomena in van der Waals heterostructures, from the appearance of flat bands in twisted bilayer graphene to Wigner crystallization in transition metal dichalcogenides (TMDs). However, there are still no experimental methods for accessing the 3D atomic distributions nor models that describe the exact atomic shifts in such reconstructed structures, which significantly impedes the development of the field. Dark field (DF) transmission electron microscopy (TEM) has been conventionally employed to visualize the local in-plane atomic displacements. Here we expand this method to obtain a full description of the reconstructed atomic systems and demonstrate the quantitative relations between the local stacking and the intensity in the DF image. We show how local 3D atomic displacements and the interlayer distance can be extracted from a DF image.

Magnetic active particles offer a versatile route to externally controlled microscale transport by combining self-propulsion with field-tunable orientation, as realized in both synthetic and living magnetic microswimmers. Here, we develop a theoretical framework for three-dimensional active Brownian motion in a uniform magnetic field, incorporating coupled translational and rotational dynamics and providing analytical approximations for low-order displacement moments. At long times, the system dynamics reduces to a combination of enhanced diffusion and permanent drift absent in regular active Brownian particles. The field acts as an external controller, channeling activity toward one of these two types of motion. At intermediate time scales, the interplay between rotational noise, self-propulsion, and magnetic alignment results in pronounced non-Gaussian displacement statistics. First-passage properties exhibit strong field sensitivity, highlighting the potential of magnetic guidance to optimize search processes and targeted delivery in active matter systems. Theoretical predictions are validated by numerical simulations.

There are many situations in nature and industry where small particles are exposed to gradients of charged polymers, such as enzymes in biological gradients of DNA or RNA, virus particles in respiratory droplets, and colloidal particles in stratifying paint layers. Here, we study the phoretic propulsion of charged microparticles in a polyelectrolyte gradient. We theoretically predict the emergence of a macroscopic electric field from charge-separation dynamics in a polyelectrolyte gradient under a continuous diffusive driving force. We confirm the presence of this self-generated electric field experimentally and show that it significantly increases the phoretic velocity of the microparticles. Finally, for high molecular weight polyelectrolytes we observe that propulsion becomes gradient-independent, consistent with diffusiophoretic predictions for asymmetric electrolytes. Our results show that self-generated electric fields in polyelectrolyte gradients can enhance microparticle transport, with potential applicability wherever charged species of different mobility are continuously driven out of equilibrium.

Direct dc-current signatures of unconventional superconductivity remain scarce. Existing probes of unconventional pairing are typically indirect, relying on phase-diagram anomalies, responses to external fields, or optical measurements. Here we propose a zero-field Hall drag effect as a direct transport signature of chiral superconductivity. The effect arises from Coulomb drag between quasiparticles in a chiral superconductor and those in an adjacent time-reversal-symmetric normal layer. We develop a minimal hydrodynamic theory that includes both quasiparticle normal current and condensate supercurrent in the superconducting layer. In an open-circuit superconducting layer, the condensate generates a counterflowing supercurrent that cancels the net layer current, while a finite quasiparticle current remains and mediates the transverse drag response. This results in anomalous Hall voltage signal appearing abruptly when $T$ is lowered below $T_c$, of the sign reflecting the sign of the superconducting order parameter phase winding.

Achieving robust functionality in active matter driven away from thermal equilibrium is a current theoretical and experimental challenge. Several recent studies have reported edge currents--persistent transport along walls and density inhomogeneities--in chiral active matter. Yet, the microscopic rules that render these edge currents robust with respect to the confinement geometry and defects remain elusive. Here, we introduce a simple particle model of two-dimensional chiral active swimmers that undergo chirality switching and demonstrate that the model exhibits robust edge currents, i.e., when a single particle is confined, edge currents arise regardless of the confinement geometry or the presence of defects. We also investigate the collective behavior of interacting particles in bulk and find that chirality switching induces phase separation accompanied by edge currents along interfaces. This phase separation is distinct from motility-induced phase separation and is qualitatively explained by an effective hydrodynamic theory derived via bottom-up coarse-graining. Furthermore, by analyzing the topological properties of the linearized hydrodynamic equations, we show that the edge currents in our system are genuine topological edge modes. Notably, phase separation induced by chirality switching can be regarded as the coexistence of two topologically distinct domains. Our results provide guidelines for designing robust edge currents in active matter systems.

Addressable self-assembly asks that each building block assemble into a particular location in a target structure. Although particles may all be distinct, achieving high yield is a challenge because of monomer depletion: more target structures can nucleate than there are building blocks for, so they form partial fragments which cannot complete growth. We ask how to design the interactions between building blocks to achieve the highest yield in a given time. Using reaction equations describing all the intermediate steps of assembly, combined with numerical optimization, we show that the optimal interactions are such that (i) all bonds are either very strong or very weak, and (ii) the strong bonds form a spanning tree of the target structure. We then prove that when interactions form a spanning tree, monomer depletion cannot occur: assembly can always proceed downhill in energy space, with no kinetic traps. This result is a combinatorial property of the underlying interaction graph, and does not depend on the particular model for the kinetics. It suggests a robust design principle: create a network of strong interactions that has no loops, and make all other interactions much weaker. We validate this principle in numerical simulations of larger structures, and we further show that spanning trees that are more compact have typically better yield. Our results suggest a new framework for understanding monomer depletion and addressable self-assembly, which may be applied to DNA nanotechnology and which may give insight into the assembly pathways of certain multiprotein complexes.

Sampling rare conformation transitions between metastable states is a central challenge in atomistic simulations. While the committor function serve as an ideal reaction coordinate for driving enhanced sampling, their high-dimensional inputs and complex functional forms limit the efficacy of standard feedforward neural networks in modeling them. Inspired by recent breakthroughs in biomolecular structure prediction, we propose a novel committor learning framework grounded in the AlphaFold 3 paradigm. By integrating a lightweight, differentiable atom-level embedding with a simplified Pairformer architecture, our method inherently captures intricate dynamical features of diverse biosystems without requiring specialized prior knowledge. We demonstrate the superior expressiveness and accuracy of the proposed framework across multiple atomistic processes. For the folding of the chignolin mini-protein, our model reveals the finer-grained structure of its transition state ensemble (TSE) and a detailed bifurcated reaction mechanism. Furthermore, for calixarene host-guest systems, we develop a unified committor model that elucidates how ligand substituents regulate the ratio between distinct binding pathways, offering new perspectives for structure-based drug design.

Many subcellular condensates continuously produce biopolymers. Coupling Flory-Huggins thermodynamics to two-fluid viscoelasticity, we probe the diffusion of such source-driven polymeric droplets, and identify a universal structural compressive rim at their diffusion front. Integrating analytical scaling laws, numerical simulations, and experimental data, we show that this framework captures key structural and dynamic characteristics of the nucleolus, demonstrating the role of polymer diffusion in non-equilibrium biological transport.

Scalable quantum photonic technologies require deterministic sources of entangled photons that are compatible with established semiconductor manufacturing platforms. While self-assembled III--V semiconductor quantum dots are among the most promising sources of on-demand entanglement generation, their integration with silicon-based architectures remains a central challenge. Here, we demonstrate energy--time entanglement from a single InGaAs/GaAs quantum dot monolithically grown on a silicon substrate. Under coherent two-photon excitation, we achieve coherent control of the biexciton--exciton cascade, evidenced by Rabi oscillations and dressed-state formation. Using a four-channel Franson interferometer, we observe phase-dependent two-photon interference with visibilities up to $(64.0 \pm 7.0)\%$ for an 80 ps integration window (and $(49.4 \pm 1.9)\%$ for a 1600 ps window), approaching the threshold for Bell inequality violation at short time scales. These results establish monolithically integrated III--V-on-silicon quantum dots as promising sources of energy--time entangled photons for scalable quantum photonic technologies.

We theoretically investigate the linear and second-order nonlinear optical responses of valley-polarized bilayer graphene under uniaxial strain. Employing a low-energy effective Hamiltonian that incorporates trigonal warping and strain-induced anisotropy, we calculate the optical susceptibilities within the quantum kinetic formalism. We show that, while the second-order response vanishes in valley-balanced bilayer graphene owing to the cancellation of contributions from opposite valleys, a finite valley polarization lifts this cancellation and enables a net second-harmonic generation (SHG) signal. Uniaxial strain substantially modifies the nonlinear response by distorting the low-energy electronic structure and altering the pseudospin texture, producing a highly anisotropic SHG spectrum. Pronounced resonant enhancements occur at photon energies $\hbarω\approx E_f$ and $\hbarω\approx 2E_f$, associated with two-photon and one-photon interband resonances, respectively. Remarkably, changing the sign of the strain parameter reverses the direction of the induced second-harmonic current, providing a mechanically controlled switching mechanism for nonlinear optical transport. These results establish strain engineering as an effective route for manipulating valley-dependent nonlinear optical phenomena in bilayer graphene and suggest new opportunities for tunable mid-infrared photonic and valley-optoelectronic applications.

We study the critical dynamics of the quantum random-field Ising model using the nonperturbative functional renormalization group (NP-FRG). The static critical behavior is found to be controlled by the zero-temperature fixed point of the classical random-field Ising model, where both thermal and quantum fluctuations are dangerously irrelevant. Considering a family of quantum dynamical universality classes defined by a bare dynamical kernel $F_Λ(ω)\sim |ω|^σ$, we show how this fluctuationless fixed point nevertheless controls the quantum dynamics by computing the full Matsubara-frequency dependence of the running dynamical kernel $F_k(ω)$. This is essential at zero temperature: a naive treatment of the dynamical kernel flow leads to a divergence at a finite length scale, resulting in apparent localization. In contrast, keeping the full frequency dependence of the dynamical kernel and choosing a regulator adapted to its running scale yields a controlled flow. The resulting dynamics is of activated form, with a relaxation time given by $\ln τ\sim ξ^Ψ$. The exponent $Ψ$ is determined by the static RFIM fixed-point exponents and by $σ$. At finite temperature, the flow crosses over to the classical thermally activated scaling of the random-field Ising model. These results provide a quantitative field-theoretic realization of the heuristic activation scenario proposed earlier for the quantum random-field model and establish a framework for analyzing the dynamics of other disordered quantum systems that may exhibit similar tentative localization-like singularities.

Indentation is a practical route for probing soft materials when standard tests are difficult, destructive, or cannot be performed in situ. Conventional indentation is usually interpreted in the shallow-depth regime, where the indentation depth D is small compared with the indenter radius R. In this limit, the response is controlled by local contact geometry and primarily identifies the small-strain Young's modulus E. Here, we show that at deep indentation, D >> R, flat and spherical indenters converge to the same parabolic force-depth law, F = beta E D^2. The coefficient beta is independent of indenter radius and tip shape, only mildly affected by interfacial friction, and controlled by the hyperelastic strain-stiffening response. Finite-element simulations show that this scaling arises from strain-energy delocalization: the region where SED/mu > 0.01 expands into a spheroidal domain whose size scales with D. The activated volume therefore scales as D^3, giving stored elastic energy U ~ E D^3 and force F = dU/dD ~ E D^2. Far from contact, the strain-energy-density fields collapse toward the Boussinesq far-field solution when distances are normalized by a = sqrt(F/E), which scales as D in the deep-indentation regime. These results provide a mechanistic basis for tip-shape independence and link beta to the Ogden strain-stiffening parameter alpha, enabling hyperelastic parameter extraction from deep-indentation data.

Population immunity carried over from past epidemics of an antigenically variable pathogen influences the epidemic of new variants based on their antigenic similarity to the previous ones. We develop a recurrent SIR model where a population faces sequential, antigenically related variants. The model yields a recurrence map for the population susceptibility to successive variants under the assumption of status-based population immunity. The model reveals that stable, equal-sized recurrent epidemics occur across broad parameter ranges, but can be destabilized when transmission is strong and antigenic escape is limited, leading to period-2 or more, or even more complex epidemic dynamics. Epidemic size is maximized at an intermediate basic reproduction number: higher transmissibility boosts immediate infection but also enhances cross-immunity, reducing future susceptibility of the population. Our results clarify how immune history shapes recurrent epidemics and why success in one wave does not ensure larger future epidemics.

Off-shell Bethe vectors for a generic $\fg$ invariant integrable model are constructed through the currents of the Yangian doubles of the classical series. These off-shell Bethe vectors are shown to satisfy the defining properties which were used in \cite{LPR-RR} to prove the rectangular recurrence relations and verify the eigenvalue property of the on-shell Bethe vectors in $\ggo$-invariant integrable models.

Optimizing vaccine prioritization is often treated as the default policy response when vaccine supply is limited. Yet optimized prioritization carries administrative, ethical and communication costs, motivating an upstream question: whether differences among vaccine allocations can alter epidemic outcomes enough to make optimization epidemiologically necessary. We show that optimization is not always worth pursuing: in some regimes, vaccination markedly reduces epidemic burden, but many feasible allocation rules perform almost equally well, making the necessity of optimization low. We quantify this necessity as the range of epidemic outcomes generated by different allocations under fixed supply and show that it is governed by competition between vaccinating high-contact groups to slow transmission and vaccinating groups that benefit most directly: necessity is low when these protection routes are balanced and high when one dominates. Increasing transmission intensity changes this balance and drives a transition in the optimal allocation from transmission-focused prioritization toward direct protection. Different prevention objectives exhibit distinct transition thresholds, creating regimes in which optimizing one objective substantially compromises another, thereby revealing when the choice of prevention target matters most. This framework reframes vaccine prioritization as a prior decision problem, identifying when optimization is warranted, when simpler rules suffice, and when prevention goals conflict.

Dissipative solitons constitute a robust class of self-localized nonlinear states sustained by the dynamic balance between nonlinearity and gain-loss, possessing an intrinsic stability that stems from their fundamental attractor nature. When combined with lattice truncation, this balance gives rise to dissipative surface solitons (DSSs), whose existence and stability are jointly dictated by boundary-induced confinement and non-Hermitian dynamics. In two-dimensional truncated lattices with linear gain and loss, surface localization emerges within gap regimes, where families of DSSs bifurcate from linear surface localized gain modes as the nonlinearity increases. Increasing the number of waveguide rows at the interface enriches the diversity of supported surface modes in both linear and nonlinear regimes. Although multiple DSS families with distinct phase configurations may coexist within the same gap, their dynamical stability is strongly phase selective. These insights establish linear gain-loss engineering as a powerful mechanism for controlling nonlinear surface localization and provide practical guidelines for realizing robust nonlinear surface states in gain-loss-tailored photonic platforms.

The coordination of prices in economics is one of the most complex phenomena. In particular, the classical and neoclassical approaches related to the economic theory provide some insights into such a complex coordination based on different formulations. However, these formulations have not been successful for explaining simple mechanisms to understand and predict a set of prices that theoretically clears all markets. Consequently, elementary cellular automata can contribute to clarify such a coordination problem by using simple computational rules to describe the theoretical bases of the classical and neoclassical economics. Therefore, we propose to use this type of cellular automata for explaining different escenarios of price coordination in which simple rules of price interactions generate stable and unstable patterns of coordination. We used an explorative data analysis based on the Shannon entropy for computing the uncertainty related to such generated patterns of coordination, and a Monte Carlo simulation approximation based on a Spearman correlation for evaluating the statistical significance of such price coordination. Findings suggested that the classical economics provides a consistent approach for understanding the coordination of prices because it emphasizes human interactions based on logical choices related to an objective data. On the other hand, the neoclassical approach does not propose any type of mechanism for describing the price coordination. The neoclassical individual is just a spectator and receiver of the unpredictable and supposed event of price coordination. As a result, by modeling the economic theory based on computational concepts, we reveal facts and believes behind the classical and neoclassical economics.

The two-dimensional Breitenlohner-Maison (BM) model is a classically integrable subsector of $D=4$ general relativity endowed with two commuting Killing isometries, obtained via Kaluza-Klein reduction to $D=2$. Integrable deformations of such a theory have recently been constructed via auxiliary fields in the so-called $ν$-frame. In this work we first extend this point of view by deriving the complementary auxiliary field perspective known as $μ$-frame, and then explicitly construct the uplift to $D=4$ of both descriptions, relying on an ansatz inspired by duality-invariant Lagrangian formulations of Einstein theory. The resulting four-dimensional deformed model thus obtained is a higher-derivative theory which lacks manifest diffeomorphism invariance in both frames. We comment on possible resolutions of this puzzling feature and on the physical interpretation of the model in $D=4$.

We present the verification of a thermo--hydrologic--mechanical capability implemented within the PFLOTRAN framework, with emphasis on benchmark-based assessment of the THM implementation. The thermal--hydrologic (TH) equations for mass and energy balance are solved on control-volume blocks or Voronoi cells, while the quasi-static momentum balance is solved on an element-based dual mesh. The coupling is achieved using a strictly sequential, non-iterative fixed-stress split strategy in which the TH system is solved implicitly for pressure and temperature, followed by a mechanics update for the displacement unknowns. Several verification problems are set up against poroelastic and thermo-poroelastic benchmarks, demonstrating agreement with analytical or semi-analytical benchmark responses for pressure diffusion, the temperature field, and mechanical deformation. In addition, we propose a treatment for discontinuities (e.g., fractures) based on mapping between mechanical and flow degrees of freedom, and validate the approach by comparison to an analytical solution. This work establishes the basis for thermo-poroelastic coupling in PFLOTRAN and provides a solid modeling foundation for a range of applications (e.g., enhanced geothermal systems and other subsurface energy storage) involving coupled thermal--hydrologic--mechanical (THM) processes in geologic porous media.

Accurate 3D localization of radiation interactions in scintillation detectors is essential for nuclear and particle physics, safeguards, and medical imaging, but remains difficult in light-starved regimes with limited photon statistics. We present PRISM, a multifocal plenoptic imaging system designed for millimeter-scale 3D position reconstruction in a single-volume scintillator. PRISM uses a multifocal microlens array with diverse focal lengths and high effective numerical aperture to balance photon collection with spatial and depth encoding. A Cram'er--Rao lower bound analysis shows that the multifocal design improves axial sensitivity over conventional unifocal plenoptic systems under photon-limited conditions. We build a prototype system, calibrate its optical response with a tunable light source, and form photon-limited measurements with $\mathcal{O}(100)$ detected photons. For sparse single-vertex events, we reconstruct interaction locations using an Alternating Descent Conditional Gradient-inspired algorithm and demonstrate an average 3D localization error of approximately 1 mm. We also provide an initial evaluation of double-vertex events, showing that localization improves as the axial separation between interactions increases. These results demonstrate that multifocal plenoptic imaging can mitigate the traditional trade-off between light collection and spatial resolution, providing a photon-efficient approach to 3D reconstruction in scintillation detectors and a foundation for future multi-scattering event reconstruction.

We present a symmetric Lanczos method for computing charge-changing QRPA strength functions in atomic nuclei. Starting from the finite-amplitude-method formulation of the QRPA linear-response problem, we derive equivalent spectral representations and, in the real case, a reduced eigenvalue problem involving the matrix products $MK$ and $KM$, where $M\equiv A+B$ and $K\equiv A-B$ are formed from the usual QRPA matrices $A$ and $B$. The resulting formulation enables a matrix-free Lanczos approximation of the Lorentzian-smeared strength function over a broad energy interval from a single Krylov run, in contrast to conventional frequency-by-frequency response calculations. Numerical tests for $^{112}$Sn and $^{150}$Nd first show that GMRES reproduces the converged iterative FAM strength profiles while requiring fewer iterations. Using GMRES as the frequency-by-frequency reference, we then show that the Lanczos approximation reproduces the same strength profiles with reduced overall cost. These results indicate that symmetric Lanczos projection provides an efficient and accurate approach for QRPA strength-function calculations when spectral information is required over an extended frequency range.

Valley coherence refers to a phase-coherent superposition of inequivalent momentum valleys, in which quantum information can be encoded in the relative valley phase. Chiral nematic liquid crystals, by imposing a flip of the spin angular momentum upon light reflection, provide an effective photonic environment for optically coupling excitons of the K and K' valleys in monolayer semiconducting transition-metal dichalcogenides. We experimentally demonstrate that using such liquid crystal as a substrate, it is possible through nearfield interaction to engineer a room temperature mechanism for inducing the intervalley coupling. Our results show that this approach provides a simple and scalable route toward valleytronic functionalities based on controlled coherent emission from valleys with opposite Berry curvature.

Weaving patterns in fabrics, traditionally used for aesthetic purposes, present a largely untapped opportunity to create metamaterials that serve as passive layers for sensing, filtering, and signal processing. However, the hierarchical architecture of fabrics makes structural design and wave prediction challenging. Here, we establish fully woven fabrics as phononic crystals that passively filter and route elastic vibrations. Using double weaving, we integrate a soft cotton weave with stiff woven copper inclusions to form periodic fabric lattices with engineered dispersion. A multiscale modeling framework that combines homogenization of weave blocks with an effective-property macroscale model enables computationally efficient design of phononic crystals. Simulations and experiments confirm a pronounced phononic bandgap for out-of-plane vibrations in a finite fabric crystal, while an equivalent pure cotton weave shows no band suppression in the corresponding frequency range. Building on the same platform, we realize a fully woven higher-order topological insulator. Modal analysis and transmission measurements reveal in-gap edge states and localized corner states. These results show that phononic bandgaps and topological states can be directly encoded through weaving patterns and material contrast, enabling passive vibration management layers and multifunctional waveguiding fabrics for sensing, haptic interfaces, robotics, and noise mitigation.

Phytofluidic platforms have enabled controlled studies of plant roots, however, most existing systems either impose geometric confinement without flow or introduce hydrodynamics in single-channel devices that limit throughput and disrupt downstream analysis. New experimental platforms are therefore needed to investigate how roots integrate mechanical confinement and hydrodynamic nutrient transport, two defining features of the rhizosphere that remain difficult to reproduce under controlled laboratory conditions. Here, we present the Plant-on-a-Disc (POD), a phytofluidic platform that enables the parallel cultivation of eight seedlings under controlled hydrodynamic conditions while allowing non-invasive, in situ multimodal analysis of the intact root-shoot system. The device is fabricated in PDMS using a cost-effective wire-drawing technique to generate radial microchannels that converge into a central sump beneath an optical window. This design enables sequential bright-field, fluorescence, and Raman measurements using a single microscope objective without disturbing neighbouring seedlings. Dimensionless transport analysis and finite-element modelling confirm that the radial architecture equalizes hydraulic resistance across channels, establishing creeping laminar flow with convection-dominated nutrient transport under physiologically safe shear conditions. Using Brassica seedlings, we show that hydrodynamic flow drives coordinated root responses across multiple scales. Roots grown in flow condition exhibit accelerated elongation, substantial ROS generation and anisotropic cortical cell expansion, accompanied by carotenoid signatures detected by Raman spectroscopy.

In previous work we sought to address the fundamental question ``What is life?''. Building on that conceptual foundation, we here integrate traditional criteria for living systems with our recent proposals on physical identity and self, and apply them to concrete biological cases. We suggest that life can be characterized as the preservation of a well-defined identity ``at all costs'' together with the presence of a physically grounded self, and we show how this perspective illuminates the organization and social behavior of Dictyostelium and other multicellular systems.

Searches for new physics at the LHC often look for localized excesses on smoothly falling background distributions. Several classes of background models have been considered, including polynomials and other parametric families; however, these approaches can require extensive analysis-specific development as datasets grow. In this work, we motivate the finite exponential mixture as a flexible semi-parametric class of functions for approximating falling distributions, drawing on results from extreme value theory. Using two published datasets ($n=28,619,185$ and $n=5,036$), we show that the exponential mixture performance is comparable to existing methods for both small and large datasets. Finally, in simulation studies ($n = 5,036$), we find that the finite exponential mixture exhibits small bias relative to the true statistical uncertainty while maintaining consistent nominal coverage in the bulk.

Developmental robustness is framed as progress through a fixed Waddington-type landscape. We argue that in morphogenesis this landscape evolves through coupled bio-signaling, mechanical, and physiological processes, while fluctuations aid exploration. In Hydra regeneration, stochastic Ca activity plays a major role in reshaping the landscape of accessible morphologies as regeneration unfolds, including the early progressive confinement of tissue fluctuations. We propose testing this framework of dynamical noisy canalization in developmental systems.

We present the Axion Polarimetric Experiment (APE), a cavity-enhanced polarimeter designed to search for ultralight axion and axion-like-particle dark matter through a time-dependent rotation of the linear polarization of laser light. In cavity-based schemes, intracavity quarter-wave plates can restore coherent buildup of the axion-induced orthogonal polarization, but their transmissive loss limits the achievable finesse. To avoid transmissive intracavity optics, we propose a folded Fabry-Perot cavity that employs dielectric phase-shifting mirrors. At an incidence angle near $45^\circ$, these mirrors provide a reflection-phase difference $Δφ\equiv φ_s-φ_p \simeq π/2$ between $s$ and $p$ polarizations and therefore act as reflective quarter-wave plates. We present the coating design, thickness optimization, and measurements of the phase shift and optical loss of the phase-shifting mirrors. Using a heterodyne polarimetric readout and an explicitly stated noise model, we derive design-level sensitivity projections for the axion-photon coupling $g_{aγγ}$. These projections should be interpreted as target sensitivities for the proposed cavity configuration, since the full-system birefringence noise and angular-jitter coupling remain to be measured.

Quantitative interpretation of atomic-resolution STEM-EELS requires dynamical simulation of the electron probe before and after core-loss transitions, which is computationally expensive. While the PRISM algorithm accelerates this by reusing scattering matrices, we introduce beam partitioning for both the probe-forming ($\mathcal{S}_1$) and detector-propagating ($\mathcal{S}_2$) PRISM matrices to further reduce computational and memory costs. Each matrix is calculated on a sparse set of parent beams and reconstructed via natural-neighbor interpolation locally at the ionized atom. A locality result demonstrates that the total error is governed entirely by this on-atom reconstruction error. The resulting BiP-PRISM algorithm removes per-scan exit wave propagation and significantly reduces memory requirements, enabling full-resolution elemental mapping, 4D cubes, and momentum-resolved qEELS on consumer-grade GPUs. We characterize the approximation's validity regime and demonstrate the simulation of a multimodal five-edge oxide-interface map and an FePt nanoparticle Fe-L map at 5x memory reduction, showing that the algorithm achieves high accuracy with significantly lower computational demands.

We propose the Long-Short-Range Neural Network (LSR-Net), an extensible operator-learning framework for predicting pattern dynamics on planar domains, spherical surfaces, and general manifolds. The method decomposes the forward evolution operator into a long-range component, represented by a compact Fourier multiplier constructed via the Sum-of-Exponentials (SOE) approximation, and a short-range component adapted to the underlying geometry and its intrinsic symmetries. For general manifolds represented by irregularly sampled point clouds, the long-range component is implemented by Gaussian gridding onto an auxiliary regular grid, where the Fourier multiplier is efficiently applied in k-space using FFT and the result is interpolated back to the original sample points. We evaluate LSR-Net on several benchmark systems, including the Allen-Cahn, Cahn-Hilliard, Schnakenberg, and Turing systems, over planar domains, spherical surfaces, and a blob-shaped manifold. Numerical results demonstrate that LSR-Net consistently achieves higher accuracy and improved stability compared with baseline operator-learning models. In particular, for Allen-Cahn dynamics on the sphere, the RMSE is reduced by approximately three orders of magnitude compared with the Spherical Fourier Neural Operator (SFNO). Rotation and reflection equivariance tests further confirm that the learned operator is consistent with these geometric transformations. These results indicate that LSR-Net provides an effective and robust approach for learning pattern dynamics on complex geometries.

The integration of artificial intelligence with physical modeling offers a transformative route for accelerating the design of active nanophotonic devices. Here, we present NanoPhotoNet-Lase, a physics-informed neural network (PINN) framework that embeds the electromagnetic and rate equations of lasing directly into its learning process to expedite the design of metasurface nanolasers. By coupling Maxwell's vector Helmholtz equation with the four-level population dynamics of dye gain media, the model achieves physics-guided prediction of optical responses, enabling rapid estimation of lasing thresholds across arbitrary nanostructure geometries and material configurations. Using high-index metasurfaces cavity, the NanoPhotoNet-Lase model identifies optimized geometries supporting quasi-bound states in the continuum (BICs) with strong confinement and high-quality factors. The predicted lasing was experimentally realized using Rhodamine B dye as gain medium. The measured lasing threshold (Pth = 565 uJ/cm2) and emission wavelength of 620 nm exhibited below 1% deviation from model predictions. Importantly, the framework enables design phase-gradient nanolaser metalens and beam steering that demonstrated coherent, directional, focused or steered emission. This work bridges physics-informed machine learning with experimental nanophotonics, establishing a scalable paradigm for real-time, physically interpretable design of coherent light-emitting metasurfaces.

We have shown a pattern that connects the refractive index, area and the cavity modes of the optomechanical crystals (OMCs) by the same order function. By keeping the fundamental and second cavity modes within a range of -+16 nm and -+23 nm we have shown the link between the design area of the OMC and the refractive index of the material, by keeping the design area same we have shown the link between the refractive index and the cavity mode wavelength and by keeping the refractive index the same, we have shown the link between the cavity mode wavelength and the design area. We have performed simulations for 2 different OMC designs and 10 different refractive indices (9 different materials) to prove the first two claims and we have performed both simulations and experiments on a 3C-SiC OMC, which resulted as 100 nm shift of the second cavity mode, to prove the last claim. Our findings prove that a universal design for optomechanical crystals is possible, making the transition to different material bases easier to exploit their specific properties, suggesting a path to commercialize such devices for hybrid quantum technologies and having flexibility of tuning such devices for their own relative applications.

Magneto-infrared spectroscopy under strong magnetic fields provides a powerful probe of Landau quantization and field-induced collective excitations, yet its full potential has long been constrained by the lack of in-situ polarization control, because the highly divergent infrared beam propagating through narrow light tubes undergoes multiple wall reflections, leading to severe polarization degradation. Here we report a collimated magneto-infrared spectroscopy system that integrates continuous in-situ polarimetry. The system employs incident and exit collimation chambers forming a Kepler type optical architecture, which converts the large-aperture FTIR output into a low-divergence beam and strongly suppresses multi-reflection trajectories inside long gold-plated light tubes, thereby enhancing both optical throughput and polarization fidelity. A remotely controlled polarization module, consisting of an automated linear polarizer and a switchable Fresnel rhomb positioned entirely outside the high-field region, enables continuous in-situ tuning between linear, circular, and arbitrary elliptical polarization states without thermal cycling, manual realignment, or breaking vacuum. Interchangeable compact focusing modules further support Faraday and Voigt geometries in both transmission and reflection experiments within a 50 mm magnet bore, providing efficient beam focusing and signal collection while maintaining polarization fidelity. The setup achieves a minimum root-mean-square noise of 0.0033%, an average noise of 0.0082%, and a linear polarization extinction ratio up to 40:1. We demonstrate the capability through continuous in-situ linear polarimetry and broadband circular polarimetry in the magneto-infrared spectroscopy of various single crystals. This platform establishes a robust experimental framework for in-situ polarization-resolved magneto-infrared spectroscopy.

Self-selected phase-matched second harmonic generation is introduced as an all-optical probe of refractive-index dispersion in birefringent nonlinear optical materials. Rather than requiring wavelength or angular tuning, the exposure with a spectrally broad, intense ultrashort pulse allows the material to self-select the fundamental spectral component that satisfies the type-I noncritical phase-matching condition. This produces a narrow peak in the second harmonic spectrum whose position is governed by the refractive indices and is therefore highly sensitive to material parameters that affect the optical dispersion. We demonstrate the application of this phenomenon for the optical inspection of stoichiometry and temperature gradients in technologically relevant lithium niobate, as well as composition inhomogeneities in newly grown lithium niobate-tantalate solid solutions. These results establish self-selected phase-matched second harmonic generation as a rapid, non-contact method for inspecting nonlinear optical materials, with potential relevance for bulk crystals, wafers, and thin-film platforms.

We address two distinct nonlinear propagation problems in nonlinear optical waveguide arrays (WGAs) with non-reciprocal (non-Hermitian) couplings. First, we investigate the light propagation launched by initial excitations of two different types. The single-channel excitation creates stable solitons supported by the interplay of the Kerr nonlinearity and non-Hermitian skin effect (NHSE). In this case, we derive, by means of the symbolic-regression method, an analytical formula defining the soliton existence boundary. For the broad-pulse excitation, we produce perturbed soliton solutions analytically in the continuum approximation, which is accurately corroborated by numerical results. We thus conclude that NHSE accelerates the propagation of the broad soliton towards the boundary, ultimately causing tight localization at the edge, which is a hallmark of the NHSE in the continuum limit. Second, we identify stationary solitons in the system -- specifically, nonlinear bulk modes in the Hermitian regime and near-edge skin solitons in the non-Hermitian one. The nonlinear bulk modes are compressed toward the edge of the WGA under the action of the non-reciprocality, which is the nonlinear extension of NHSE.

We analyze the relationship between open-accessibility in data, code, and paper text in astrophysics using a sample of 53,194 peer reviewed papers published between January 2021 and April 2025, drawn from NASA's Astrophysics Data System (ADS). We measure eleven quantities: open accessibility of text, open-code status, open-data status, number of grants received, code size, programming language, data repository size, citation count, number of authors, paper length, and publication date. We break down citation advantages based on six astrophysical sub-fields: Solar System, Planet, Stellar, ISM, High Energy, and Galaxies+Cosmology, determined by keywords. This is accomplished by tuning a multivariate least-squares regression model with alongside partial correlations and non-parametric tests to isolate the contribution of each facet of openness. After controlling for the aforementioned quantities, we find significant citation advantages associated with all three forms of openness: open data (+32%, p < 10^-24), open access (+26%, p < 10^-67), and open code (+16%, p = 0.003). The open-data citation advantage is present in all six sub-fields, and especially in Galaxies+Cosmology and ISM, which have the strongest cultures of sharing simulation outputs and observational data products. Open-code and open-data sharing rates are highest in Galaxies+Cosmology and HEA (~0.9% and ~2.9%), reflecting their more developed community data infrastructure, and lowest in Solar System and ISM, where data is distributed on platforms not taken into account by this study. Our findings support the long held notion that public access comes with concrete personal incentives for authors in terms of citations.

We propose and analyze a nonstandard finite difference (NSFD) scheme for nonlinear parabolic equations involving a p-Laplacian-type diffusion operator in one- and two-dimensional spatial domains. Following Mickens' design principles, the proposed discretization employs a nonlinear denominator function phi(.) together with a nonlocal approximation of the nonlinear diffusion term Delta_p, yielding a structure-preserving discrete model. The scheme is designed to retain key qualitative properties of the continuous problem, including positivity, boundedness, and stability, which may be lost by standard finite difference methods (FDMs). We establish the well-posedness of the continuous model, derive the NSFD scheme, and investigate its consistency, convergence, and local truncation error. Numerical experiments confirm the theoretical results and demonstrate that, unlike the standard explicit FDM, the proposed NSFD scheme avoids spurious oscillations and nonphysical negative solutions even for relatively large time-step sizes.

This work reports the spatial-spectral engineering and finite-size quantization of optical modes within a triangular-lattice silicon nitride membrane metasurface. Truncating the lattice into a finite square cavity breaks translational symmetry and lifts modal degeneracy, splitting optical modes into discrete cavity-envelope sub-modes. High-resolution photoluminescence (PL) scanning reveals distinct spatial field distributions. The corner-localized sub-mode features the highest Q-factor due to multipolar far-field destructive interference, whereas the core-localized sub-mode exhibits strong radiative coupling. PL mapping reveals a symmetric, four-fold clover-like wavelength arrangement. These results demonstrate that boundary-induced deterministic symmetry can override underlying lattice characteristics, offering a robust strategy for precise spatial-spectral tailoring of light-matter interactions at the nanoscale.

We present a theoretical framework for spontaneous Raman scattering that fundamentally bridges quantum-mechanical and nonlinear-optical approaches. By conceptualizing spontaneous Raman scattering as a stimulated Raman gain or loss event seeded by the quantum vacuum field, we rigorously derive the spontaneous Raman cross-section directly from the third-order nonlinear susceptibility. Crucially, this framework predicts the existence of a hitherto unrecognized phenomenon: "spontaneous Raman loss" (sRL), which acts as the vacuum-seeded counterpart to stimulated Raman loss, complementing traditional spontaneous Raman scattering (spontaneous Raman gain, sRG). Furthermore, we establish a rigorous connection to the traditional Kramers-Heisenberg-Dirac (KHD) theory, revealing that the spontaneous process is governed by interference before a detector between the signal field emitted from molecules and the vacuum field itself that stimulates the molecules. This insight uncovers a direct correspondence between the sRG susceptibility and the rotating/counter-rotating interference terms in the KHD formula. Ultimately, we extend the foundational KHD theory by incorporating previously unrecognized essential terms, achieving perfect analytical agreement between the quantum mechanical and nonlinear optical descriptions of Raman scattering.

Spin correlations are among the most fundamental quantum observables in many-body systems, yet they remain difficult to access experimentally in relativistic heavy-ion collisions. Existing spin measurements, including hyperon polarization and vector-meson spin alignment, have revealed important single-particle spin phenomena, but genuine two-particle spin correlations in the produced hadronic system remain largely unexplored. Here we propose spin femtoscopy, a framework for accessing genuine two-particle spin correlations through spin-resolved femtoscopic measurements. The key principle is that different two-particle spin configurations can give rise to different femtoscopic correlation functions because of quantum statistics, spin-dependent final-state interactions. Using $ΛΛ$ pairs as a proof of principle, we exploit the self-analyzing weak decay of $Λ$ hyperons to construct spin-sensitive femtoscopic correlation functions with different singlet and triplet admixtures. We show that these observables provide experimental access to the spin-state populations of the pair and allow genuine spin correlations to be separated from spin-dependent femtoscopic mixing caused by quantum statistics and final-state interactions. This work extends femtoscopy from a probe of source geometry and final-state interactions to a framework for revealing the quantum spin structure of strongly interacting matter.

A photon entering a plane parallel scattering slab performs a random walk and eventually escapes through one of the two faces or is absorbed. The scattering distribution is a Henyey Greenstein phase function and the step length distribution is exponential. The central result of this paper is that the reflectance, transmittance, absorptance, and emergent angular distributions can all be expressed in terms of the first passage statistics of the walk. Two approaches are used. In the Monte Carlo MC approach, an extremely long random walk with many steps is efficiently generated without regard to any boundaries. The intersection of this walk with a large collection of target objects creates an ensemble of excursions of the objects. The MC approach relies explicitly on the memoryless property of the exponential distribution so that the portion of the first and last steps inside the object follow the same length distribution as the walk steps. The details of each excursion are recorded and any statistics can be extracted, to the sampling precision, from the database of excursions. In particular, first passage statistics are extracted from this ensemble. In this work the objects are slabs with different positions and thicknesses. In the radiative transfer RT approach the slab is divided into thin layers with scattering treated to first order in each layer. The RT equations are then directly integrated over the slab to give the desired first passage statistics. In the RT approach reflection, transmission and absorption are found to the precision of the RT solver. The two methods agree to the precision of the MC over the tested range of random walk parameters.

X-ray free electron laser (XFEL) experiments using self-amplified spontaneous emission (SASE) pulses typically achieve temporal resolutions of order several femtoseconds, as the pulse duration puts a practical limit to pump-probe or probe-probe schemes. Even with the emerging capabilities to generate pulses with attosecond durations with new single-spike SASE schemes, direct access to attosecond electron dynamics remains an experimental challenge. Here we show how X-ray coherent attosecond pulse-pair spectroscopy (X-CAPPS) provides a powerful new approach to access the ultrashort time-delay window. Coherent attosecond pulse pairs with time delays varying from ~500 as to ~5 fs are generated with Cu K$α_1$ stimulated X-ray emission from a gain medium pumped by intense SASE XFEL pulses. These pulse pairs are analyzed with two subsequent Bragg crystal spectrometers, and the resulting interference spectrum is captured on two sequential 2D image detectors encoding their time separation, relative amplitudes, and phases with high precision. X-CAPPS requires no split-and-delay X-ray optics, nor XFEL pulse modifications, making it broadly implementable across existing facilities. This technique enables the investigation of attosecond processes with $\mathring{A}$ngström resolution, providing a new tool for probing ultrafast dynamics across a wide range of atomic, molecular, and solid systems.

The fidelity of a quantum gate is sensitive to small deviations in the physical control parameters. Unfortunately, it is generally difficult to exactly model the implemented Hamiltonian for a set of user-defined parameters, necessitating on-device calibration. Here, we present an active learning framework based on Bayesian optimization with a Gaussian Process surrogate to find the optimal parameter set. We validate the technique through numerical calibration of the laser amplitude and frequencies that implement the trapped-ion Mølmer Sørensen gate. We show that a Gaussian process can model the Hamiltonian dynamics. The addition of active learning accelerates the discovery of the optimal parameter set with speed and final fidelity dependent on the quantum projection noise of the data. These results establish the utility of active learning and surrogate models for quantum calibration and control.

Dynamic metasurface antennas (DMAs) enable programmable wave-domain signal processing that can be jointly optimized with downstream digital processing in an end-to-end manner. Existing studies, however, typically assume ideal analog-to-digital conversion (ADC) and often rely on simplified electromagnetic models. Here, we study ADC-aware end-to-end optimization of a monostatic sensing pipeline based on a DMA with strong mutual coupling (MC). We model the wave domain using an MC-aware multiport-network model whose parameters were experimentally estimated for a fabricated chaotic-cavity-backed DMA with 96 one-bit-programmable meta-elements. We perform ADC-aware end-to-end optimization of the DMA configurations and digital classifier, either with awareness of a fixed uniform ADC or, optionally, with jointly learned ADC decision thresholds, and compare against baselines that assume an ideal ADC and/or ignore MC. Our results show that ADC awareness is essential in low-resolution ADC regimes: with one-bit ADCs and eight DMA configurations, deploying an ideal-ADC-trained system with a uniform one-bit ADC reduces the test accuracy from 95.5% to 56.0%, whereas ADC-aware training with the same fixed uniform one-bit ADC achieves 87.2%. We also show that without MC awareness the accuracy drops to the random-guess level. Learning non-uniform ADC thresholds provides at most modest additional gains over fixed uniform ADCs in the considered DMA-based sensing pipeline.

Corruption is embedded in networks of access, coordination, and protection, yet little is known about how gender shapes actors' positions within them. This article examines whether corruption networks in Colombia's territorial press reproduce gendered patterns of exclusion. Drawing on an access-to-power perspective, we argue that women's lower presence may reflect unequal incorporation into the spaces where exchanges are organized. Empirically, we use Transparencia por Colombia's Radiografía de Hechos de Corrupción, integrating case- and actor-level information to build co-participation networks. We analyze gender differences in composition, position, recurrence, and institutional access to resources. Results show that these networks are strongly masculinized: men dominate actors and ties, and women appear less often among recurrent actors. However, women are not absent from connected or dense areas, suggesting uneven rather than absolute exclusion. The findings shift attention from whether women are ''less corrupt'' to how unequal access to power structures participation in corruption networks.

We propose the arbitrary Lagrangian-Eulerian (ALE) form of the Galerkin-Boltzmann formulation for the simulation of nearly incompressible flows with moving boundaries. The continuous Boltzmann equations are mapped to a reference state to compensate the mesh motion with an advection term. The resulting system is discretized in space using the discontinuous Galerkin method on unstructured meshes. A semi-analytic Runge-Kutta time discretization is used to overcome the stiffness introduced by the continuous Boltzmann equations. The well-known geometric conservation law is shown to be satisfied by the time and space discretizations and consistent update of geometric factors of the discretization. The implementation is on the GPU accelerated kernel library libParanumal and validated by a free stream preservation and moving Taylor-Green vortex test cases. Then, the capabilities are shown using a plunging symmetric airfoil in two-dimensions and moving carangiform fish in three-dimensions using perfectly matched layers.

\emph{Background:} Standard numerical methods accurately simulate seismic waves but are computationally expensive, particularly for inverse problems. Machine learning approaches have been proposed as alternatives that can reduce computational cost while maintaining acceptable physical accuracy. \emph{Objective:} To map how physics-informed machine learning methods have been applied to seismic wave propagation modeling based on partial differential equations. \emph{Methods:} A scoping review was conducted using the OpenAlex and Scopus databases. Selected studies were classified by problem type (forward or inverse) and machine learning strategy to identify research trends, methodological patterns, and gaps in the literature. \emph{Results:} Physics-informed machine learning has been applied to both forward modeling and inversion in seismology, often reaching accuracy comparable to standard numerical methods at lower computational cost. Application of three mechanisms for incorporating physical knowledge were identified: observational bias, inductive bias, and learning bias. To evaluate methodological reproducibility of a representative method, the original PINN framework was replicated in PyTorch, obtaining results consistent with and in most cases more accurate than those originally reported. From the reviewed literature, limitations remain in benchmarking consistency, training cost, and scalability to three-dimensional and experimentally validated problems. \emph{Conclusions:} Standard numerical methods remain the basis of seismological workflows, while physics-informed machine learning offers complementary approaches that are useful for inverse problems and surrogate modeling. Future work should focus on consistent benchmarking, hybrid formulations, and validation under realistic geophysical conditions.

The Spiral Phase Transform (SPT) is a generalization of the Hilbert transform for 2D signals and, as such, can be used for AC signal demodulation. However, phase demodulation with the SPT is complicated by a multiplicative term that depends on the fringe directional map. We derived an analytical formula for the twofold directional map and applied the SPT for the blind reconstruction of phase-shifted digital holograms. Possible phase ambiguities in the unfolded directional map were resolved by satisfying the spatial uniformity condition of the phase shifts. The method was experimentally verified using on-axis and off-axis digital holograms of specularly reflecting subjects.

High-order accurate simulations of special relativistic hydrodynamics (RHD) are prone to numerical breakdown if intrinsic physical constraints (positive rest-mass density/pressure and subluminal velocity) are violated near strong discontinuities. In this work, we develop a robust and efficient physical-constraint-preserving (PCP) flux-limiting framework for high-order schemes, using finite-difference WENO as a representative example. By leveraging the geometric quasilinearization (GQL) representation, which equivalently reformulates the nonlinear RHD constraints into a family of linear inequalities, we integrate a Zalesak-type Flux-Corrected Transport (FCT) update into a scalar-style limiter that acts directly on conservative variables. A critical innovation is the explicit, non-iterative determination of limiting parameters via a rational stereographic parameterization of the GQL normal vector. This technique transforms the required worst-case minimization over auxiliary variables into a generalized Rayleigh-quotient formulation, allowing the optimal parameters to be obtained by solving small symmetric eigenvalue problems ($2\times2$ in 1D; $(d+1)\times(d+1)$ in $d$ dimensions). Relaxed variants are further introduced to reduce computational costs in multidimensions while retaining the PCP guarantee. Extensive numerical benchmarks ranging from 1D to 3D, including ultra-relativistic Riemann problems and astrophysical jets, demonstrate that the proposed method robustly enforces physical admissibility, sharply resolves discontinuities, and maintains design-order accuracy for smooth solutions.

Accurate aerodynamic modeling of satellites in very low Earth orbit (VLEO) requires gas-surface interaction (GSI) models that capture the full velocity spectrum from thermal to orbital speeds. Atmospheric particles initially strike spacecraft surfaces at hypersonic velocities of 6 000 - 10 000 m/s. Due to surface roughness and complex geometries, especially within air-breathing electric propulsion (ABEP) intake systems, multiple collisions occur, progressively reducing the particle velocities. A recent machine learning framework for deriving scattering kernels from molecular dynamics (MD) simulations has shown promise, but remains limited to high-velocity single impacts and possibly violates fundamental equilibrium principles such as detailed balance. This work extends this machine learning based scattering kernel to cover the complete velocity range using conditional normalizing flows trained with physics-informed constraints, enabling accurate modeling of multi-bounce scenarios in realistic VLEO applications. We train a conditional Real-valued Non-Volume Preserving (cRealNVP) model on expanded molecular dynamics simulations covering velocities from thermal to hypersonic speeds, incorporating a detailed balance loss term. The resulting model demonstrates improved accuracy compared to previous approaches even in the original high-velocity regime, while successfully capturing thermal-velocity scattering. Quantitative assessment shows that thermalization is approximated within acceptable tolerances. This framework provides essential capabilities for accurate ABEP intake optimization and VLEO mission planning while offering a general methodology applicable to broader rarefied gas dynamics problems requiring thermodynamic consistency.

Many bipartite social networks exhibit pronounced asymmetries in selectivity and matching opportunities: members of one side can afford to be highly selective, while members of the opposite side are forced to accept less desirable matches. While it is natural to try to explain this asymmetry in terms of the intrinsic characteristics of the two sides or other exogenous factors, here we show that such asymmetries can also emerge endogenously through a feedback process generated by the matching process itself: as one side becomes more selective, the other side is pushed to be less selective due to reduced matching opportunities, and vice versa. We develop a model in which individuals repeatedly form one-to-one matches across two groups and adapt their selectivity to achieve a target matching rate. Using both analytic and numerical methods, we show that when encounters are sufficiently frequent, the unique equilibrium is for one group to be highly selective and the other non-selective. This qualitative outcome holds even for heterogeneous groups with overlapping, almost indistinguishable distributions of target matching rates. The model makes several testable predictions, and it provides a mechanism for behavioral differentiation in repeated matching environments, with applications ranging from online dating to hiring and housing markets.

Understanding how information propagation affects epidemic dynamics has become an emerging topic of interest. However, the influence of interpersonal relationship heterogeneity on information acquisition and disease transmission has been largely overlooked. In this work, we introduce a hypergraph structure for Cyber-Physical Systems (CPSs) with two distinct layers. The upper layer, referred to as the cyber layer, consists of a mixed hypergraph, capturing both pairwise propagation and higher-order diffusion of epidemic-related information. The lower layer, referred to as the physical layer, employs a Susceptible-Infected-Susceptible (SIS) process to capture epidemic spreading. This work introduces an adaptive perception-protection mechanism based on Jaccard similarity, which accounts for interpersonal heterogeneity. In this mechanism, individuals receive information based on their relationships with neighbors and take protective measures accordingly. We analyze the impact of interpersonal relationships and the adoption of neighborhood-based self-protection strategies on epidemic dynamics. Furthermore, we conduct a theoretical analysis based on the Microscopic Markov Chain Approach (MMCA), analytically derive the outbreak threshold, and confirm the results with extensive Monte Carlo (MC) simulations. The results show that stronger interpersonal relationships can promote information propagation, significantly increase the threshold for epidemic outbreaks, and effectively suppress the scale of the epidemic. The study provides theoretical support for designing epidemic control strategies considering interpersonal heterogeneity and improves the understanding of epidemic spreading on hypergraphs.

We propose PIFLUX, a single-molecule localization scheme combining deep-subwavelength plasmonic illumination with widefield detection. Interference between counter-propagating gap plasmons and a normally incident optical field generates an illumination pattern whose position can be tuned through the plasmon phase while preserving its spatial period. A Cramér-Rao analysis shows PIFLUX reaches few-nanometer precision matching MINFLUX while doubling that of SIMFLUX over a micrometer field of view, and a maximum-likelihood estimator confirms this on a synthetic nuclear pore complex.

Passive acoustic sensing is an attractive modality for counter-unmanned aerial system (counter-UAS) defence: it is covert, low-cost, and effective against drones with small radar cross-sections or minimal radio emissions. We present EchoHawk, an open and fully reproducible reference pipeline that detects a drone from its rotor harmonics, estimates its blade-passing frequency, and localises it with a microphone array via classical wideband beamforming (delay-and-sum, MVDR, MUSIC) and time-delay processing (GCC-PHAT, SRP-PHAT), followed by temporal tracking. We evaluate the system on a physically transparent synthetic benchmark that pits drones against hard low-frequency harmonic confusers, such as ground vehicles, and on real recorded audio. Our central methodological contribution is a documented case of session-level data leakage in a widely used public dataset: because its recordings are pre-segmented into short clips, naive clip-level splits place adjacent slices of the same continuous recording in both training and test sets, inflating reported performance. Enforcing recording-session-grouped cross-validation reduces, for example, a random-forest baseline's detection probability at a 1% false-alarm rate from 0.796 to 0.745, yielding honest numbers. All code, figures, and a synthetic data generator are released so that every result runs without any download.

Estimating peak prevalence is a central problem in epidemic modeling because it determines the period of greatest infectious burden and is closely linked to health-care demand. In multistage SIR models, however, peak prevalence is generally less tractable than in the classical model with exponentially distributed infectious periods. Motivated by the use of weighted infectious-stage aggregates as surrogates for prevalence, we investigate the relationship between the prevalence peak and the maximum of a weighted stage functional in deterministic SI$(k)$R epidemic models. We show that this relationship depends critically on how the stage-progression rate is scaled as the number of infectious stages increases. Under naive scaling, in which the progression rate remains fixed, the weighted peak is asymptotically equivalent to the prevalence peak and the commonly used factor-two approximation fails. Under Erlang scaling, which preserves the mean infectious period, the multistage model converges to a delay formulation in which prevalence and the weighted stage functional become unweighted and triangularly weighted moving averages of incidence. This limiting representation provides a theoretical basis for the factor-two approximation and identifies the regimes in which it is accurate. It also explains why this approximation deteriorates as epidemic waves become more sharply peaked. We derive analytical error bounds and develop curvature-based and parameter-based corrections that substantially improve accuracy. Numerical studies confirm these improvements across a broad range of epidemiological parameters. Overall, the results show when weighted-stage peaks can be used reliably as proxies for peak prevalence and how the resulting estimates can be refined when the standard approximation loses accuracy.

Brain activity spans single-neuron, population, and network levels, and core questions in neural coding require moving between them. Yet current tools target a single paradigm and incompatible data formats, leaving cross-level questions hard to address. We present DRIADA, an open-source Python framework that unifies neural signals and time-aligned behavior in a shared data model, so selectivity testing, dimensionality reduction, and network analysis operate within a unified workflow. We evaluate it on synthetic data with known ground truth, hippocampal calcium imaging from 13~mice in an open field, and a simulated toroidal attractor network. In the hippocampal data, selectivity-based filtering restored a two-dimensional spatial embedding from a collapsed all-neuron embedding, while reverse analysis showed that ${\sim}57\%$ of neurons informative about leading manifold dimensions were not selective to any of the 11 measured behavioral features. On the toroidal benchmark, four independent modules recovered the expected topology. DRIADA makes cross-scale analysis routine across calcium imaging, spike trains, and simulated networks.

In multispecies birth-death processes, how population regulation -- through suppressed replication, elevated mortality, or both -- affects macroscopic stochastic dynamics has escaped detailed analysis. Here, we show that the distribution of regulation mechanisms can be invisible in deterministic or mean-field dynamics but play a significant role in the diffusive evolution of population frequencies. By introducing a tunable regulation partitioning parameter $α_i$ and projecting a $d$-species birth-death process onto a $(d{-}1)$-dimensional Moran process, we find a regulation-mechanism-dependent diffusion tensor. For the simple two-species case, we derive exact fixation times and probabilities to show how different regulation mechanisms stochastically favors a more birth-regulated species, even under complete deterministic neutrality. Our model also allows us to define an $α$-dependent effective population size $N_{\rm e}(α)$ among neutral species, generalizing its classical interpretation. For near-neutral populations or populations that are heterogeneous in their regulation mechanism, we used perturbation theory to calculate the spectral gap, identifying it with a diversity loss timescale which can also be interpreted as setting an effective population size. Our results are particularly applicable to interacting subpopulations of T cells ("clones") which are near-neutral, are regulated through proliferation and apoptosis, and lose diversity with time.

This work addresses an optimal control problem for a SI epidemic model incorporating heterogeneities in resistance and viral load at the population level. Building upon the heterogeneous SI framework developed in [1], a minimization problem constrained to the macroscopic counterpart of the SI dynamics derived therein is proposed. Unlike traditional optimal control problems in homogeneous epidemic models, the present approach focuses on an optimal control problem that accounts for population heterogeneity, offering insights from a microscale perspective. The contribution aims to minimize the final size of the infection within a finite time horizon by developing a pharmaceutical strategy, under a supply constraint that translates into an integral equality constraint in the control function. By applying the Pontryagin Minimum Principle, a characterization of an optimal control is provided.

Colored sectors in a microbial range expansion encode more than lineage survival counts. We formulate a computer-vision inverse problem: from one endpoint image of an accretive multi-type expansion, recover the radius-indexed pairwise boundary-flow field and test whether the visual pattern is compatible with a transitive scalar fitness hierarchy. The observable is a geometric signal extracted from sector-boundary curves in log-polar coordinates. We prove endpoint observability and stability for frozen fronts, weighted transitive/cyclic decomposition, contact-complete circular design, physical-clock and mechanism non-identifiability, exact Gaussian cyclicity testing, and Bonferroni-valid interval scanning. The benchmark is deterministic: analytic endpoint images, blurred/noisy pixel round trips, scalar-null stress tests, public-image tracing, multi-resolution mechanistic endpoints, and a non-learning frozen-front simulator. The implementation recovers pairwise edge-flow histories from endpoint images, detects cyclic residuals in a mechanistic four-type expansion, and uses those residuals as forcing signals for a dimensionless active design-control layer covering reaction-diffusion control, phenotype-frontier optimization, protocol synthesis, Monte Carlo robustness, and a downstream population-state bridge.

Insecticide-treated nets (ITN) are an effective and low-cost intervention for controlling vector-borne disease (VBD), however, their use depends on individual decisions based on perceived cost and risk of infection. This study investigates a nonlinear multi-host model for the transmission of VBD with endogenous strategic control. We assume that hosts' adoption of ITN emerges from the payoff-based decision-making, creating a nonlinear coupling with disease prevalence. We model vector preference as a function of ITN coverage to probe the complex interplay among individual choices, disease prevalence, and its control in a multi-host setting. The qualitative behavior of the system is characterized by the thresholds $R_0$ and $R_c$, which determine the existence and local stability of the disease-free and endemic equilibria. The system exhibits rich dynamical behavior; hence, we provide a bifurcation analysis identifying the conditions for saddle-node and Hopf bifurcations. Our results demonstrate that the interaction between the perceived cost of ITN and the infection risk can induce critical transitions, including regime shift from stable endemic states to sustained periodic oscillations. Furthermore, we identify a counterintuitive effect whereby complete ITN adoption by the primary host can increase the overall prevalence in the secondary host due to adaptive shifts of vector feeding behavior.

Evolutionary game theory provides a framework by which to study the emergence of cooperation in a population of self-interested actors. In such a framework, players' decisions on whether or not to cooperate evolve according to decision rules called population dynamics. However, often games are studied under the assumption that all individuals play under the same conditions, and many common choices of update rule are not well suited for a heterogeneous population. In this paper, we categorise and compare four different population dynamics in such a population as ``extrinsic'', where players learn by looking outward at the payoffs of other players, and ``intrinsic'', where players look inwardly at their own attributes or potential payoffs. We show that extrinsic population dynamics admit a ceiling on the rate of cooperation which can be exceeded by intrinsic population dynamics, and demonstrate this using the public goods game with heterogeneous contributions.

The Evolvable Soma Theory of Ageing is a recently proposed model that frames development as a continuous process of change accompanying organisms throughout the lifespan. This process is driven by developmental genes which encode epigenetic changes on target cells, whereas ageing reflects the expression of late-acting modifications, that are subject to ongoing evolutionary optimisation and function as somatic "experiments" to explore phenotypic novelty. In this work we examine the role of transposable elements in the model. Our proposal acknowledges that these elements facilitate the expansion and diversification of gene regulatory networks by providing transcription factor binding sites. To minimise disruption, their regulatory activity is tightly repressed by epigenetic mechanisms during early development, which may be progressively released by genetically driven, age-associated epigenetic changes in later life, thereby contributing to transcriptional pseudo-randomness and ageing-associated phenotypes. Within this framework, transposable elements are integrated into a unified view of evolution, development and ageing, providing a conceptual basis for their dual role in regulatory innovation and age-related decline.

Data analysis in the medical domain often encounters scenarios involving a limited target dataset and a large, unannotated dataset with a general distribution. Under such circumstances, self-supervised learning (SSL) methods are highly effective for utilizing large datasets, making them a popular choice for electrocardiogram (ECG) analysis. This work presents the Event Reconstruction Joint-Embedding Predictive Architecture (ER-JEPA), a lightweight SSL framework for multivariate time series, whose name and two-fold hierarchical structure are inspired by the diagnostic approach of cardiologists. At its core, ER-JEPA features: (1) a two-stage structure that constructs representations for each time interval and subsequently processes these representations as a univariate time series, (2) the hierarchical integration of two Joint-Embedding Predictive Architectures (JEPAs), and (3) a Vision Transformer (ViT) backbone. The structural concatenation of two JEPAs categorizes the model as a Hierarchical JEPA (H-JEPA), designed to encode multiple levels of abstract representations for enhanced prediction on complex tasks. This study reports a successful application of H-JEPA to 12-lead ECG data as a multivariate time series alongside an analysis of the sensitivity of hierarchical representation during the pretraining stage. Pretrained on approximately 180,000 10-second recordings, the model achieves state-of-the-art downstream performance on the ST-MEM benchmark, with rapid computation and minimal resource usage.

Integrated sensing and communication (ISAC) has emerged as a key technology for future wireless networks by enabling communication and environmental sensing through a common waveform and hardware platform. Among the candidate waveforms for ISAC, Affine Frequency Division Multiplexing (AFDM) had attracted significant attention due to its robustness in high-mobility environments, but it suffers from a high peak-to-average power ratio (PAPR). In this paper, we propose a sensing-aware chirp-subcarrier reservation (CSR) framework that reduces PAPR while improving ranging performance. The proposed method combines low-complexity gradient-based PAPR minimization with a randomized local search that exploits the phase sensitivity of the AFDM autocorrelation function to suppress delay low-ambiguity-zone (LAZ) sidelobes. Numerical results show that the proposed scheme achieves significant PAPR reduction together with significant sidelobe suppression, resulting in improved weak-target detection performance.

Short-horizon prediction is essential for electro-optical UAV tracking, especially when the target is small, maneuvering, or intermittently observed. Image center, line-of-sight, and range measurements provide direct constraints on target position, but their constraints on acceleration are weak. As a result, prediction can lag during aggressive maneuvers. This paper proposes an image-domain tilt constrained distributed fusion method for maneuvering UAV tracking. The method uses the apparent roll and pitch of a rotorcraft target in the image as low-level maneuver cues. A weak-prior auto-labeling pipeline first generates oriented bounding box and image-domain tilt labels from synchronized video, gimbal IMU, and UAV IMU data. A YOLO-OBB detector is then trained to provide online target position and tilt measurements. The front-end Python implementation is publicly available at github.com/ShineMinxing/PythonYOLO. In the fusion stage, the UAV state is modeled by position, velocity, and acceleration. Image-domain roll and pitch are introduced as acceleration-related pseudo-observations. For distributed tracking, one mobile gimbal camera and two fixed ground cameras are fused asynchronously. Camera attitude error states are augmented into the filter to absorb extrinsic drift and cross-camera systematic inconsistency. A Mahalanobis gate with time-since-last-valid covariance widening is used to reject false detections and handle dropouts. In simulation, adding roll/pitch observations reduces the prediction RMSE from 1.991 m to 0.821 m and decreases the cumulative prediction error by 60.75\%. In real distributed experiments, a self-consistency evaluation shows an 18.10\% reduction in cumulative prediction error. The results show that image-domain tilt can provide useful acceleration constraints for robust short-horizon UAV prediction.

Forthcoming Low Earth Orbit (LEO) satellite networks such as Starlink's Mobile Satellite Service (MSS) will incorporate the New Radio (NR) Non-Terrestrial Network (NTN) standard. The Synchronization Signal Block (SSB) specified as part of NR is periodically broadcast for cell search and initial access. We propose to exploit the SSB for opportunistic receiver positioning. Doppler shift measurements are modeled and pseudoranges are derived from SSB while also taking into account the receiver's clock bias and drift. The resulting per satellite integer ambiguity in the pseudorange is resolved by geometry alone, without inter-satellite differencing or an a-priori position. Measurements are taken from SSBs of multiple satellites and at multiple occasions per satellite, whereby the SSBs are subject to different transmission timings and varying propagation delays. Finally, a simulation model is developed for positioning based on the actual Starlink constellation and the NR NTN standard to evaluate the positioning accuracy to be expected. The proposed approach achieves a mean positioning error of less than 10m without requiring any modification of the NR NTN standard.

Microwave linear analog computers (MiLACs) have recently gained attention for future gigantic multiple-input multiple-output (MIMO) systems by enabling beamforming with greatly reduced hardware and computational cost. However, channel estimation for MiLAC-aided multiuser systems remains an open problem. Conventional channel estimation requires many radio-frequency (RF) chains to access full-dimensional received signals, followed by massive digital processing, which undermines the advantages of MiLAC-aided systems in reducing the number of RF chains and computational complexity. In this paper, we propose computationally efficient channel estimation and beamforming schemes for MiLAC-aided multiuser multiple-input single-output (MU-MISO) systems with a limited number of RF chains. We consider the general case where different user groups experience different channel correlation matrices. By exploiting the rank deficiency of these matrices, the proposed schemes use MiLAC to compress the full-dimensional received signals in the analog domain, making them compatible with the available RF chains while preserving the essential channel information. Then, in the digital domain, only low-dimensional channel estimation is performed based on these compressed observations, substantially reducing computational cost. We further show how regularized zero-forcing beamforming (R-ZFBF) can be efficiently realized from the low-dimensional channel estimates through a cascade of two MiLACs, which offers greater computational flexibility than a single MiLAC. Numerical results show that the proposed schemes reduce computational complexity up to $1540\times$ and $16108\times$, for channel estimation and beamforming, respectively, while achieving performance comparable to digital baselines.

Sensing-aided beam tracking is a promising approach to reduce the overhead for millimeter-wave beam management. However, real-world application remains challenging due to rapid channel variations and substantial environmental differences across deployment scenarios. Developing low-complexity sensing assisted approaches that generalize to diverse environments can alleviate the problem. With this motivation, this paper proposes a lightweight vision-aided model for cross-environment beam tracking. The task is formulated as a sequence-to-sequence classification problem, where the model jointly predicts the current and future optimal beams from past visual observations. We develop a low-complexity model based on depthwise separable convolutions and introduce hierarchical data augmentation and beam power-based label smoothing to improve robustness and generalization. Experimental results on real-world images from two geometrically distinct DeepSense 6G scenarios show that the proposed strategies consistently improve cross-environment beam prediction accuracy up to 84% across the current and three future time slots, outperforming the state-of-the-art solution. Notably, this performance is achieved while reducing the number of model parameters and computational complexity by factors of approximately 52 and 79, respectively, compared with the high-capacity ResNet baseline.

This paper studies the stochastic performance of a downlink multiple-input multiple-output integrated sensing and communication (ISAC) system over Rician fading channels. Rician fading is important in line-of-sight (LoS)-dominated deployments, where a deterministic propagation component can strongly affect sensing and communication reliability. The base station (BS) simultaneously serves a user and senses a target. The BS-user channel contains LoS and non-line-of-sight components. The user LoS angle may be fixed or random, and the target angle may follow an arbitrary distribution potentially correlated with the user angle. Compared with Rayleigh fading, the deterministic LoS component introduces angle-dependent terms and leads to generally independent but non-identically distributed random vectors, requiring new analysis. We analyze two beamforming strategies: subspace joint beamforming (SJB), optimal for the shared waveform structure, and linear beamforming (LB), a practical alternative using separate sensing and communication beamformers. For both schemes, we derive communication outage probability (OP) and sensing OP based on the Cramer--Rao bound (CRB). We also identify special cases with simpler expressions. For LB, we derive upper and lower bounds on sensing OP and a tractable approximation. We characterize large-system and high-power scaling laws. LB without dirty paper coding (DPC) is interference-limited at high power due to radar self-interference. Results show the Rician K-factor affects communication more strongly than sensing, with non-monotonic behavior across regimes. LB with DPC achieves the best overall performance in strong LoS environments and is the only scheme achieving ultra-high communication reliability in Rayleigh fading, while SJB provides a robust lower-complexity alternative across operating conditions.

This paper examines the use of radiofrequency (RF) channels for hemodynamic monitoring in cardiac pacemakers. It analyzes RF signal variations between intracardiac transceivers in the right ventricle (RV) and right atrium (RA), as well as subcutaneous receivers, to determine their correlation with cardiac dynamics. The study shows that temporal RF signal variations closely align with cardiac rhythm, allowing for the estimation of parameters such as chamber volume, valve behavior, and pressure changes. These results underscore the potential of RF-based sensing as a novel method for real-time cardiac monitoring in pacemaker systems.

This paper proposes an Affine frequency division multiplexing (AFDM)-empowered integrated sensing and communications (ISAC) design, referred to as AFDM-ISAC. We first design a novel AFDM-ISAC frame structure that consists of both ISAC and pure data symbols. Each ISAC symbol consists of a single chirp subcarrier for both sensing and channel estimation, while the remaining subcarriers are allocated for communication. Building upon this structure, we present an analog-domain sensing receiver that down-mixes the received echo with a local chirp to fully exploit \textit{chirp compression} gains avoiding the need for full-duplex hardware. In addition, a sensing fusion algorithm, guided by AFDM modulation parameters, is further proposed in the digital domain. Leveraging the distinct features of the proposed AFDM-ISAC frame, we present a low-complexity channel estimation scheme for high mobility channels based on a generalized complex exponential basis expansion model (GCE-BEM), along with an optimal power allocation strategy between pilot and data symbols. Moreover, to support frame-based AFDM communications, a GCE-BEM-based Kalman filter is also employed for robust intra-frame channel estimation.

Precision-agriculture networks based on private 5G NR should ensure reliable connectivity for IoT sensor nodes throughout the crop growing season, yet the propagation environment changes dramatically as vegetation grows and matures. We formulate $K$-base-station~(BS) placement as a \textit{maximin seasonal coverage} problem that maximizes the worst-case coverage fraction across all crop growth stages. Since each objective evaluation requires expensive ray-tracing simulations across all stages, we adopt a Gaussian-process Bayesian optimization~(GPBO) framework that builds a probabilistic surrogate of the robust objective using ray tracing. On a $1\,\text{km}^2$ multi-crop farm with three distinct crop zones at $3.5\,\text{GHz}$, the proposed scheme achieves $72.8\%$ worst-case coverage with $K{=}3$ BSs in fewer than fifty ray-tracing evaluations, outperforming budget-matched state-of-the-art approaches by at least $4.6\,\text{pp}$ across all four seasonal stages.

Unmanned aerial vehicle (UAV) communications have been recognized as a key component of future sixth-generation (6G) space-air-ground-sea integrated networks. Accurate characterization and modeling of air-to-ground (A2G) channels are essential for the design and optimization of low-altitude communication systems. This paper presents a wideband A2G channel measurement campaign in an urban environment at 2.85 and 4.6~GHz in FR1 and 7.25~GHz in the FR3 frequency band, each with a bandwidth of 250~MHz. To enable reliable line-of-sight (LoS) and non-line-of-sight (NLoS) propagation state identification, a weakly supervised method is developed by fusing geometric priors, channel features, and spatial consistency constraints. Furthermore, based on the measured data, A2G channel characteristics are extracted and analyzed under LoS/NLoS conditions across different frequency bands, including path loss (PL), shadow fading (SF), power delay profile, root-mean-square delay spread (RMS-DS), and Rician $K$-factor. The results show that the close-in model fits the measured PL more accurately than the 3GPP reference model, and that NLoS propagation leads to larger path loss exponents and stronger SF than LoS propagation. For channel delay characteristics, higher-frequency channels exhibit fewer effective MPCs and weaker delay dispersion, indicating increased channel sparsity. Specifically, the mean RMS-DS under LoS conditions decreases from 93.11 to 46.84~ns, while the mean Rician $K$-factor increases from 9.16 to 12.88~dB. The statistical results further show that the RMS-DS and the Rician $K$-factor can be well characterized by lognormal and normal distributions, respectively. Moreover, the movement of the receiver in a complex scattering environment intensifies the spatial non-stationarity of the A2G channel.

This article proposes the concept of \emph{brain-body-to-everything (B2X)} networks to facilitate the integration of wireless networks and embodied intelligence. In this framework, the \emph{brain} refers to the intelligence functions for reasoning, planning, and decision-making, the \emph{body} denotes the physical embodied agent that senses and acts in the real world, and \emph{X} represents the surrounding ecosystem involved in the brain-body interaction loop. Two B2X architectures with \emph{distributed} and \emph{centralized} brains are introduced to characterize different placements of intelligence across the body, base station, and core network. The uplink and downlink designs of B2X networks are then discussed under a representative base-station-side brain setting. For the uplink, communication is redesigned for B2X state acquisition under event urgency, sensing volume, and simultaneous multi-body access. For the downlink, communication is redesigned to coordinate command delivery and conventional service under shared radio resources. Based on these uplink and downlink considerations, a communication-control Pareto boundary is further used to characterize the loop-level trade-off between wireless transmission performance and control quality in B2X networks. Finally, several open research problems are discussed to guide future B2X network design.

For multiple-input multiple-output (MIMO) channels with phase modulation, we recently proposed a method of unitarily transforming the channel matrix into a certain row-echelon form, by which the original MIMO channel can be converted into a certain number of scalar sub-channels with two phase inputs, thereby forming an annulus constellation geometry, and corrupted by both the additive white Gaussian noise and weak self-interference. In this paper, several bounds are derived to evaluate the fundamental limit of such a specific transceiver architecture. Two upper bounds are obtained by upper-bounding the capacity of a scalar channel with an annulus support constraint from the perspective of the convex geometry, while a lower bound is obtained by the standard entropy power inequality. Numerical results show that the gaps between these bounds are small at high signal-to-noise ratios for the MIMO phase-modulated channels over the Rayleigh fading and the single-input multiple-output symbiotic communication system assisted by a reconfigurable intelligent surface.

Recent advances in generative artificial intelligence (AI) and embodied intelligence (EI) enable autonomous agents to interact with the physical world. However, scaling these systems into networks of multiple agents, namely the Internet of EI (IoEI), faces critical bottlenecks. These include the overhead of massive multimodal data transmission and the decoupling of logical reasoning from physical constraints. To address these challenges, we envision the Semantic-based IoEI (SIoEI), which leverages semantic information as a unified metric throughout the agent lifecycle. We systematically define four key dimensions of EI: perception, intelligence, control, and communication. We further elaborate how semantic empowerment revolutionizes environmental perception, cognition and task planning, action generation and robust control, and communication and networking. We also present a case study to verify that, the semantic-empowered end-to-end process significantly improves channel robustness and reduces end-to-end latency for EI. Finally, we outline critical open research directions for the SIoEI paradigm.

Future engineering infrastructures are evolving into large-scale, open, heterogeneous, and wirelessly interconnected complex systems. These systems present significant challenges in optimizing network resource utilization, managing high-dimensional information spaces, and accommodating diverse business requirements. Intellicise networks, characterized by Intent-driven operation, semantic-native capability, and distributed intelligence, offer a promising paradigm for enabling such intelligent complex systems. We provide a systematic exploration of future intelligent complex systems from the perspective of intellicise networks. Specifically, we propose a cross-domain intelligent communication network architecture based on intellicise networks, grounded in information theory, systems theory, game theory, and cybernetics. The architecture comprises a cross-layer organizational framework, multi-functional planes, and novel information flows. The cross-layer framework defines the vertical evolution from perception and cognition to decision, while the control, user, data, computation, intelligence, and security planes deliver horizontal intellicise capabilities. Moreover, data, knowledge, model, and task flows interconnect the various layers and planes, forming a closed-loop process that derives simplicity from high-level intelligene while concurrently pursuing enhanced. Building on this architecture, we review key enabling technologies, tracing their evolution from semantic extraction to intent understanding, from heterogeneous resource integration to self-configuration and self-optimization, from generative artificial intelligence (AI) to agentic AI, and from embodied AI to symbodied AI. Additionally, we present a case study on intellicise networks for embodied agent communications and discuss representative applications and services for intelligent complex systems.

This paper examines several model based approaches for retrieving surface soil moisture from ALOS-2 PALSAR-2 quad-pol imagery, over a lime stone quarry in southeastern Finland. The study primarily targets physically interpretable semi-empirical modeling approaches, with generic ML modeling used as a benchmark. Along with common polarimetric observables, we propose a generalization of the SAR time series based TU Wien soil moisture index (SMI) retrievals examined across several representational spaces derived from polarimetric coherency matrix $[T3]$. This study was conducted over a closed tailing storage facility and a landfill, with a set of 9 repeat pass PALSAR-2 images. The best semi-empirical configuration combining temporal context SMI and current observation PolSAR parameters achieved $R^2=0.67$ and RMSE $=5.65$ volumetric \% units. The strongest $SMI_{[T3]}$ approach with sediment-specific calibration, achieved $R^2=0.66$ and RMSE $=5.67$ vol. \%, which was considerably better than using $SMI_{HH}$ or $SMI_{VV}$. The proposed approach was sensitive to representations: dB-based projection outperformed linear or trace-normalized $[T3]$ representation. Factoring in sediment information dramatically improved retrieval performance compared to using global model fitting. Machine learning results closely approached but not outperformed semi-empirical model based methodologies. Similarly, they highlighted the need for sediment-specific modeling as well as the importance of including time-series/temporal backscatter dynamics during SSM retrieval. Our study demonstrated the utility of physics based SSM retrieval approaches in the complex multi-sediment mine environment under relatively scarce reference data conditions.

The sensing capability of the pinching-antenna system (PASS) is analyzed from a Ziv-Zakai bound (ZZB) perspective, motivated by the sensing ambiguity arising from the multimodal observation model inherent to PASS. In comparison to other Bayesian sensing bounds, the ZZB provides a lower bound on the mean-squared error (MSE) across a broad range of signal-to-noise ratios (SNRs) and accounts for ambiguity in the likelihood functions. First, an observation model is developed for an uplink sensing scenario where a single sensing target transmits uplink pilots to a single-waveguide PASS receiver equipped with multiple pinching antennas (PAs). Building on this model, general ZZB expressions are derived for arbitrary prior distributions of the target's position, and are then specialized to the Gaussian and uniform cases. Second, the asymptotic ZZBs in low- and high-SNR regimes are characterized, and the relationship between the ZZBs and the conventional Bayesian Cramér-Rao bound (BCRB) is further studied by introducing the concept of an ambiguity function. Furthermore, to reduce the high computational complexity of direct evaluation of the ZZB, SNR-free and SNR-aware surrogate objective functions are proposed to facilitate ZZB-based optimization for enhancing sensing performance. Numerical results demonstrate that: i) Compared with the BCRB, the ZZB provides a tight sensing performance lower bound over a wide range of SNRs, ii) the ambiguity-awareness of the ZZB can address the multimodality-induced ambiguity in sensing, thereby yielding a reliable lower bound on the MSE, and iii) the proposed surrogate objective functions enable effective ZZB minimization with a lower computational complexity.

This work considers a Differential Reflecting Modulation (DRM) scheme for Reconfigurable Intelligent Surfaces (RIS) not requiring channel state information (CSI). When operating over time-varying fading channels, such schemes with Conventional Differential Demodulation (CDD) receivers experience high error floors and performance degradation. To address these issues, we propose a Decision Feedback Differential Detection (DFDD) technique for DRM. We explore the application of DFDD for RIS DRM and conduct extensive Monte-Carlo simulations to analyze performance. Results demonstrate the viability of our DFDD technique across various RIS scenarios and highlight the importance of proper parameter selection to achieve good performance. The DFDD scheme is also compared with uncoded and Differential Space-Time Modulation (DSTM) coded DRM using CDD based receivers. We observe that at low SNR, the DFDD scheme performs almost as well as the DRM with CDD scheme, but worse than the DSTM coded DRM. As the SNR increases however, both CDD-detected systems encounter high error floors while the error rate of DFDD based scheme continues to improve until it reaches a relatively low error floor. It is shown that the chief merits of employing DFDD receivers in such RIS systems is the low error floors they provide over time varying fading channels, albeit at expense of a small increased complexity.

For a remote estimation system, we study the optimization of age of incorrect information (AoII), which is a recently proposed semantic-aware information freshness metric. In particular, we assume an information source that observes a discrete-time finite-state Markov chain (DTMC), and occasionally transmits status update packets to a remote monitor which is tasked with remote estimation of the source. For the forward channel from the source to the monitor, we assume the channel delay to be modeled by a general discrete-time phase-type (DPH) distribution, whereas the reverse channel from the monitor to the source is assumed to be perfect, ensuring that the source has perfect information on the AoII and the remote estimate at the monitor, at all times. Push-based transmissions are initiated when AoII exceeds a threshold depending on the current estimation value, i.e., multi-threshold policy. In this very general setting, our goal is to minimize a weighted sum of the time average of a polynomial function of AoII, depending on the remote estimate, and energy consumption from transmissions. We formulate the problem as a semi-Markov decision process (SMDP) with the same state-space of the original DTMC to obtain the optimal multi-threshold policy, whereas the parameters of the SMDP are obtained by using a novel stochastic tool called dual-regime absorbing Markov chain (DR-AMC), and its corresponding absorption time distribution named as dual-regime DPH (DR-DPH). The proposed method is validated with numerical examples using comparisons against other policies obtained by exhaustive search, and also various benchmark policies.

Integrated sensing, communication, and computation (ISCC) is envisioned as a key enabler for intelligent services in future wireless networks. However, in multi-device ISCC systems, directly offloading full orthogonal frequency division multiplexing (OFDM) sensing data to the edge may incur excessive overhead, thereby limiting sensing performance under practical resource constraints. In this paper, we propose a subcarrier selection-based sensing framework for multi-device ISCC systems, where frequency-domain redundancy in OFDM sensing data is removed during local preprocessing to reduce sensing data transmission and processing overhead. Based on the proposed framework, we establish analytical models for sensing accuracy, delay, and energy consumption, and formulate a sensing accuracy maximization problem under practical resource constraints. To solve this problem, we develop an alternating direction method of multipliers (ADMM)-based algorithm. Experiments on commodity wireless devices validate the effectiveness of the proposed framework and show that it consistently outperforms three baseline schemes under various resource constraints.

To maintain the antenna apertures offered by 5G massive MIMO systems operating at the sub-6GHz band, known as FR1, 6G base stations (BSs) using the upper-mid band, FR3, should increase the number of antennas by a factor 4-8, giving rise to gigantic MIMO. This poses challenges in terms of processing complexity and interconnection bandwidth. The WAX framework, previously introduced for exploring trade-offs in decentralized architectures, may offer the flexibility needed to tackle these challenges. However, no results have been established on the applicability of this framework in the presence of hard-boundary constraints. The current work explores gigantic MIMO implementations based on a novel adaptation of the WAX framework, where the decentralized processing is performed by non-cooperating hardware modules. These modules may be implemented through state-of-the-art massive MIMO baseband units (BBUs). The results show the potential of the proposed framework towards exploiting trade-offs between complexity and performance in practical gigantic MIMO implementations.

This paper presents a spatially coupled sparse code multiple access (SC-SCMA) framework to overcome the performance and scalability limitations of conventional SCMA systems. By analyzing the pairwise error probability associated to multi-user error patterns, we show that spatial coupling projects the superimposed SCMA codewords into a higher-dimensional effective signal space, leading to a strictly improved minimum Euclidean distance (MED) compared with conventional SCMA, while simultaneously enhancing the coding gain through global message propagation and the diversity gain through inter-block resource spreading. Such a distance gain is shown to be governed by the effective access dimensionality (EAD) induced by the coupled factor graph. With the aid of spectral graph theory, we establish a direct relationship between the spectral gap of the factor graph and a lower bound on the EAD, providing a computable structural metric that guarantees MED improvement under various error patterns. Building upon these theoretical insights, we introduce a low-complexity structure-aware codebook design approach, including a spectral-gap-oriented construction of spatially coupled factor matrices and a localized codebook optimization strategy that exploits the dominant error-inducing local user group. Simulation results validate the analysis and demonstrate that the proposed SC-SCMA consistently outperforms conventional SCMA in overloaded massive access channels.

Unresolved measurements occur in many inference problems where two or more hidden processes may, at times, jointly generate a single measurement. For instance, such phenomena are encountered in multiobject tracking owing to the limited resolution capabilities of practical sensors; or in camera-aided autonomous driving due to shadowing or occlusions. Substantial performance degradation, such as track losses, are incurred when unresolved measurements are not accounted for. In this paper, we address multiobject tracking under a generalized unresolved measurement model, where any subset of objects may generate a single unresolved measurement according to a probabilistic model. Our innovation lies both in modeling and algorithm-design directions. First, we develop a probability distribution for object partitions based on a model of pairwise coupling of objects and subsequently a probability distribution for object-to-measurement association variables. This generic model incorporates sensor resolution capabilities, sensor detection, and sensor noise characteristics for object groups. Second, a generic Loopy Belief Propagation (LBP) method as well as a specialized Gaussian-LBP (GLBP) algorithm are proposed that perform object state inference under the aforementioned model. In contrast to direct marginalization methods, which involve a computational complexity of $O(m^n)$, for $m$ measurements and $n$ objects, the proposed GLBP algorithm achieves a computational complexity on the order of $O(m n 2^{n})$. Numerical results demonstrate the effectiveness of our proposed GLBP, with estimation performance that closely matches that of exact marginalization for only a fraction of the computational resources.

WiFi channel state information (CSI) sensing must coexist with data communications, which constrains the acquisition rate of fresh CSI measurements. To model this, we formulate CSI-based human activity and identity recognition under a sensing rate constraint that limits the fraction of time slots, within a measurement session, where CSI samples are available. This framework captures sensing-communication resource sharing and uncontrolled packet loss or traffic-driven irregularity. To satisfy the sensing constraint, two fixed CSI sampling policies are considered: a deterministic policy and a stochastic Bernoulli policy. We propose a low-cost age-aware WiFi sensing framework that explicitly incorporates sample freshness into the model training. The age of each retained CSI sample is first encoded and then fused with the CSI embedding via multiplicative fusion. On the NTU-Fi human activity recognition and person identification datasets, the proposed model consistently outperforms both a CSI-only baseline and the state-of-the-art time-aware attention model from the UniFi benchmark. For example, it yields up to a 10-percentage-point improvement over the UniFi method for person identification, with the largest gains observed under strict sensing budgets.

Active systems can exhibit a broad range of phenomena forbidden in equilibrium. Their dynamics are often specified by abstract local update rules, and it is generally unclear when the same behavior can arise from physically natural driving. Here we show that two simple driving mechanisms can universally simulate any local active dynamics in spin systems. The first is the familiar setting of a time-periodic Hamiltonian coupled to a cold bath, which we call a "many-body Brownian pump." As a second mechanism, we promote the Brownian ratchet, traditionally a mechanism for transport, to a "many-body Brownian ratchet": a static Hamiltonian coupled to a hot bath and a cold bath, where the resulting steady heat current can be harnessed not only to drive transport but also to generate local active dynamics. Using probabilistic cellular automata as an explicit model, we prove that for any continuous-time (or discrete-time) local active dynamics, there is always a many-body Brownian ratchet (or pump) that approximates the dynamics, up to noise that can be made arbitrarily weak by tuning energy scales and other parameters. As a concrete demonstration, we construct a simple ferromagnetic Ising ratchet on a bilayer lattice. When the two layers are coupled to baths at different temperatures, this model serves as a robust classical memory even under a symmetry-breaking field, something impossible in equilibrium. More broadly, our work shows that ratchets can use steady heat currents to autonomously generate and stabilize novel collective behavior, realizing a new static setting for nonequilibrium many-body dynamics.

The transverse-field Ising model has attracted a lot of attention in recent years, especially in the quantum simulation and quantum computation literature. This interest is driven by many platforms for analog quantum computation, which implement the transverse-field Ising model for solving optimization problems, such as quantum annealing. However, it has remained an open question whether the Ising model with a global transverse field is equivalent to the gate model of quantum computation. Here we answer this question affirmatively for the case of a non-monotonic time-dependent transverse field. Building on a recent result by Cesa and Pichler on global control of Rydberg atoms, we provide a construction that allows simulating arbitrary quantum circuits using the Ising model with global transverse field with polynomial overhead in time, qubit number, and energy scale. Although the polynomial overheads we establish here are large relative to what is feasible on real-world quantum hardware, our result motivates the development of more sophisticated methods for simulating quantum circuits using the Ising model with a global transverse field. Additionally, under the assumption that quantum computing is strictly more powerful than classical computing, our result serves as a no-go theorem for efficient classical simulation of the transverse-field Ising model with a time-dependent global transverse field. Therefore, our finding is relevant for multiple communities, from analog quantum simulation and quantum optimization on various platforms to complexity and control theory.

I revisit Type-IIB axion--dilaton Euclidean saddles in a specified axion charge sector. In that sector, the solution with $E=0$ is the BPS instanton, while $E>0$ gives non-BPS wormholes with a smooth throat. The two cases solve the same radial equations but define different fluctuation problems. For the $E=0$ instanton, the Hamiltonian constraint, gauge quotient, charge-sector boundary condition, and removal of collective zero modes reduce the quadratic action to a physical Hessian. This Hessian factorizes, $ {\cal H}_ν={\mathcal Q}_ν^\dagger{\mathcal Q}_ν$. I interpret this as an endpoint theorem, beyond a stability theorem for the full $E>0$ wormhole. This puts Type IIB wormhole spectra on firmer grounds. I also separate the connected two-ended wormhole throat from its long-distance two-end multipole operator term. Once the coefficient matrix $C^{ij}$ is derived, the different-component and same-component placements of the two end insertions are terms in the same quadratic expression. Removing either term requires a genuine projection or cancellation.

In this work, we perform an analytical and numerical analysis of quantum landscapes for controlling special four-level quantum systems for which we prove that the null control is a five-order trap: a $V-V$ system and an anharmonic system. As a control goal, an observable optimization is considered. The rigorous theoretical analysis is followed by the numerical experiments based on the GRadient Ascent Pulse Engineering (GRAPE) algorithm and Gradient Projection Method (GPM), performed to investigate the behavior of the efficiency of optimization for unconstrained (using GRAPE) and constrained (using GPM) controls. As the main result, we observe an interesting phenomenon with a diverse behavior of the optimization efficiency depending on the system Hamiltonian -- sharp increase of the optimization efficiency up to 100% at certain distance from the null control for a V-V system, while much slower and less significant increase (and even small decrease) for a system with the chain interaction. This sharp difference might be related with the fine structure of the subspace of controls where second derivative of the objective functional is zero.

Extracting the full power of non-local correlations in prepare-and-measure (PM) scenarios requires precise control over the timing and structure of the receiver's measurements. Indeed, recent developments in entanglement-assisted classical communication scenarios have shown that adaptive strategies-where the receiver uses the transmitted message to guide their measurement choice-can outperform standard non-adaptive protocols. Moving beyond quantum theory, however, the ultimate limits of such advantages remain largely unexplored. In this work, we thoroughly study adaptive and non-adaptive non-signaling (NS) assistance in PM scenarios with classical communication. We provide simple characterizations of the sets of behaviors that can be realized using both non-adaptive and adaptive NS assistance in arbitrary PM scenarios. As a consequence, we show that non-adaptive NS assistance is already strong enough to reproduce quantum communication with the same message dimension: the transmission of a qudit can be simulated by a classical dit assisted non-adaptively by NS correlations. We then compare adaptive and non-adaptive NS assistance. We prove that any adaptive NS advantage can be traced back to scenarios in which the receiver has no measurement choice, ruling out the genuinely multi-setting advantages found in entanglement-assisted quantum protocols. Finally, we identify all PM scenarios where adaptive NS strategies provide a strict advantage over non-adaptive ones.

Monolayer tungsten diselenide (WSe$_2$) has become a suitable platform for quantum transport and spintronics and valleytronics applications because it possesses an intrinsic band gap and strong spin-orbit coupling and spin-valley coupling features. The electrostatic confinement of Dirac fermions proves challenging in graphene because of Klein tunneling, yet WSe$_2$ provides an environment that supports both carrier localization and the development of confined quantum states. In this work, we theoretically investigate the confinement of massive Dirac fermions in a WSe$_2$ quantum dot generated by a localized magnetic field. Using the effective Dirac Hamiltonian in the presence of a magnetic flux, we derive the exact wave functions and scattering coefficients by employing Kummer's confluent hypergeometric functions together with Bessel and Hankel functions. Our results show that the localized magnetic field provides an efficient mechanism to suppress Klein tunneling and promote the formation of stable quasibound states. We systematically examine the scattering efficiency and carrier density distributions as functions of the incident energy, magnetic field strength, and quantum dot radius. We find that low-energy carriers are strongly confined by the magnetic barrier, while the interplay between magnetic localization and geometric confinement gives rise to sharp and tunable resonance peaks. These results provide valuable insight into the control of spin-valley transport in transition metal dichalcogenide nanostructures and establish a theoretical basis for the development of quantum confinement devices and quantum information technologies.

Symmetry is a powerful inductive bias, but in quantum reservoir computing (QRC) it cannot be imposed only by making the reservoir symmetric. QRC maps inputs through fixed quantum dynamics into nonlinear expectation-value features and trains only a classical readout, so the relevant symmetry must be visible in the measured feature map. We study cyclic forecasting tasks, such as sensors around a turbine or weather stations along a latitude circle, where the same local pattern should be forecast by the same rule wherever it appears on the ring. Thus, rotating the input by one site should rotate, not change, the predicted field. We show that a symmetric Hamiltonian is not enough: even large Pauli measurement sets can fail if their channels do not match the data symmetry, since optimization cannot recover channels that were never measured. We address this through observable-orbit completion, which measures symmetry-related observable channels and aligns encoding, dynamics, measurement, and readout. The strongest gains arise from aligning all four interfaces together, with matched spin-ring, real-weather, and IBM hardware checks supporting the same measured-span mechanism.

We propose an Ensemble Feature Selection (EFS) method for fast estimation of process infidelity of involutory multi-qubit gates, including non-Clifford targets, for which standard Clifford-based benchmarking does not apply. The method selects a compact set of experimentally executable circuit measurements from a candidate pool through offline training on a physically motivated ensemble of noisy channels, and combines them into a linear estimator with weights learned by ridge regression. The training ensemble is an explicit and tunable component of the protocol, incorporating prior knowledge about dominant hardware noise mechanisms. The estimator is validated on ibm_kingston using two Clifford validation benchmarks structurally related to the transpiled CCZ circuit, against independent Interleaved Randomized Benchmarking (IRB). Both show close EFS-IRB agreement across a wide range of process infidelities, with an estimation precision of approximately 0.01 over a process infidelity range of 0.02-0.2. EFS is subsequently applied directly to CCZ on the same device.

Non-perturbative processes in the early universe may create overdense structures in scalar fields like the inflaton, called oscillons. In this work, we explore whether the leading order higher derivative contributions to the scalar-tensor theory change the formation and growth of these structures, and investigate the limits in which the effective field theory (EFT) description breaks down. We find that whilst the properties of the oscillons are not significantly modified, and black holes do not generically form, for large couplings the period of formation can result in the evolution leaving the regime of validity of the EFT, at which point predictivity is lost and the next order terms in the EFT should become relevant. If the oscillons survive their formation, they tend to be stable and the EFT corrections remain bounded. The EFT breakdown is triggered by large curvature terms in the metric in the densest regions of the oscillon, meaning that approximations of such modified theories that neglect the local backreaction and non-linear dynamics of the fields may miss important effects.

In a series of papers in the 1980's Alexander Holevo proved a classification theorem for continuous quantum measurement processes, or, as they would today be called, stationary quantum trajectories in continuous time. His main tools were functional analytic in character: starting from a Bochner-type inequality he employed dilation techniques for positive definite kernels. Here we give an alternative, more probabilistic proof: we use weak convergence of measures and employ Levy's Continuity Theorem. We clarify the boundedness conditions in Holevo's theorem, and supply a simple example from quantum optics.

Non-invertible defects are usually diagnosed through scaling spectra or infrared CFT data. We show that the Fibonacci duality defect of the critical golden chain already carries an exact categorical fingerprint at finite lattice size. The even-length antiferromagnetic ground state has fixed cut-charge weights, giving P_tau/P_1=phi^2 and log g=log phi without finite-size extrapolation. The proof is a finite-dimensional operator identity for the sandwiched cut projectors, combined with a Perron-Frobenius sector theorem for the even-length ground state. This gives a sharp lattice-level boundary entropy for a non-Abelian duality defect. We also separate this exact two-charge result from the finer six-primary tricritical-Ising resolution: the latter is located by the standard scaling-limit Virasoro branching of A_4 affine-TL packets, and is not an assumption in the finite-size theorem.

We study a one-dimensional electronic chain coupled to a homogeneous quantized vacuum field and electron-electron interactions. In the absence of the latter, we derive a low-energy effective description in the presence of light-matter coupling, which we identify as a strange Luttinger liquid. Although it retains a formal resemblance to conventional Luttinger liquid theory, the coupling to the quantum field qualitatively modifies the low-energy sector and breaks the standard velocity relation underlying Luttinger universality. For finite electron-electron interactions, we recover a phase diagram featuring several phases as a function of interaction strength and hopping amplitude, including a phase hosting Majorana-like zero modes. Using exact diagonalization, we compute observables that characterize the phase boundaries and show that the cavity field significantly shifts them. We also study the fate of Majorana-like states under the influence of the cavity field, highlighting their modification by light-matter coupling. Finally, we investigate whether the strange Luttinger liquid description identified in the noninteracting regime continues to hold when electron-electron interactions are introduced.

Non-invertible symmetries are characterized by topological defects of irrational quantum dimension, but their imprint on the entanglement of a quantum many-body state has not been resolved at the level of the spectrum. We show that the categorical data of the canonical example -- the Kramers--Wannier (KW) duality defect of the critical Ising chain, with quantum dimension d_sigma=sqrt(2) -- is encoded in the single-particle entanglement spectrum of its ground state: a maximally mixed Majorana zero mode is the spectral origin of the boundary entropy log g=(1/2)log 2, hence of d_sigma itself. Reading the same duality-twisted ground state along two independent routes -- the transfer-matrix momentum shift and the Casimir curvature of the energy -- pins the twist-field weight h_sigma=1/16 twice over, and the defect Hilbert space organizes into a half-integer sigma-twisted conformal tower. This promotes the boundary entropy from an integrated number to a level-resolved spectral signature of non-invertibility, and supplies an exactly solvable calibration target for tensor-network studies of duality defects that lack a free-fermion shortcut.

How stellar-mass black holes grow after their birth is a central open question in astrophysics. Gravitational-wave observations have revealed a subpopulation of coalescing black holes with both high masses and high spins, but whether these properties arise from hierarchical mergers in dense stellar environments or from accretion onto isolated black holes has remained unresolved. Here, using a flexible mixture population model applied to the 259 binary black hole mergers in GWTC-5, we show that the mass function of the high-spin subpopulation traces, peak by peak, the predicted remnant-mass distribution of the low-spin, stellar-collapse-origin subpopulation up to $\sim80\,M_\odot$. This morphological match, quantified by a Bhattacharyya coefficient as high as $\sim0.95$, is naturally expected if the high-spin black holes are themselves the products of earlier mergers, whereas any alternative scenario would require fine-tuning, thereby providing smoking-gun evidence for hierarchical mergers. In addition, the sharp upper-mass cutoff of the low-spin subpopulation at $m_{\rm max,1}=54.2^{+7.7}_{-7.2}\,M_\odot$ yields an astrophysical $S$-factor of $S_{300}=151^{+30}_{-26}$~keV~b (68\% credible interval) for the $^{12}{\rm C}(α,γ)^{16}{\rm O}$ reaction, in agreement with the benchmark theoretical value. These results establish that the entire observed black-hole population can be accounted for by stellar collapse followed by dynamical hierarchical assembly, without invoking primordial black holes.

Hawking radiation - the emission of quantum particles at the event horizon of a black hole - connects gravity with quantum mechanics and thermodynamics; the Bekenstein-Hawking entropy has been the benchmark for potential quantum theories of gravity. But Hawking radiation has never been observed in astronomy, only in laboratory analogues and the chances of ever observing it in space are astronomically small. The energy of Hawking radiation must come from the gravitational field around the black hole, but how field quanta generate Hawking quanta has been unknown. Here we report on experimental and theoretical evidence for the process that generates Hawking radiation in a fibre-optical analogue of the event horizon. There, as in gravity, it has been believed that Hawking radiation comes from a complicated, cascaded process; here we have found a simple, direct process and measured its backreaction on the field. Our findings suggest an equally direct process for other laboratory analogues and perhaps also for gravitational fields, shedding light on how black holes might radiate.

Quantum metrological advantage is certified by averaged Fisher responses: contrast, susceptibility, or quantum Fisher information (QFI). This fails in quenched sensors, where slow environmental variables freeze within a session but vary between repetitions: shallow nitrogen-vacancy (NV) centers, superconducting qubits with slow two-level fluctuators, and semiconductor spin qubits in drifting charge noise. They sample session-resolved Fisher geometries, not an averaged channel. Certification conditions on the latent session, projects nuisance directions, inverts to attainable loss, then tail-certifies; this inverse upper-tail loss defines quenched tail-certified information. A no-go theorem: no averaged Fisher data determine this certificate; ensembles sharing averaged Fisher matrix, QFI, and projected information have finite or zero certified precision. A Fisher-zero integrability transition governs collapse: the inverse-loss tail exponent $β$ sets the boundary, with nonintegrable certified loss for $β\le 1$, even when annealed information is large or scaling. The certified quantum resource is response transverse to latent disorder, not raw amplification sharing its generator; universal design laws: safe windows, nondegenerate portfolios, Fisher reserves, action separation, Fisher-cut criteria. A shallow-NV Ramsey tournament shows average-QFI optimization is tail-catastrophic, whereas tail-certified designs recover nearly three orders of magnitude in certified information at equal shot budget and latent ensemble. These non-self-averaging phases are Fisher glasses, governed by Fisher-zero rare-event statistics.

We show that combining quantum error detection (QED) with probabilistic error cancellation (PEC) gives more accurate and lower-variance estimates than PEC alone, provided that the symmetry measurements required for QED are carefully chosen. Because noisy symmetry measurements can negate the benefits of the PEC+QED approach, we cast the selection of measurement configurations as a classical optimization problem that systematically suppresses the impact of noise. Applying optimized PEC+QED to GHZ-state output distributions and to simulating the time-dynamics of a generalized superfast encoded Fermi-Hubbard model, we find consistent improvements over PEC. For GHZ states, the optimization over symmetry measurement configurations is essential for achieving an advantage. For the Fermi-Hubbard model, PEC+QED improves observable estimation on a $2 \times 2$ lattice and for larger systems the mitigation overheads can be reduced by measuring only subsets of stabilizers. Our results demonstrate the importance of circuit-specific tailoring of QEM techniques and that fault-tolerant design principles may already provide value for near-term devices.

We investigate and establish an analytic connection between the exact solutions describing the dynamics of a two-level system driven by periodic external fields, focusing on the cases of linear driving and the so-called rotating-wave approximation, or circular driving. In both cases, the exact solutions can be obtained by mapping the Schrodinger equation onto Heun equations: the confluent Heun equation for linear driving and the Heun equation for the rotating-wave case. In particular, we demonstrate a direct analytic connection between the exact solutions for linear driving and those for the rotating-wave case. This result is obtained by analyzing local solutions expressed in terms of hypergeometric functions, which, in the case of the confluent Heun equation, can be derived by considering path-multiplicative Floquet solutions involving a bilateral series. This series leads to two continued-fraction expansions that can be perturbatively solved by imposing a suitable consistency condition. The connection between the linear-driving and rotating-wave solutions is established through a perturbative procedure that allows us to recover not only the rotating-wave approximation itself, but also the correct Stark and Bloch-Siegert shifts, as well as the so-called high-frequency approximation.

We investigate the shadow and polarization images of a Konoplya-Zhidenko rotating non-Kerr black hole surrounded by a geometrically thick and optically thin accretion flow. The accretion flow is described by an analytical ballistic approximation accretion flow model. The numerical results show that the shadow image exhibits two main features, an outer bright ring and an inner dark region. The former corresponds to higher order images, while the latter is produced by the black hole event horizon. Increasing the deformation parameter $η$ does not significantly change the overall shape of the higher order images, but it enlarges their size. Increasing the spin parameter $a$ and the observer inclination angle $θ_o$ enhances the asymmetry of the higher order images and makes the intensity on the left side much larger than that on the right side. This behavior is associated with frame dragging and the relativistic Doppler effect. In the polarization images, the degree of linear polarization is much smaller in the higher-order image region than in other regions, and the polarization vectors extend over the whole image plane. These results indicate that the thick disk model produces features in both intensity and polarization images that differ markedly from those in thin disk models. Within the framework used in this work, the observed intensity and polarization signatures can serve as effective probes of the underlying spacetime geometry and near horizon accretion dynamics.

Efficient classical simulation of large-scale quantum networks frequently relies on noise approximations, which consider a restricted set of operators to describe noisy channels and operations. In this work, we demonstrate how such simplified error models, such as Pauli twirling or reset channels, can lead to severe quantitative and qualitative discrepancies in protocol performance predictions. We analyze, in particular, how small differences can accumulate in iterative and sequential protocols such as entanglement purification, entanglement swapping, and repeater chains. Our results reveal that neglected error contributions can lead to important performance under- and over-estimations, measurement-outcome dependency, and oscillations in the fidelity, which are entirely overlooked by the simplified error model approximations. These results show that rigorous validation of complete noise architectures is indispensable for accurately predicting operational thresholds in future quantum technologies.

We give an analytical expression for the dynamical Lie algebra corresponding to the QAOA-MaxCut problem on complete graphs, and show that the variance of the associated loss function scales linearly in the number of qubits. This solves an open problem from [ASYZ26] and confirms that such systems do not exhibit barren plateaus. The proof is based on projecting the dynamical Lie algebra generators onto subspaces given by the Schur-Weyl duality between irreducible representations of the unitary and symmetric groups.

We give a prescription to compute a dissipative action describing the large-scale thermal stress tensor dynamics of a holographic quantum field theory dual to AdS$_4$ gravity, in the context of the Schwinger-Keldysh formalism. Our prescription is valid to quadratic order in perturbations about the thermal equilibrium state. The hydrodynamical degrees of freedom of this action are realised in gravity as relative diffeomorphisms between the black hole horizon and the two asymptotic boundaries of the Crossley-Glorioso-Liu contour. We explicitly compute the action to first order in derivatives, and confirm it correctly reproduces the known hydrodynamic Green's functions. Our prescription requires a choice of horizon boundary conditions for the metric. We study the horizon symmetries that preserve these, and their relation to conjectured hydrodynamic symmetries responsible for many-body quantum chaos.

Gravitational-wave signals from massive black hole binaries observed by LISA can carry imprints of both the astrophysical environment of the source and possible deviations from general relativity. We investigate whether environmental effects leave a detectable imprint on the LISA binary population, and whether they can mimic modified-gravity effects with the same frequency dependence. As representative channels we adopt accretion and viscous migration in a circumbinary disk for the environmental sector, and a time-varying Newton constant $\dot G$ for the modified-gravity sector. All three effects enter the waveform at the same negative post-Newtonian order and are described, at leading order, by a common phase-deformation parameter, which makes them formally degenerate at the single-event level. Combining Fisher-matrix forecasts with a hierarchical nested-sampling analysis of synthetic catalogs from astrophysically motivated population models, we find that, even under extreme astrophysical assumptions -- an active fraction of $50\%$, together with a super-Eddington accretion tail -- the population-level posteriors remain fully compatible with vacuum. However, a hierarchical population-wide analysis may yield a non-trivial upper limit on the active fraction and a mild lower bound on the slope of the Eddington-ratio distribution. Environmental effects are therefore unlikely to bias LISA's tests of general relativity with massive black hole binaries in astrophysically realistic scenarios.

We introduce the Twisted Gaussian Schell (TGS) state, a two-mode mixed Gaussian state defined as the quantum-optical analog of the Twisted Gaussian Schell-model beam of classical paraxial optics, characterized by the so-called twist phase. In the TGS state, the twist parameter arises when an asymmetric two-mode thermal state is subject to local squeezing after the action of phase shifters and a beam splitter. Its defining quantum feature is nonclassicality: although the state is separable in its natural bipartition, when the twist parameter is nonzero there are global quadratures that can be squeezed below the shot-noise limit. The nonclassicality has also a direct signature in the joint photon-number distribution, which we obtain in closed form. Moreover, coupling each mode to an ancillary vacuum at a balanced beam splitter yields a four-mode state with entanglement in select $2\times2$ bipartitions, with local description given by two TGS states, and all $1\times3$ bipartitions. For fixed input squeezing, increasing the twist parameter activates entanglement where the state is otherwise separable and deepens it where already present. The classical physicality bound on the twist parameter coincides with the quantum physicality condition. These results advance the two-way bridge between classical beam engineering and quantum information.

We propose a photonic platform for simulating arbitrary single-qubit open-system dynamics using a single photon in an open Mach-Zehnder interferometer. A birefringent quartz plate induces a coupling between the polarization and frequency degrees of freedom. By treating the latter as an effective environment, we analytically derive the reduced polarization dynamics. We show that the resulting evolution is characterized by a controllable interplay between populations and coherence, instead of the usual dephasing caused by quartz plates. By adjusting the photon frequency distribution and interferometric parameters, we demonstrate that target single-qubit states can be efficiently reproduced through a tunable optical protocol expected to work under accessible experimental conditions. The simulator is benchmarked against paradigmatic open-system evolutions, including depolarization and non-Markovian dynamics, achieving high accuracy. Our results establish polarization-frequency engineered photonic interferometers as a versatile protocol for simulation of open quantum systems.

In this paper, we investigate timelike geodesic motion, periodic orbits, and the associated gravitational-wave signals around a Schwarzschild-like black hole (BH) embedded in a generalized Dehnen-type dark matter (DM) halo. We show that the Dehnen-type $(1,4,γ)$ DM halo profile modifies test-particle dynamics, with increasing the parameter of density profile, $γ$, leading to larger marginally bound orbit (MBO) and innermost stable circular orbit (ISCO) radii and angular momenta, together with a higher ISCO energy. These findings provide further insight into the role of the DM distribution in modifying the orbital dynamics, energy, and angular momentum of timelike test particles near the BH. Furthermore, we investigate the gravitational-wave signals produced by a stellar-mass compact object moving along periodic orbits around a supermassive BH embedded in a generalized Dehnen-type DM halo. Using the numerical kludge approach, we calculate the orbital trajectories and the corresponding gravitational-wave polarizations. We find that increasing the halo parameters $γ$, $ρ_s$, and $r_s$ produces larger periodic orbits, longer orbital periods, and lower waveform amplitudes. The resulting spectra lie mainly in the millihertz frequency range, while several characteristic-strain peaks lie above the sensitivity curves of future space-based gravitational-wave detectors such as LISA, Taiji, and TianQin. These results suggest that the surrounding DM halo may leave observable imprints on extreme mass-ratio inspiral (EMRI) gravitational-wave signals.

Understanding the correlation between material properties and microwave losses in superconducting films is a crucial subject for developing low-loss materials for quantum circuits. We focus on vanadium (V) as a novel material for superconducting quantum devices and discuss loss in V films in relation to their structural properties. Using a sputtering method, we grow four V-film structures on (001)-oriented Si wafers, employing Nb and Ta as the buffer and capping layer materials, respectively: Nb/V/Ta, Nb/V, V/Ta, and V. X-ray diffraction and atomic force microscopy reveal that the V films grown on the Nb buffer layers have higher uniformity of lattice orientation and smaller grain size than that directly grown on the Si wafer. Coplanar waveguide resonators are fabricated from the four V-film structures, and averaged photon number ($\langle n_{\rm ph} \rangle$) dependences of internal quality factor ($Q_{\rm int}$) are obtained by performing microwave measurements. By analyzing the obtained $Q_{\rm int}$ vs $\langle n_{\rm ph} \rangle$, it is found that loss at the V surface is dominated by $\langle n_{\rm ph} \rangle$-independent non-two-level-system (non-TLS) losses, which can be mitigated by introducing the Ta capping layer. Furthermore, the V films on the Nb buffer layers exhibit lower $Q_{\rm int}$ in the $\langle n_{\rm ph} \rangle$ range from 10$^{0}$ to 10$^{6}$ and higher non-TLS loss than that directly grown on Si wafers, even though the former has higher lattice-orientation uniformity than the latter. Origins of these trends might be relevant to V oxides, of which presence at surfaces and grain boundaries in bulk regions in the V resonators is suggested by energy dispersive X-ray spectroscopy and X-ray photoelectron spectroscopy, and/or V hydrides.

Quantum nonlocality admits several operational manifestations, one of which emerges from sets of orthogonal quantum states that cannot be perfectly distinguished by local operations and classical communication (LOCC). Such sets are regarded as nonlocal because their perfect discrimination requires global measurements. In contrast, sets that are perfectly distinguishable by LOCC are generally considered locally accessible and operationally classical. In this work, we investigate the role of incompletability in local state discrimination and introduce the notion of \emph{activation of incompletability}. Specifically, we demonstrate the existence of orthogonal sets that are initially perfectly distinguishable by LOCC and free from local redundancy, but which can be transformed via LOCC into strictly incompletable sets. We prove that activation of incompletability necessarily implies activation of nonlocality, whereas the converse fails in general, thereby establishing a hierarchy between the two activation phenomena. Furthermore, within the framework of local incoherent operations and classical communication (LICC), we show that any set whose incompletability can be activated can nevertheless be extended to a complete orthonormal basis of the Hilbert space, although the resulting completed basis is no longer perfectly distinguishable by LOCC. Our results uncover a fundamental interplay among local distinguishability, incompletability, coherence, and nonlocality, and provide new insight into the structure of locally accessible quantum information.

Quantum nonlocality paradoxes, such as that of GHZ, provide maximally sharp logical obstructions to classical probabilistic models of quantum correlations. They are key resources in a broad variety of information-theoretic tasks that exhibit unconditional quantum advantage. For example, in nonlocal games, which are communication tasks that serve as core technical tools in recent landmark results in quantum computational complexity theory. Their role in establishing quantum advantage motivated their study by Abramsky et al. who introduced an infinite family of three-qubit paradoxes exhibiting novel conditional structure. This was later extended by de Silva et al. into a full classification program. In this work, we completely classify all three-qubit nonlocality paradoxes established via a biconditional parity proof; this is a very large class of paradoxes that encompasses all earlier-known examples. We do this by introducing a suite of new structural and combinatorial techniques. We find that the landscape of nonlocality paradoxes is far richer than previously understood, violating regularity conditions underlying all prior constructions.

We construct an exact vacuum spacetime that develops closed timelike curves from regular, asymptotically flat initial data respecting the weak, dominant, and strong energy conditions. Einstein's equations fix the geometry through two functions, a harmonic transverse profile and a traveling-wave mode, whose evolution drives the closed curves of the time-machine core from spacelike to timelike. The chronology horizon admits a degenerate limit in which its generating closed null geodesic has vanishing boost and optical scalars, the conditions for a Killing spinor.

Scalable fault-tolerant quantum computation requires quantum error-correcting codes that simultaneously support universal logical operations, suppress hardware-specific noise, and enable efficient handling of photon-loss errors. Bosonic encodings such as the dual-rail and cat codes each offer attractive features but also exhibit important limitations when used in isolation. The dual-rail code enables efficient single-photon-loss detection by converting leakage out of the computational subspace induced by photon-loss errors into an erasure error. In contrast, the cat code provides a resource-efficient, bias-tailored error-correction scheme with bias-preserving logical gate operations. Here, we introduce the dual-rail cat code (DRCC), a concatenated bosonic encoding that combines an inner cat code with an outer dual-rail structure, thereby inheriting and enhancing the advantages of both constituent codes. We analyse the error-correction properties of the DRCC and propose a deterministic single-photon-loss correction protocol by concatenating it with an outer repetition code. Exploiting the code's intrinsic noise bias, we construct a universal set of logical gates using only beam-splitter interactions and demonstrate that all logical operations preserve the erasure-biased noise structure. The DRCC offers several distinctive advantages, including the absence of relative geometric phases during gate operations, deterministic erasure detection and correction, and simultaneous syndrome extraction without interrupting stabilisation. These features make the DRCC a promising bosonic code for hardware-efficient, bias-preserving, and erasure-resilient fault-tolerant quantum computation.

We study the the scattering of light by a binary system of identical atoms in which one of them is incoherently pumped. This system belongs to the kind of non-parity symmetric optical systems in which gains and losses are partially compensated. We carry out a fully quantum analysis of the directionality of the radiation scattered from the atoms when the incident light strikes the system either perpendicular or alongside the interatomic axis. We find that, generally, while the degree of asymmetry depends on the pump rate, the preferred direction for emission depends on the interatomic distance and the detuning of the probe field with respect to the resonant frequency. On physical grounds, for the case of frontal illumination, the asymmetry is the result of the interference of the photons emitted from different atoms. On the contrary, for side lighting, the asymmetry with respect to the side of incidence is caused by both the phase difference between the probe field photons that strike each atom and the interference between the photons emitted from different atoms. Further, for side lighting too, our quantum approach demonstrates that the forward scattered power depends on the side of incidence, which reveals the lack of reciprocity in the quantum optical response of the system. This result conflicts with what is obtained within a classical approach.

We perform experiments to quantify error suppression in fiber-connected superconducting QPUs using combined error mitigation techniques, demonstrating over 20\% improvement in operational fidelity across interconnected quantum processing units under realistic noise conditions.

We systematically investigate the role of squeezing in improving dissipation-induced nonreciprocal coupling mediated by a common reservoir, by incorporating a $χ^{(2)}$ nonlinearity into either the cavity mode, the reservoir, or both. Squeezing is capable of exponentially enhancing the effective coupling, but this enhancement is not universal in such a nonreciprocal interaction. Instead, it depends on the system configuration and specific parameter regimes. Specifically, the squeezing of the reservoir injects additional energy into the system via the dissipation channel, modifying the underlying dynamics in a nontrivial manner. Furthermore, we demonstrate that the proposed approach enhances the performance of the quantum battery, including stored energy, charging power, and ergotropy. We provide the corresponding analytical expressions, together with a systematic analysis of energy transport and parameter optimization. Extending the framework to optical isolation, we observe an exponentially amplified output signal. Our results open a new avenue for nonreciprocal quantum information processing and nonreciprocal quantum device design.

We present GsOQDC, an open-source graphical framework integrating optical-network design, distributed quantum-circuit compilation, scheduling, and DES-based remote-gate simulation, enabling end-to-end cross-layer evaluation of entanglement-resource dynamics and system-level performance in Optical Quantum Data Centers.

Geometric Machine Learning (GML) successes have been achieved through the thorough study and design of new equivariant neural networks. In comparison, geometric quantum machine learning (GQML) models lack such a detailed understanding and, despite already several proposals, a unifying perspective on their design remains elusive. In this work, we focus on GQML models for graph problems that showcase a lot of structure and still remain frontier in machine learning. For the case when n-node graphs are encoded in n-qubit states, we provide a comprehensive characterization of their constituents. Taken together, these furnish us with a toolbox for the design of quantum graph models, and we further probe its benefits including the natural integration with classical models, generalization of known GQML models (sometimes extending their expressivity at virtually no cost), and straightforward classical pre-training strategies. The latter two features are demonstrated in dedicated numerical experiments.

Qubit leakage is a noticeable source of errors for quantum computing. In quantum processors, leakage excitations traveling between qubits generate correlated errors and perturb gate implementations. Leakage mobility can also be utilized for creating dedicated leakage removal pathways and removal units. To quantitatively characterize leakage mobility and to guide better design of processor architectures, we study here leakage dynamics in transmons with tunable couplers through numerical and analytical methods. Even if the couplers are tuned to cancel the single-excitation exchange or the ZZ interaction, the leakage hopping rates still persists in the range of 0.8-10 MHz due to transmon nonlinearity. In typical operation regimes, however, transmon frequency detuning localizes leakage excitations. The next-nearest-neighbor transmons can be still be near-resonant opening leakage tunneling channels. To suppress longer-range hopping, we find that the frequency spread of the next-nearest-neighbor transmons needs to be in the range of 1-4 MHz. Utilizing leakage mobility, we propose two passive leakage removal units. One is based on a tunable coupler and a pumped transmon, and another on a junction readout scheme. Based on realistic experimental parameters, our results on selectively mobilizing or localizing leakage excitations are readily applicable in superconducting quantum devices.

We construct a residue-based framework connecting the thermodynamic geometry of a Schwarzschild-type black hole in $f(Q)$ gravity with its gravitational-wave quasinormal-mode spectrum. The analysis is based on the symmetric teleparallel formulation of gravity, in which the gravitational field is encoded by the nonmetricity scalar $Q$ rather than by curvature or torsion. For the Schwarzschild branch, the Robson--Villari--Biancalana (RVB) method gives the Hawking temperature through the simple-pole residue of the inverse blackening function. We show explicitly that the same residue also controls the logarithmic monodromy of the tortoise coordinate near the event horizon, and therefore enters the ingoing quasinormal-mode boundary condition. In the strict general-relativistic Schwarzschild limit the heat capacity is negative and finite, the one-dimensional Ruppeiner geometry contains no intrinsic curvature singularity, and no genuine thermodynamic phase transition occurs. In the extended $f(Q)$ state space, however, the modified horizon function and the effective Wald entropy generate a non-trivial thermodynamic Hessian. Its degeneracy condition coincides with singular behavior of the thermodynamic curvature and is reflected in the quasinormal-mode spectrum through shifts of the photon-sphere frequency, Lyapunov exponent, damping time, and near-horizon monodromy. This gives a precise statement of the internal relation between thermodynamic-geometric phase structure and gravitational-wave ringdown: both are different projections of the same analytic structure of the corrected black-hole metric.

The intrinsic spin of fermions induces torsion in spacetime, leading to an effective four-fermion interaction. This affects the bending of null and timelike geodesics inside a star with a spherically symmetric distribution of gravitationally dense fermionic matter of constant density. Our analysis shows an additional torsional contribution to the geodesic deflection, alongside the conventional curvature-induced effect. This change depends on the fermion number density, coupling constants of the different species of fermions with different chiralities, and the temperature of the matter distribution. We have shown that the torsion-induced corrections to null geodesic bending remains very small for both low- and high-mass white dwarfs, whereas significantly larger effects may arise in more compact astrophysical objects such as neutron stars.

High-fidelity quantum gates are essential for scalable quantum computation. However, at short durations, short-timescale waveform distortions break the time-reflection symmetry of control pulses, preventing the precise closure of cyclic evolution. This mechanism renders conventional symmetric protocols intrinsically over-constrained. Conventional strategies typically rely on smoothing the pulse envelopes or embedding the interaction pulse within a longer qubit pulse to bypass short-timescale distortions, which inevitably leads to a persistent speed-fidelity trade-off. To overcome this limitation, we introduce a cyclic control strategy based on parameter-space expansion, which restores controllability by incorporating an additional degree of freedom. We experimentally demonstrate this approach in a superconducting controlled-Z gate, achieving robust suppression of coherent errors without increasing gate duration, reducing the average coherent error from 0.27% to 0.12% across multiple two-qubit gates, as validated by cross-entropy benchmarking. Our results establish a general route to fast, high-fidelity cyclic quantum gates beyond the conventional speed-fidelity trade-off.

Far-field spatial coherence dictates the interference properties of scattered light and thermal emission. Traditionally, these phenomena are treated through disjointed paradigms: classical scattering descriptions assume cold objects lacking quantum fluctuations, idealized quantum scattering schemes ignore dissipation, and semiclassical fluctuational electrodynamics relies on phenomenological noise currents, precluding the consistent treatment of incident quantum states. Here, we develop a first-principles framework based on the modified Langevin noise formalism to unify the scattering of quantum light and the intrinsic thermal emission of finite dissipative objects. We demonstrate that the outgoing far-field spatial coherence separates into an algebraic superposition of two geometry-driven mechanisms, coupled by the global unitarity of the radiation-matter dynamics. The first mechanism, elastic scattering, acts as a non-unitary spatial filter, mode-selectively attenuating and reshaping incident quantum correlations. The second mechanism, thermal emission, originates from localized material dissipation and projects the object's absorption profile into the far field, providing a quantum-vectorial derivation of the macroscopic van Cittert-Zernike theorem. Applying this framework across optical regimes, we determine operational bounds for lossy quantum photonics. Under chaotic thermal illumination, we analytically demonstrate thermal cloaking at equilibrium and show that a passive sink casts a structured thermal shadow geometrically identical to a primary emitter. Under coherent illumination, we derive a thermodynamic phase diagram bounding macroscopic phase correlations, demonstrating that subwavelength nanostructures undergo substantial coherence degradation compared to bulk objects. Finally, under spatially entangled illumination...

Information geometry has traditionally been formulated within the framework of Riemannian geometry and dual affine connections. In this work, we reframe this foundational structure by introducing the geometric machinery of symmetric teleparallel gravity. By requiring both curvature and torsion to vanish globally on the statistical manifold, we demonstrate that the fundamental properties of the information space can be entirely encoded into the non-metricity tensor. This approach allows us to distinguish the general $ξ$-parameterized space from the $θ$- (or $η$-) parameterized space, mirroring the relationship between conventional general relativity and symmetric teleparallel gravity. Specifically, the $θ$- or $η$-coordinates emerge as the special coordinates in the coincident gauge, where the connection coefficients vanish.

We investigate how degeneracies in quasi-de Sitter backgrounds, in the sense of Wands' duality, are reflected in real-space quantum correlations of primordial perturbations. Using the continuous-variable Gaussian formalism for coarse-grained scalar fluctuations, we construct the covariance matrix of a pair of spatially localized modes in inflationary spacetime, and extract the symplectic invariants of the system. For a generic Wands-dual pair of backgrounds, we find that while the individual entries of the covariance matrix are highly background-dependent, the symplectic eigenvalues -- and hence the entanglement entropy, mutual information, quantum discord and log-negativity -- all coincide for the two dual realizations. Our results unveil a new ''quantum-informatic symmetry'' of the de Sitter vacuum, according to which local linear entanglement witnesses constructed from coarse-grained fields cannot distinguish between Wands-dual inflationary histories, even though their background trajectories differ. We show that the special nature of the Wands-duality symmetry (of being local, scale-independent canonical transformations) is at the heart of this duality.

We develop a quantized chiral-field mechanism for chirality-induced spin selectivity (CISS). The corresponding quantum is a heliciton: a helical mode with phase coordinate $φ-qz$, screw momentum $\hbar q$, and energy $\hbarΩ_q$. A helical electron can absorb or emit this quantum, converting the static chiral vertex developed in our preceding work into an inelastic resonant scattering process. Using first Born scattering theory, we show that an incident two-spin-channel state generates two heliciton-assisted sidebands. Absorption converts the $\uparrow k$ channel into the $\downarrow,k+q$ sideband, while emission converts the $\downarrow k$ channel into the $\uparrow,k-q$ sideband. Thus the heliciton supplies both the screw momentum and energy needed to turn the handedness-conversion into a resonant spin-selective channel. The two sidebands inherit the same sampled-current overlap $J_χ(k)$ from the static theory, but acquire different kinematic weights and different resonance detunings. The sideband sector reaches full spin polarization at the respective isolated heliciton resonances, with $P_{\rm sb}(k,q)\simeq +1$ for $Δ_-(k,q)=0$ and $P_{\rm sb}(k,q)\simeq -1$ for $Δ_+(k,q)=0$. Reversing the screw handedness, $q\rightarrow -q$, interchanges the two sideband channels and reverses the polarization. No ad hoc spin-dependent potential is introduced. The spin selectivity comes from three ingredients: helical Dirac-current texture, quantized screw-symmetric environmental motion, and resonant exchange of screw momentum and energy. This identifies CISS as a heliciton-assisted resonance mechanism that produces spin polarization in the inelastic sideband sector.

We investigate the effect of interferometer delay and imbalance on the performance of a BB84 time-bin quantum key distribution system. We simulate the impact of interference visibility on system performance and measure the visibility of a pair of interferometers as a function of their relative time delay and intensity imbalance. In addition, our analysis highlights the effect of laser chirp on system performance.

Population transfer via chirped rapid adiabatic passage is studied using open quantum and semiclassical models, with and without the rotating-wave approximation. A time-dependent variational approach based on the multiple-Davydov D$_2$ trial state is employed to simulate the quantum models with an arbitrary finite mean photon number. We examine the accuracy of both the semiclassical field description and the rotating-wave approximation. Robust population transfer is identified over a wide parameter regime controlled by the laser spectral chirp and is found to be insensitive to the spin--phonon coupling strength, Gaussian pulse area, and energy gap of the two-level system.

Adaptive Derivative-Assembled Problem-Tailored variational quantum eigensolvers (ADAPT-VQE) represent one of the most promising approaches for quantum chemistry on near-term quantum devices. However, their optimization is slow and may stall due to vanishing parameters and redundant operators in the ansatz. In this work, we propose a simple strategy of operator elimination that removes non-contributing operators from the pool once they are detected, enabling the optimization to continue progressing toward convergence. We examine two variants, with and without pool restoration after elimination, and find that the former converges more smoothly and faster than the latter and the standard ADAPT-VQE. To capture dynamical correlations between the active space and its environment, we combine ADAPT-VQE with our recently developed downfolding approach, the one-body downfolding framework (OBDF). In OBDF, the bare molecular Hamiltonian in the active space is replaced by a correlated effective Hamiltonian that incorporates dynamical correlation effects outside the active space. We benchmark our implementation on a linear \ce{H_6} chain, an \ce{H_6} lattice, an \ce{H_6} ring, and the \ce{N_2} molecule using the OpenFermion simulator. Our results show that operator elimination significantly reduces circuit depth and iteration count, and that OBDF-ADAPT-VQE yields energies closer to the full configuration interaction (FCI) reference than the standard approach within the same active space.

Deterministic single-photon sources are indispensable core devices for quantum information technology, yet high-performance implementation remains a long-standing bottleneck for linear optical quantum computing. We propose a feasible scheme for deterministic single-photon emission based on a $\triangle$-type three-level atom coupled to a single-mode cavity, driven by two classical external fields, which is adaptable to both strong and ultrastrong cavity-atom coupling regimes. Under continuous-wave driving, the system achieves excellent single-photon characteristics: the normalized equal-time second-order correlation function reaches $g^{(2)}(0)\sim10^{-6}$, with a photon indistinguishability of $98.73\%$ and a state purity of $99.95\%$ in the strong coupling regime, while the ultrastrong coupling regime further suppresses $G^{(2)}(0)\sim10^{-8}$, yielding an indistinguishability of $99.10\%$ and a purity of $99.99\%$. For pulsed driving in the ultrastrong coupling regime, the source realizes superior performance, with an emission efficiency, indistinguishability, and purity of $99.96\%$, $98.98\%$, and $99.99\%$ under resonant conditions, and $100\%$, $95.91\%$, and $99.93\%$ under detuned conditions, respectively. The near-ideal optical performance of the proposed scheme provides a viable route for constructing high-quality deterministic single-photon sources, which offers a promising solution to the limitations of conventional single-photon devices and facilitates the further development of quantum information science and fundamental quantum optical research.

Compact objects with black-hole-like exteriors may hide new strong-field physics in their interiors, making their dynamical response a sensitive probe of gravity beyond General Relativity. We present an analytically tractable, gravitationally bound compact object with a genuinely new dynamical signature: under a minimal passive boundary prescription, its exactly controlled odd-parity sector exhibits purely dissipative relaxation poles, rather than the oscillatory modes usually associated with black holes and exotic compact alternatives. The object we study is a regular, vector-supported compact solution of a vector--tensor theory, matched without any surface layer to an exterior Schwarzschild geometry. Owing to its anisotropic stress, it can violate the Buchdahl bound and be continuously connected to the black-hole compactness limit. Its unusual response follows from a hidden chiral symmetry, which turns the perturbation problem into one-way transport rather than ordinary wave propagation. The exterior region alone has no conventional quasinormal-mode spectrum; instead, the regular interior and the matching conditions break the symmetry and quantize the fluctuation spectrum. We analytically compute the retarded Green function and susceptibility, and derive an effective membrane response by integrating out the object's interior. In the black-hole limit, the relaxation times diverge, the poles collapse toward zero frequency, and finite-frequency exterior perturbations decouple from the interior. Black-hole behaviour is therefore approached through the disappearance of relaxation modes, not through the emergence of ringdown.

The emergence of multiple superradiant phases in Dicke lattice models has attracted considerable attention in the quantum optics community. However, a unified understanding of the origin of multistability and its relation to different superradiant phases is still lacking. Here, we develop a configuration-based understanding to classify the superradiant phases in Dicke lattices. We show that photon hopping naturally organizes the possible superradiant configurations according to the lattice symmetry, providing a unified interpretation of the nonequilibrium phase diagram and the emergence of multistability. For the dissipative four-site Dicke lattice, we obtain the complete phase diagram and identify the coexistence of up to four stable superradiant phases. The proposed classification is further extended to five- and six-site lattices. Moreover, we demonstrate that the same configuration-based understanding also applies to the closed Dicke lattice, where the ground state uniquely selects one of the allowed configurations. Finally, we show that different configurations may belong to either same or distinct nonequilibrium universality classes in the dissipative Dicke lattice, while they share the same equilibrium universality class in the closed Dicke lattice. Our results provide a unified picture for understanding equilibrium and nonequilibrium superradiant phase transitions in Dicke lattices.

Gravitational wave (GW) polarizations are traditionally classified in a fixed frame ($E(2)$ classification), which does not account for how polarization patterns change under Lorentz boosts. In this work, we derive the explicit transformation laws for all six GW polarizations under longitudinal and transverse boosts. For gravity theories devoid of preferred frames, we propose a symmetry-based classification of the GW polarizations they admit. Among our key findings, we demonstrate that a propagating mode with five degrees of freedom strictly locks its longitudinal and breathing scalar amplitudes via the universal relation $A_l/A_b = -2(1-k^2/ω^2)$. For theories with a preferred frame, we analyze Bumblebee gravity and reveal that preferred-frame effects induce significant GW birefringence and observer-dependent polarization mixing. Crucially, we identify a novel vector-to-tensor polarization conversion mechanism, where vector modes in the preferred frame inevitably generate observable tensor polarizations for moving detectors, offering a new pathway to test Lorentz-violating gravity. Our framework provides a novel, observer-independent classification of GW polarizations and reveals previously unnoticed polarization mixing effects.

Hybrid quantum platforms based on highly excited Rydberg atoms coupled to nanophotonics devices offer a promising route toward scalable quantum networks and integrated quantum technologies. However, the close proximity of Rydberg atoms to dielectric nanostructures makes these systems particularly susceptible to uncontrolled surface electric fields that can lead to a degradation of the excitation process. Here, we experimentally investigate Rydberg excitation of laser-cooled $^{87}$Rb atoms via the evanescent field of an optical nanofiber in the presence of fiber-guided red- and blue-detuned light fields as used to trap ground state atoms in fiber-based dipole traps. We observe a time evolution of the Rydberg excitation spectrum when both the dipole trapping fields are on and the additional spectral features that appear can be suppressed by applying an external oscillating electric field to the system, strongly indicating that surface charge accumulation is responsible for the observed spectral feature. The experimental results are reproduced qualitatively by a model that incorporates DC energy level shifts arising from electric fields generated by charges deposited on the nanofiber surface. We identify Rydberg-ground state collisional ionization, which is enhanced by the dipole trapping fields, as the dominant mechanism for charge generation. These results provide new insight into charge dynamics at dielectric nanophotonic interfaces and establish practical guidelines for mitigating surface charge-induced electric fields in fiber-integrated Rydberg quantum systems.

Quantum contextuality is widely recognized as an essential non-classical resource underlying quantum technology, yet illuminating the precise mechanisms through which it translates into unconditional computational advantages remains an ongoing challenge. We demonstrate an exponential, noise-resilient memory advantage for quantum finite automata arising from graph-theoretic approaches to contextuality. We define a promise problem on an exclusivity graph $G$ for which any classical deterministic automaton acts as a non-contextual hidden variable model requiring at least $N=χ(G)$ states, where $χ(G)$ is the graph's chromatic number. In contrast, by exploiting a structural phenomenon we term \textit{representational contextuality}, a QFA solves this task using a memory of dimension at most $d=ξ(G)+1$, where $ξ(G)$ is the graph's orthogonal rank. This separation scales exponentially ($d=\mathcal O(n)$ versus $N=2^{Ω(n)}$) for Boolean-orthogonality graphs. Crucially, this memory advantage maintains an $\mathcal{O}(1)$ threshold against both depolarizing and coherent noise.

Superconducting quantum chips commonly utilize quarter-wavelength (λ/4) transmission line resonators as readout circuits. An analytical model for the accurate determination of resonance frequencies and coupling Q-factors of feedline-coupled superconducting resonators is introduced. The model leverages four-port microwave network analysis, integrating boundary conditions and conformal mapping techniques to compute even- and odd-mode impedances in edge-coupled coplanar waveguide (CPW) structures. Its versatility allows application to both planar and 3-D heterogeneous architectures, making it a powerful tool for resonator design. To validate the model, a test chip with λ/4 resonators of varying geometries is fabricated and measured in a cryogenic environment. Comparisons with finite element method (FEM) simulations and experimental measurements confirm the model's accuracy, with resonance frequencies and coupling Q-factors aligning closely across configurations. This proposed model facilitates the design of superconducting resonators in readout circuits for more effective, scalable, and adaptable quantum computing architectures.

Multi-core quantum computing architectures have emerged as a promising solution to the qubit scalability limitations of monolithic NISQ devices. Quantum algorithms are expressed as quantum circuits composed of single- and two-qubit gates. However, circuit scheduling in multi-core quantum systems remains largely unexplored. Reducing overall execution time (makespan), increasing core utilization, and hiding communication latency behind computation depends on effective scheduling. In this paper, we first introduce a layered scheduling approach as a baseline where quantum gates within the same layer are executed in parallel, while layers themselves are executed sequentially. We then propose a greedy scheduling strategy which schedules each gate as soon as all its dependencies and required resources are available. This allows fine-grained parallelism across cores. Our evaluation shows that on real benchmarks, greedy scheduling achieves an average 40% reduction in makespan and improvement in core utilization. The results suggest that the use of intelligent circuit scheduling to exploit parallelism can greatly enhance the speed of circuit execution in multi-core quantum architectures.

State-of-the-art two-qubit gates with trapped ions employ open-loop control that rely on simplified models to precompute control sequences. Our aim is to introduce closed-loop control for two-qubit gates to correct disturbances as they occur during the gate implementation. We introduce a spectator ion into the ion chain used for quantum logic processing, where it couples with the other ions through collective motional modes. The spectator ion's position is continuously monitored by driving dipole transitions and detecting the resultant fluorescence. We show that incorporating a spectator ion is feasible for linear Paul trap implementations and is expected to reduce the two-qubit gate Bell-state preparation infidelity by an order of magnitude with the deleterious effects of position monitoring being negligible compared to the thermal effects that exist in the system, even in the absence of the spectator ions. Mathematically, we describe driven ion-trap dynamics, including the spectator ion, by a stochastic quantum master equation involving the amplitude-modulation multimode-motional coupling gate, motional drift, thermal effects, recoil from photon scattering, spontaneous decay, and light shift. Our on-the-fly control method employs reinforcement learning with the reward function based on the actual geometric phase of the spectator ion. A key advantage of our approach is that we introduce a control method that involves `learning' and correcting disturbances happening in the trap on-the-fly, thus achieving high-fidelity gates. Our approach will lead to a significantly higher two-qubit gate fidelity at a reduced calibration overhead owing to the small parameter drift in the control system.

In this paper, we employ spinning test particles as probes to investigate the regulatory effects of particle and black hole parameters on the violation of the chaos bound in Kerr-Newman-AdS spacetime. Our results demonstrate that the chaos bound violation is governed by the interplay of spacetime geometry, electromagnetic forces, and particle dynamics. The particle spin modulates the direction dependence and parameter thresholds of the violation through its coupling with the orbital angular momentum, which contributes to the total angular momentum. The negative cosmological constant acts as a potential well: a larger value enhances the chaotic behavior. A competitive coupling exists between the black hole rotation and charge -- its prograde rotation exerts a stabilizing effect that can suppress or even completely quench charge-driven violations, while the charge serves as a condition for triggering the violation, with its effect modulated by the spin stabilization. In the Kerr-AdS limit, the violation occurs only when the black hole rotates opposite to the $z$-axis with a sufficiently large rotation parameter and a sufficiently small cosmological constant. In the RN-AdS limit, the violation condition is jointly determined by the charge and the cosmological constant, with electromagnetic repulsion more readily inducing the violation than electromagnetic attraction.

In a recent article, Hod [1] has concluded the non-existence, heretofore unnoticed, of a unified Thorne's hoop conjecture that holds for the quasi-local mass, $\mathcal{M}(r\leq R_{+})$ for static Reissner-Nordstrőm black hole, but does not hold for the spinning electrically charged Kerr-Newman black holes, when the same quasi-local mass is used. While the conclusion is correct and important, we wish to point out a curious exception for which the conjecture holds in a unified way for both the static and spinning \textit{tidal} charged $4d$ braneworld black holes of string theory, when the mass in the conjecture is the same quasi-local mass.

The Damour-Solodukhin wormhole (hereinafter DSWH) is known to mimic Schwarzschild black hole (hereinafter SBH) horizon in some properties. To act as a mimicker, the DSWH parameter $λ$ by definition is required to be extremely tiny, i.e., $λ\sim 0$. Our comparative analyses show that such a requirement may be too restrictive at least as far as the Bondi accretion profiles of the two objects, the DSWH and SBH, are concerned. Intriguingly, it turns out that some profiles of DSWH mimic those for the SBH near the horizon even at values of $λ$ considerably higher, i.e., $λ\sim 1$.

Quantum steering ellipsoid (QSE) provides a faithful representation of a two-qubit state. When extended to tripartite systems, the steerability from a trusted party to different receivers is subject to volume monogamy relations, which only constrain the total steerability but cannot individually eliminate the steerability of an untrusted third party, leaving a potential channel for information leakage via steering. Here, we show that this residual steerability can be completely suppressed by selectively engineering bound states in local qubit-environment subsystems, without compromising the steerability between trusted parties. Specifically, when bound states are formed in the subsystems formed by the trusted parties and their environments but absent in the untrusted one, the untrusted party's QSE volume decays to zero, while the trusted party's QSE volume remains finite. Our results establish selective bound-state engineering as a mechanism for extreme volume monogamy, with potential applications in secure quantum communication with an untrusted third party.

We present a variational derivation of the Wallis product and its reciprocal from the four-dimensional singular harmonic oscillator. The inverse-square interaction is absorbed into an effective angular parameter $ν$, so that the lowest exact energy in a fixed sector is $E_{4d,\mathrm{exact}}=\hbarω(ν+2)$. Motivated by the radial Kustaanheimo--Stiefel relation $r=ρ^2$ between the four-dimensional oscillator and the three-dimensional Coulomb problem, we use the quartic trial family $R_a(ρ)=Nρ^νe^{-aρ^4}$. The minimized variational energy yields an accuracy ratio governed by adjacent Gamma functions. In the large-$ν$ semiclassical limit, this ratio approaches unity. Restricting $ν$ to the odd sequence $ν=2n-1$ gives the standard Wallis product, whereas the even sequence $ν=2n$ gives its reciprocal form. The Coulomb-dual interpretation further relates the two branches to integer and half-integer effective angular sectors in the dual Coulomb/MICZ description. The result shows that Wallis-type infinite products persist under an inverse-square deformation of the oscillator and arise from a common Gamma-function structure in radial variational dynamics.

We overcome the long-standing noninvasive measurability (NIM) challenge in Leggett-Garg tests by exploiting the causal structure of quantum field theory (QFT). Our protocol uses three independent ensembles of the vacuum state, each measured by a different pair of observers at spacelike-separated events, yielding the three two-time correlators. By placing these events at positions $(0,0)$, $(τ,L)$, and $(2τ,2L)$ with $L>τ+2τ_0$, we rigorously ensure that no measurement can influence another. We investigate the vacuum state of a free massive scalar field in 1+1 dimensions, employing the dichotomic observable $Q(f)=\operatorname{sign}(φ(f))$ where $φ(f)$ is the smeared field. In the Heisenberg picture, the time evolution is absorbed into a translation of the time-window function, allowing us to derive the two-time correlation function $C(τ,L)$ and the Leggett-Garg parameter $K_3=2C(τ,L)-C(2τ,2L)$. For non-overlapping time windows, we find that the correlation function decays exponentially with $τ$ for a massive field. For overlapping windows, our numerical computation for a rectangular time window yields $K_3<1$ across the entire mass range, firmly establishing that the vacuum does not violate the LGI. Thus, under strict noninvasive conditions, the vacuum shows no violation of macrorealism, in stark contrast to its well-known violation of spatial Bell inequalities. Our spacelike-separated protocol provides the first LGI test in QFT with rigorously satisfied NIM, setting a methodological benchmark for future studies and highlighting the fundamental distinction between spacelike entanglement and temporal macrorealism in relativistic quantum fields.

Quantum illumination offers a significant theoretical advantage for target detection in high background noise environments. However, its practical deployment in free-space channels is hindered by atmospheric turbulence. Stochastic fluctuations in atmospheric turbulence inevitably degrade the quantum signature, rendering the real-time evaluation of quantum advantage under such dynamic conditions a critical yet unresolved challenge. To circumvent the reliance on costly direct turbulence measurements, we propose a physics-driven framework that integrates Kolmogorov-Arnold networks directly bridge macroscopic meteorological observations with microscopic quantum channel dynamics. Trained on 105,120 samples from 12 climatically diverse sites and validated on 26,280 unseen samples from three extreme boundary conditions (arid continental, tropical maritime, high-altitude plateau), our approach establishes a physically consistent mapping from standard meteorological variables to the temporal evolution of the quantum advantage. This end-to-end system dynamically quantifies the degradation of quantum advantage across diverse turbulence conditions. Our results provide a rigorous theoretical and data-driven pathway for environmental adaptation, facilitating the transition of quantum radar networks from proof-of-principle demonstrations to all-weather operational systems.

We develop a unified treatment of scattering and discrete spectra for the one-dimensional Dirac equation with scalar and vector interactions. Under the spin-symmetry condition, the coupled first-order Dirac system maps exactly onto an effective Sturm--Liouville (Schrö\-din\-ger-like) problem for a single spinor component. This mapping provides a convenient framework for analyzing transmission, reflection, and analytic continuation. As an explicit application, we consider effective interactions of hyperbolic Pöschl--Teller type and exploit supersymmetric quantum mechanics and shape invariance to obtain a closed-form expression for the transmission probability. The bound-state spectrum is then recovered from the poles of the analytically continued transmission amplitude, reproducing known results and offering a unified description of scattering and bound states. For the barrier configuration, we briefly comment on the resulting pole pattern in the complex momentum plane and its connection with resonance and quasi-normal-mode behavior. Moreover, we use the chiral transformation to relate the spin- and pseudospin-symmetry sectors and translate results between them without repeating the full derivation.

We study the problem of extracting a quantum complexity resource from a mixed Gaussian state of the multimode light. We present the first complete, certificate-checked solution to this problem in a genuinely coupled sector. We carry this out for the two-mode case, the smallest case in which modes are genuinely coupled. Even in this case the solution is highly nontrivial, and we rigorously prove that it cannot be given in a closed form.

Predicting the overlap of quantum states with specified low-energy subspaces is a key diagnostic for quantum many-body dynamics, with direct applications in state preparation, subspace-based algorithms, and the study of thermalization. We study the supervised prediction of subspace overlaps O_K between time-evolved states and K-dimensional low-energy eigenspaces of a 10-qubit Heisenberg spin chain following a local perturbation. We compare two quantum information extraction strategies: measurement-based learning, in which classical shadow features are processed by convolutional neural networks, and coherent quantum learning, in which quantum convolutional neural networks process the state directly. We further introduce physics-informed variants for both approaches, including Hamiltonian-aware shadows and QCNN gates aligned with the Heisenberg exchange structure. Across five dataset configurations spanning weak, moderate, and strong quench regimes, physics-informed QCNNs achieve stable performance, with mean test-set coefficients of determination R^2 = 0.753-0.846. Shadow-based methods show stronger regime dependence: they outperform QCNNs in the moderate-quench regime, reaching R^2 = 0.886, but underperform in weak and strong quenches at default shot budgets, where the best shadow results are R^2 = 0.615 and 0.672, respectively. Hardware validation on Quantinuum and IBM noise models shows that arbitrary state preparation is the dominant limitation, requiring approximately 2,044 two-qubit gates and causing near-complete depolarization before inference. These results identify a regime-dependent tradeoff between measurement-based and coherent quantum learning, with shadow methods excelling when the target remains locally accessible and physics-informed QCNNs providing more robust performance across dynamical regimes.

The production of magic states is studied in two settings. The first is the electroweak (EW) sector of the Standard Model (SM). The second is an extension featuring a new broken $U(1)$ gauge symmetry and a Dirac fermion charged under it. This setup resembles a dark $U(1)$ scenario, with the additional fermion playing the role of a dark matter candidate that annihilates into SM particles through its coupling to the new gauge boson. In the EW sector, the low-energy regime reproduces earlier magic production results obtained for Quantum Electrodynamics, whereas the high-energy and $Z$-resonance regimes generate new magic distribution functions and non-trivially reorganize the stabilizer state classes, with Bhabha scattering exhibiting the strongest sensitivity to electroweak effects. Also, a subset of fixed stabilizer states is identified, for which the magic distributions remain unchanged across the different energy regimes. In the dark sector, the main effect of the new massive mediator is the appearance of new magic distributions functions for Moller-like, Bhabha-like, and inverse pair-annihilation processes in the low-energy limit. These reach the maximal magic value at the SM-to-dark fermion mass ratios $m_f/m_χ\to 0$ and $m_f/m_χ\to 1.83929$.

We classify two-qudit finite-field Clifford dual-unitary gates and apply the classification to exact local masking and operator transport in homogeneous brickwork circuits. Under ordered one-qudit Clifford equivalence, the dual-unitary locus contains $q-2$ perfect-tensor cores, one rank-one core, and one SWAP core. The perfect cores are indexed by \[ δ=\det B=\det C, \] and are related to the ordered cross-ratio $λ$ of the associated $[4,2,3]_q$ MDS configuration through \[ λ=\fracδ{δ-1}. \] Homogeneous repetition separates these cores into five transport phases whose algebraic codimensions in $\operatorname{Sp}(4,q)$ are $0,1,2,3,4$. The one-site Weyl edge channels determine exact local-masking distances. Perfect-tensor circuits attain \[ d_1(t)=4t, \qquad d_2(t)=4t-2, \] whereas delayed erasers satisfy \[ d_1(t)=4t-2, \qquad d_2(t)=4t-4 \] for $t\geq 2$. Consequently, sufficiently short quantum messages are completely hidden from every one- or two-qudit output subsystem, even when the input is entangled with a reference, while remaining exactly recoverable from the full output. For $q=3$, we construct an explicit perfect-tensor Clifford gate from two inverse SUM gates. Exhaustive Weyl-support searches for $t=1,2,3$ reproduce the predicted masking distances, and a one-period Choi-channel calculation for a four-qutrit periodic circuit gives numerical residual leakage below $2\times 10^{-16}$. Under the coherent perturbation considered here, local leakage is linear in the perturbation strength, whereas the infidelity of recovery using the ideal inverse is quadratic near the perfect point.

We present a resolution of the Wigner's Friend paradox within a framework of quantum mechanics (QM) on the hybrid space RxQ_{p}, where Q_{p} denotes the field of p-adic numbers, regarded as a model of discrete microscopic space at the Planck-Bronstein scale. In this framework, wavefunction collapse is not an independent postulate but a dynamical consequence of the Schrödinger equation with non-local Hamiltonians: wavefunctions localize onto compact supports during measurement interactions, producing definite pointer readings without the intervention of observers or the exchange of information between subsystems. We model both Wigner and his Friend as classical apparatuses and show that each produces a definite reading through independent applications of the collapse mechanism, thereby eliminating the conflict between their descriptions of reality. The framework is consistent with the principal no-go theorems in finite- and infinite-dimensional Hilbert spaces associated with extended Wigner's Friend scenarios -- including those of Frauchiger-Renner, Brukner, Bong et al., and Guérin et al. -- since it requires no agents capable of recording or reasoning about outcomes, thereby vacating the observer-dependent assumptions that drive those theorems. We illustrate the collapse mechanism explicitly through a toy model of a particle in a box, comparing the standard description with the new one. The non-locality intrinsic to QM on L2(RxQ_{p}) permits realism at the cost of locality, and the Absoluteness of Observed Events holds in our framework without requiring observer independence.

Left alone, open quantum systems relax toward a Kubo--Martin--Schwinger (KMS) equilibrium state, yet the inner workings of this process remain opaque. It remains unclear why the intricate fourth-order time-convolutionless (TCL4) population generator reproduces the comparatively simple second-order stationary state corrections. Even more surprisingly, stationary-state corrections of such precision can be compressed into a simple virtual coherence pathway: an open quantum systems analogue of virtual transitions, in which populations communicate through coherences that are never occupied. Here we show that this simplification arises through a sequence of stationary-state-preserving transformations that progressively eliminate asymptotically redundant components while preserving the stationary state, ultimately yielding the virtual coherence pathway. This resolves the longstanding stationary-state problem of the Redfield equation and reveals that much of the apparent complexity of the TCL4 generator is asymptotically redundant.

We show that Bose--Einstein condensation produces a cusp in the time-of-arrival (TOA) statistics of a harmonically trapped gas released into free fall. In the semiclassical long time-of-flight regime, with $ε=σ_V/\sqrt{2gH}\ll1$, both the mean and standard deviation of the arrival time distribution, which are governed by the longitudinal velocity variance, remain continuous, but acquire a cusp whose one-sided slope ratio is universal within the ideal-gas far-field limit, $\mathcal{R}_\infty=2.5556\ldots$, and equals the trapped-gas specific-heat ratio $C(Tc^-)/C(Tc^+)$. Finite atom number rounds the cusp and weak interactions perturb it only weakly, leaving a measurable time-domain signature of condensation.

In this work, we examine the theoretical framework of modified teleparallel gravity with the inclusion of the boundary term in the action and investigate its cosmological implications by considering the power-law model $f(T, B) = -T + α(-B)^β$, with the aim of addressing the late-time accelerated expansion and the dark energy. In this context, $T$ denotes the torsion scalar and $B$ represents the boundary term, whose presence allows for departures from standard teleparallel dynamics and provides a unified description that connects torsion and curvature-based formulations, reproducing $f(T)$ and $f(R)$ gravity in appropriate limits. The viability of the model is assessed by confronting its theoretical predictions with observational data while constraining the cosmological and model parameters through a Markov Chain Monte Carlo (MCMC) analysis using cosmic chronometers (CC), the Pantheon Plus sample (PPS), and the DESI baryon acoustic oscillation (BAO) Data Release 2 (DR2) datasets, and comparing its performance with the standard $Λ$CDM model. The Akaike Information Criterion (AIC) analysis shows that the combined CC+PPS dataset strongly favors the $f(T, B)$ model, suggesting an improved phenomenological fit to late-time observations relative to $Λ$CDM. Our result further shows an alleviation of H0 tensions, although a dedicated analysis is required to establish its full statistical significance. Furthermore, the background cosmological quantities indicate that the model exhibits a dynamical phantom-divide crossing while remaining consistent with late-time observations and yielding a viable expansion history and characteristic dark energy evolution.

We extend our study of a cosmological scenario in which dark matter is non-minimally coupled to gravity at the fluid level. In previous work, we showed that this interaction can drive an early phase of accelerated expansion, addressing the horizon and flatness problems, and can also lead to a cosmological bounce in the presence of spatial curvature. Here we analyse the evolution of linear perturbations in this framework. We derive the equations governing scalar, vector and tensor perturbations, and obtain analytic solutions in the relevant cosmological regimes. We find that perturbations generated during the accelerated expansion phase produce a strongly blue scalar power spectrum and are therefore incompatible with observations. By contrast, in bouncing solutions primordial fluctuations can originate during the contracting phase before the bounce. In this case, the model yields an approximately scale-invariant scalar power spectrum while keeping the tensor-to-scalar ratio compatible with current bounds, without introducing additional scalar fields. Although our treatment relies on simplifying approximations that should be refined in future work, these results indicate that non-minimally coupled dark matter may provide a viable alternative mechanism for the generation of primordial cosmological perturbations.

We investigate the cosmological implications of the mass-to-horizon relation, which provides a unified framework for thermodynamically consistent generalized horizon-entropy functionals. Using the Cai-Kim formulation of the first law of thermodynamics, we derive the corresponding modified Friedmann equations and examine the resulting background evolution. We find that cosmological viability sharply restricts admissible deviations from the Bekenstein-Hawking area law: phenomenologically acceptable scenarios are confined to a narrow neighborhood of the standard entropy, while more pronounced deviations generically spoil the standard radiation-matter-dark-energy sequence. Power-law entanglement corrections can give rise to a moderate early-dark-energy component, but only within a tightly constrained region of parameter space, whereas quantum-gravity corrections are suppressed by the Planck scale and remain observationally irrelevant. Consequently, all viable models predict a $Λ$CDM-like cosmological background at the present epoch. These findings demonstrate that background cosmology alone imposes stringent constraints on thermodynamically consistent generalized entropy constructions of this class.

Semiconductor hole spins offer a pathway to extended coherence times by decoupling from nuclear magnetic noise, while their spin-orbit coupling enables fast all-electrical control. In ZnSe, however, realizing this potential has been limited by p-doping challenges. Here, we circumvent this limit by optically activating acceptors within the ZnSe quantum well. We isolate a single-hole spin bound to a shallow acceptor, confirmed by antibunching and accessed via the fast (244 ps) radiative recombination of a bound exciton. Magnetic and Raman spectroscopy of the ground state reveal an effective hole g-factor of 0.7 and an optical resonance linewidth of 26.7 GHz. Complementary first-principles simulations, together with the experimental results, provide evidence that points toward nitrogen as the most likely acceptor impurity. These results introduce a promising new platform for optically active spin qubits and single-photon sources in ZnSe.

The Obara-Saika (OS) method is one of the most widely used techniques in quantum chemistry for evaluating electron repulsion integrals (ERIs) via a set of recurrence relations that build higher angular momentum integrals from lower-order ones. The original derivation by Obara and Saika proceeded by directly relating integrals of differing angular momentum. In this work, we present a compact novel derivation of the OS vertical recurrence relation based solely on differential relations between Gaussian basis functions, expanding on a method suggested in earlier work. By explicitly deriving the required derivative expressions we identify all non-zero primitive terms contributing to the full ERI to develop a hierarchical formulation of the OS recursion relations. This approach has pedagogical value as a rigorous and self-contained derivation. Additionally, the resulting organization exposes independent primitive derivative quantities and may be useful for code generation and parallel implementations on modern GPU architectures.

Recently, Sharoglazova et al. [Nature 643, 67 (2025)] proposed a procedure for determining the speed of a quantum particle in the classically forbidden region of a potential step, and implemented it in a beautiful experiment. The inferred speeds disagree significantly with the Bohmian velocities, which the authors presented as an experimental challenge to Bohmian mechanics. This is puzzling because the speeds are inferred from particle populations in coupled waveguides for which Bohmian mechanics and standard quantum mechanics make identical predictions. We resolve the puzzle by a detailed theoretical analysis of the experimental setup. We show that the speed inference rests on an assumption that fails in the relevant (evanescent) regime according to both Bohmian mechanics and standard quantum mechanics -- namely that the inter-waveguide tunneling time is set by the transverse coupling and is not affected by entanglement with the longitudinal degree of freedom. We also consider a second speed estimate suggested by Sharoglazova et al., which is based on the Büttiker dwell time formula for particles in the forbidden region. We show that the authors applied the formula incorrectly, and that a correct application yields exact agreement with the predictions of Bohmian mechanics. Our analysis includes explicit calculations of Bohmian trajectories, dwell times, and longitudinal speeds in the two-dimensional waveguide model of the experiment.

We present a counterexample to Ostrogradsky's famous "no go" theorem as usually interpreted in quantum field theory (QFT), namely a four-derivative, UV-complete QFT with a consistent perturbative expansion which describes high energy scattering processes. We carefully quantize the theory on an $\textit{indefinite}$ space of states - a Krein space - using covariant methods which ensure perturbative causality and unitarity (in the form of the optical theorem) to all orders. We generalize the Born rule to Krein spaces and prove that all tree level transition probabilities are positive in spite of the presence of ghosts. A key role in the proof is played by a hidden "ghost parity" symmetry which becomes explicit when the theory is embedded in a two-derivative, two-field $O(1,1)$-symmetric perturbative field theory.

Maxwell-Chern-Simons (MCS) theory in $2+1$ dimensions provides a paradigmatic example of a topological gauge theory with both dynamical and topological degrees of freedom. Its Euclidean formulation suffers from a sign problem, making Hamiltonian numerical approaches particularly attractive. As a first step toward the non-perturbative Hamiltonian study of MCS theory, we investigate the constant mode sector on a spatial torus. Being analytically solvable in the continuum, it provides an ideal benchmark for understanding how the topological properties of the theory are encoded in a finite-dimensional lattice Hilbert space. We construct a finite-dimensional discretization of the torus of flat connections and show that the resulting lattice problem maps onto a generalized Harper-Hofstadter model with twisted boundary conditions. We identify the commensurability conditions under which the finite lattice exactly reproduces the magnetic translation algebra and the topological degeneracy of the continuum theory. A systematic analysis of gauge field truncation and its convergence toward the continuum limit is then presented.

Silicon spin qubits are promising for large-scale quantum-computer integration because they can fully leverage the well-developed semiconductor infrastructure. However, the low fidelity of two-qubit entanglement gates remains a key barrier to large-scale integrations. Recent simulations of silicon spin-qubit two-qubit gates have been performed on silicon-on-insulator (SOI) platforms, while nanosheet-based charge-qubit work has been limited to single-qubit operation using a two-dimensional Schrödinger approximation. In this work, we study silicon spin-qubit double quantum dots built on nanosheet technology using the Quantum Technology Computer-Aided Design (QTCAD) simulation suite to run three-dimensional Poisson and Schroedinger solvers, followed by a many-body solver to extract exchange interactions. We evaluate the exchange energy sensitivity to process and bias variations and then use QuTiP to solve the master equation for a two-qubit gate. The results show that millivolt-level bias variations at the plunger and middle barrier gates can reduce the gate fidelity below 99%, a common threshold target for many fault-tolerant quantum-computing algorithms. Gate-referred 1/f charge-noise effects are also analyzed through the resulting coherence time.

Fast analytic and semi-analytic studies of neutron stars often require an equation of state that is convenient to evaluate while producing relativistic stellar sequences compatible with current multimessenger constraints. We construct such a benchmark by scanning a smooth double polytropic relation for the energy density as a function of pressure, $\hatε(\hat p)=a_1\hat p^{Γ_1}+a_2\hat p^{Γ_2}$. The parameters are selected with filters based on massive pulsars, tidal deformability from the binary-neutron-star event GW170817, and NICER mass-radius measurements. A single polytropic baseline scan finds no model passing all filters, whereas a double-polytrope scan identifies a viable region. A curve-integral score, evaluated against public NICER and GW170817 posterior data sets, is then used to choose benchmark equations of state within this region. The selected representatives support $M_{\max}=2.44$--$2.49\,M_\odot$, with $R_{1.4}\simeq 11.3$ km and $Λ_{1.4}=485$--$512$, and remain causal on the stable branch. This compact analytic family provides reference cases for relativistic stellar-structure tests at current observational scales.

The axion represents a strong candidate for weakly interacting dark matter. To date, high sensitivity lab based experiments and astrophysical observations have ruled out a substantial part of the axion mass and photon coupling parameter space. However, a challenge remains in searching for the presence of the axion in the higher mass range 0.01-1eV corresponding approximately to axion field oscillation at THz frequencies. This work investigates via numerical simulation the feasibility of a high sensitivity, lab-based axion sensor operating in this range, based on plasmonic electric field enhancement by a nanostructured metasurface, combined with heterodyne detection and quantum sensing via nitrogen-vacancy (NV) centers in diamond. Estimates of the sensor response to anomalous electromagnetic fields resulting from axion coupling are given using Ti/Au nanopillars on LiNb at axion mass corresponding to telecommunications wavelength ($\approx$0.8eV, 196 THz). Finally, the possibility of sensing in the lower axion mass $<$10$^{-2}$ to 10$^{-1}$eV range is explored using alternative materials, with CdTe as an example.

The gate and qubit requirements of quantum computations of electronic structure have been extensively studied. However, the quantum resources present in electronic ground states, as measured by entanglement and magic, remain less well understood. We study the relationship between correlation in electronic structure Hamiltonians and magic as measured by the 2-stabilizer Renyi entropy (2-SRE). Perturbative calculations show that the 2-SRE of a given state is proportional to its overlap with a reference stabilizer state. In the context of quantum chemistry, this links the magic of electronic structure ground states to their Hartree-Fock weight, an established measure of electronic correlation. We then show that the 2-SRE of post-Hartree-Fock ground states is proportional to the correlation energy they recover. We explore this connection through the contextual subspace (CS) method. We present a theoretical framework showing that the CS method can be used to monotonically vary the magic of approximate CS ground states, and we prove that the correlation energy recovered by the CS ground states is proportional to the magic present in the approximate ground state. We present simulation results using 190 molecular species under Jordan-Wigner encoding at a range of bond lengths. The linear relationships between magic and correlation are robust across the Hamiltonians in our dataset, but break down at bond lengths beyond the Coulson-Fischer point, where Hartree-Fock fails to capture key physical features of the true ground state wavefunction. By establishing linear relationships for both correlation energy and Hartree-Fock reference weight with the 2-SRE, we conclude that for weakly- and moderately-correlated electronic structure Hamiltonians, the correlation is directly represented by 2-SRE, and thus by the magic.

Two-qubit-gate error dominates the failure budget of near-term quantum circuits, so the decomposition chosen for each Toffoli (CCX) gate should minimize hardware two-qubit infidelity, not gate count. The cheapest decompositions - relative-phase and approximate Toffolis - are only correct in context: their residual phase or bounded error must be cancelled or absorbed downstream. We present the first compiler pass that selects a per-Toffoli decomposition to minimize a two-qubit-infidelity error budget. It admits each context-dependent decomposition only when an exact, instance-specific equivalence check certifies its validity in that circuit context, coupling an error-budget objective with per-instance verification and closing the gap between context-aware-but-unverified and verified-but-context-free optimizers. The central result is a safety one: pattern-matched relative-phase substitution is silently incorrect. Our verifier flags 66 library rewrites of a deployed open optimizer as non-equivalent without a context check, and count-greedy substitution silently corrupts 6 of 12 benchmark circuits; the verification gate certifies 0 errors while still applying every valid decomposition. The two-qubit-gate reduction is real but workload-dependent: up to 39.5% fewer two-qubit gates and 36.7% lower infidelity over exact-only on a compute/uncompute-heavy suite (approx. 39%/35% versus Qiskit opt-3 and tket), and 15.6% aggregate on a larger 12-24-qubit suite, with decision-diagram checking certifying every substitution past the exhaustive-verification limit. At current superconducting and trapped-ion error rates, the certified substitutions lower estimated circuit infidelity by 36-43%, and on a quantum state-resetting circuit, the pass removes 48.8% of the native two-qubit gates, every substitution verified.

In Ref. 1, a method was given for self-consistently generating sum-of-squares decompositions of quartic fermionic Hamiltonians. Perturbation theory was used to generate a useful choice of cubic operators in this sum-of-squares. On a range of model problems, this method, which is only a fragment of degree-six sum-of-squares, was able to outperform the full degree-four sum-of-squares in both speed and accuracy. Unfortunately for applications, many problems in chemistry have strong density-density interaction terms, as well as moderately strong density-dependent hopping and spin-spin interaction terms, limiting the power of the perturbative choice of the cubic operators. Here we propose a method for generating these decompositions in the presence of these strong interaction terms, hopefully extending the range of applicability of this method.

Gravitational wave observations are beginning to probe the upper edge of the pair-instability supernova (PISN) black hole mass gap, a key prediction of stellar evolution. In this work, we quantify the sensitivity of this boundary to uncertainties in stellar evolution using a suite of simulations that vary inputs including nuclear reaction rates, mixing processes, and stellar winds. We find that the $^{12}{\rm C}(α,γ)^{16}{\rm O}$ reaction rate is the dominant source of uncertainty, shifting the upper edge by $ΔM\sim30\,{\rm M}_\odot$, with the triple-$α$ rate producing a comparable shift of $\sim25\,{\rm M}_\odot$. Notably, $^{16}{\rm O}+^{16}{\rm O}$ reactions shift the upper edge by $\sim15\,{\rm M}_\odot$ while leaving the lower edge unchanged, implying they can widen or narrow the mass gap. Other processes affect the location at the $\lesssim10\,{\rm M}_\odot$ level. In contrast to the lower edge, we find that the upper edge is robust to variations in spatial and temporal resolution, indicating that it is reliably resolved in current simulations. Our results demonstrate that the upper edge carries substantial theoretical uncertainty and, while comparatively less affected by astrophysical contamination than the lower edge, provides a direct probe of the nuclear processes governing pair instability. We discuss the implications for interpreting high-mass black hole detections in gravitational wave data.

Recently, Doosti et al. introduced the problem of distributed quantum state verification, where $m$ distributed nodes are given a copy of an unknown state $ρ$, and can send limited one way communication to a central node, who has a complete description of a known state $σ$. They ask how many distributed nodes $m$ are required, before the central node can succeed at distinguishing whether $ρ=σ$ or $\|ρ-σ\|_1\geq\varepsilon$ with high probability. In the setting where only quantum communication is allowed, Doosti et al. exhibit conditional lower bounds in both the public and private-coin settings, and a matching upper bound in the public-coin setting. We extend these results, and show unconditional lower bounds for when both classical and quantum communication are permitted. We show the public-coin lower bound is tight by giving an algorithm with a matching upper bound. We also show an almost tight upper bound in the private-coin setting when only quantum communication is permitted.

The circular Unruh effect is the excitation of a detector moving along a planar circular trajectory within an electromagnetic vacuum. We demonstrate that the magnetic dipole transitions in an atom, acting as the detector, dominate the electric dipole transitions. Our analysis of both free-space and cavity schemes shows that the sensitivity to the circular Unruh effect can be maximized by balancing the minimization of mode volume against the resulting decrease in mode density. Moreover, we propose a novel measurement scheme that uses the atom's multilevel structure to suppress the spontaneous emission rate, thereby enabling the experimental detection of the circular Unruh effect.

Electron and positron can form a meta-stable bound state called positronium that decays via pair annihilation. We show how polarization-dependent Compton scattering can be used to verify that the two annihilation photons in the spin-zero case (para-positronium) are emitted in a maximally entangled Bell state. Our theoretical approach based on two-photon density matrices connects concepts from relativistic quantum electrodynamics and quantum information theory.

Following the holographic proposal that identifies the growth rate of Krylov complexity with the proper radial momentum of an infalling massive probe, we study Krylov complexity in Lifshitz and hyperscaling-violating backgrounds. For pure Lifshitz geometries, we derive exact analytic solutions and obtain quadratic complexity growth for all values of the dynamical exponent. For hyperscaling-violating backgrounds, we extract the asymptotic scaling, revealing that the hyperscaling-violating exponent directly controls the late-time growth exponent. In a special limiting case, the complexity exhibits oscillatory behavior with a logarithmic envelope, signaling a transition to a qualitatively distinct regime. Our analysis establishes that the momentum-Krylov correspondence extends naturally to non-relativistic holographic settings and remains well-defined despite the causal pathologies of Lifshitz spacetimes.

Merging optomechanics and polaritonics opens stimulating perspectives like the giant enhancement of optomechanical interaction and the enrichment of optomechanics with effective nonlinear photons. The experimental implementation of these concepts has however remained elusive. Here we report on the resonant optical control of polaritonic optomechanical resonators constituted of semiconductor disks embedding quantum wells. Whispering gallery photons and quantum well excitons strongly couple, leading to the emergence of polaritons that couple to the mechanical vibrations of the disk. We perform resonant optomechanical frequency response experiments on these resonators, modeled introducing a minimal set of constitutive equations, from which we extract the polariton-modified optomechanical coupling $g_0$ and the polariton nonlinearities. We observe a boost of $g_0$ by more than a decade compared to bare photons, reaching to a record $g_0$ for whispering gallery resonators of $22$ MHz, and analyze experimentally and theoretically its evolution as function of the polariton's composition. We also measure a clear hierarchy of three polaritonic nonlinearities, again analyzed as function of polariton composition, establishing a bridge between past unconciliated reports in polaritonics. Grounded on experimental and theoretical foundations, resonant polaritonic optomechanics is set ready for an optomechanical exploration of quantum fluids of polaritons.

We develop a geodesic-optics description of eikonal quasinormal ringing, blackhole shadows, strong lensing, grey-body factors (GBFs), and the binding energy of massive particles for the asymptotically flat regular black-hole geometry obtained in phantom Dirac-Born-Infeld (DBI) gravity. The null-orbit structure admits an especially compact analytic treatment. The unstable photon orbit remains at the Schwarzschild-like coordinate radius throughout the black-hole branch, while the orbital frequency and Lyapunov exponent coincide exactly. Consequently, eikonal quasinormal modes (QNMs), shadow radius, GBFs, and strong-deflection observables are all governed by a single dimensionless function of the core-size parameter. For timelike circular motion, we derive exact expressions for the specific energy and angular momentum, obtain the innermost stable circular orbit (ISCO) condition in closed implicit form, and show that the ISCO binding efficiency decreases as the regular core grows. We present illustrative plots for the exact geodesic invariants, the corresponding grey-body profiles, and the timelike binding-energy curves. The resulting construction provides an exact one-parameter bridge between the regular black-hole metric and its leading geodesic observables.

We propose a quantum collocation framework for approximating solutions of one-dimensional linear and nonlinear boundary value problems. The method formulates the search for admissible solutions as a residual-based quantum search over a discretized ansatz space, where candidate solutions are evaluated through residual conditions imposed at collocation points. A residual-threshold oracle is constructed that acts jointly on spatial and parameter registers. This joint oracle structure leads to amplification dynamics that decompose into a coherent superposition of spatially conditioned amplitude-amplification processes rather than a single global amplification mechanism. We derive the corresponding amplification geometry and show that the success probability is governed by a weighted combination of spatially dependent amplification angles. Furthermore, we prove that the reversible residual oracle can be implemented with gate complexity polynomial in the logarithm of the number of collocation points, while retaining the quadratic search acceleration associated with amplitude amplification in the parameter space. We analyze how the spatially dependent oracle structure influences the amplification dynamics and corresponding success probabilities. Furthermore, we investigate how discretization, ansatz expressivity, oracle tolerance, and finite-precision effects influence both approximation quality and amplification behavior. Numerical experiments validate the theoretical predictions and illustrate the resulting search dynamics across different discretization and precision regimes.

Exceptional points, where eigenvalues and eigenvectors coalesce, are a defining feature of non-Hermitian systems and have been extensively observed in photonic, atomic, and condensed matter systems. However, they have received little attention in quantum chromodynamics (QCD), which is the fundamental theory of quarks, gluons, and hadrons. We propose that imaginary magnetic fields provide a simple realization of non-Hermitian dynamics in hadronic systems. Based on two theoretical approaches, a hadronic effective Lagrangian and a constituent quark model, we compute mass spectra of neutral mesons and find exceptional points separating the real-spectrum and complex-eigenvalue regimes. In small fields, the real spectrum exhibits level attraction between hadronic states, whereas in larger fields, hadrons are deconfined, which is a signature of a field-induced inverted potential. Our findings open a new avenue for studying QCD dynamics in non-Hermitian environments.

Gravity-induced entanglement has been proposed as a method for testing the non-classical nature of gravity via tabletop experiments. While most existing proposals are restricted to the Newtonian limit, the frame dragging effect offers access to genuinely post-Newtonian features of the gravitational interaction and remains comparatively less explored. Here, we study gravity-induced entanglement generated by frame dragging in an interferometric setting and compute the entanglement phase between the rotational degrees of freedom of a source mass and the paths of a particle in two complementary ways: (i) via Schrödinger evolution with a quantized Lense-Thirring Hamiltonian in the large angular momentum limit, and (ii) via the on-shell action of linearized quantum gravity within the stationary phase approximation. Both approaches yield the same entanglement phase, consistent with the proper time difference between the interferometer arms. The path integral derivation further reveals how gravitational retardation modifies the entanglement phase, thereby making the local, relativistically causal linearized-gravity description explicit. Under the standard locality/mediator assumptions used in existing arguments, the resulting entanglement would witness non-classicality of the gravitational interaction.

Periodically driven quantum systems offer remarkable flexibility in tailoring effective Hamiltonians and synthetic band structures. However, such driving also induces heating and dynamical instabilities that limit the coherence and lifetime of many-body states. Here, we demonstrate that these instabilities can be suppressed by employing multi-tone driving schemes. Using a Bose-Einstein condensate of cesium atoms in an optical lattice, we experimentally explore two approaches: pulsed driving composed of odd harmonics and two-tone driving with tunable amplitude and relative phase. We show that both methods allow independent control of the effective tunneling amplitude and Peierls phase factor, while significantly reducing phonon excitation and the resulting rapid decay of the condensate. Numerical simulations and theoretical modeling based on Bogoliubov-de Gennes equations confirm the suppression of unstable modes under optimized driving conditions. Our results establish multifrequency drives as powerful tools for stabilizing driven many-body systems and pave the way toward robust Floquet engineering with interactions.

We correct a subtle oversight in Eq.(19) of K. Lorek et al. [Class. Quantum Grav. 32 175003 (2015)] concerning the decay probability of a uniformly accelerated quantum clock. The original expression included unphysical cross-wedge term ($|\barγ_{K 1}|^2$) violating the spatial separation of Rindler wedges. We derive the corrected probability using the paper's Rindler conformal coordinates $(τ, ξ)$, retaining only causally consistent thermal terms, and discuss implications for relativistic clock models.

The $5S_{1/2} \rightarrow 6P_{3/2}$ transition of rubidium at 420 nm is a promising candidate for a portable warm-vapor all-optical atomic clock. Despite recent precision spectroscopy studies at 420 nm in Rb, an experimental determination of the saturation intensity of this transition has not yet been reported. The saturation intensity is a fundamental parameter that influences the identification of a potential clock transition frequency in terms of optimizing various intensity-dependent parameters and connected systematics. In this work, we report the first experimental measurement of the saturation intensity of the 420 nm transition in Rb, obtaining $(23.18 \pm 0.28)$ mW/cm$^2$ for the $^{87}$Rb $F=2\rightarrow F'=3$ transition and $(25.56 \pm 0.37)$ mW/cm$^2$ for the $^{85}$Rb $F=3\rightarrow F'=4$ transition, in excellent agreement with theoretical predictions. We further investigate the temperature dependence of the Doppler-free Lamb-dip amplitude and linewidth over 59.03$~\pm~0.37$ - 91.20$~\pm~0.90^\circ$C in a 100 mm commercial vapor cell, identifying around 82.02$~\pm~ 0.73^\circ$C as the optimal operating temperature, where the signal-to-noise ratio of the Lamb-dip amplitude with temperature reaches a maximum and the observed Lamb-dip linewidth exhibits a minimum. We also present precise measurements of the magnetic-dipole ($A$) and electric-quadrupole ($B$) hyperfine constants of the $6P_{3/2}$ state for both isotopes, with the measured values being consistent with previously reported values for the hyperfine constants.

We use the relativity principle to complete axiomatic reconstructions of quantum mechanics (QM) via information-theoretic principles that are based on Darrigol's "discreteness" requirement or its equivalent, e.g., Brukner & Zeilinger's Information Invariance & Continuity or Khrennikov's quantum action invariance principle. In this approach to the quantum reconstruction program (QRP), the Hilbert space kinematics of QM is derived most fundamentally from the experimentally motivated postulate of "discreteness," rendering QM a principle theory as defined by Einstein. Special relativity is also a principle theory, since its Lorentz transformation kinematics is derived from the experimentally motivated light postulate. While special relativity has a compelling fundamental principle (relativity principle) to account for its experimentally motivated light postulate, QRP has not produced a compelling fundamental principle to account for its experimentally motivated "discreteness" requirement. We complete QRP by showing how the "discreteness" requirement is necessitated by the relativity principle and Planck's radiation law, just like the light postulate is necessitated by the relativity principle and Maxwell's equations. Accordingly, the fundamental explanans for time dilation, length contraction, quantum superposition, and entanglement is the relativity principle. Phenomenologically speaking, the relativity principle is "the equality of all perspectives." Thus, quantum entanglement isn't evidence for the violation of intersubjective agreement or the need to abandon factive or objective models of reality, as in relational interpretations of QM. It is evidence that reality is not a fragmented collage of different subjective experiences from different perspectives, it's a comprehensive and coherent integration of those different subjective experiences per the equality of all perspectives.

In this paper we prove, both mathematically and through a simulation, how the Quantum Amplitude Estimation algorithm can obtain quadratic improvements with respect to the Monte Carlo method in gradient-based stochastic optimization, highlighting the central role of the Quantum Phase Estimation concentration guarantee in achieving the predicted advantage.

Quantifying non-stabilizerness in mixed states is provably intractable, as any strict monotone requires superexponential time. We propose a linear Stabilizer Entropy that acts as a proper non-stabilizerness monotone with overwhelming probability when restricted to non-adaptive Clifford channels acting on flat mixed stabilizer states. Analytical and numerical results for Haar-random states, Clifford orbits, and random matrix product states show that monotonicity violation probabilities decay as $\exp-ηN$. We also prove the validity of Stabilizer Entropy in specific many-body systems undergoing partial measurements, where the amount of resource never increases for each measurement outcome as well as when averaged over outcome probabilities. Given the hardness of strict alternatives, Stabilizer Entropy emerges as a practical and theoretically justified resource measure.

Michel accretion becomes transonic at the saddle point of a dynamical system. An Eulerian perturbation on the steady inflow produces the metric of an acoustic black hole. As a high-frequency travelling wave the perturbation does not destabilize the steady inflow. Acoustic waves propagating outwards against the fluid inflow are blocked at the sonic barrier but can tunnel through it with an exponentially decaying amplitude. The Hawking temperature and the frequency of the Hawking phonons are enhanced by the spacetime geometry.

The simplest point-particle description of classical matter is incompatible with Einstein's General Relativity because the stress-energy tensor of a point particle is distributional and concentrated on a one-dimensional worldline. For such higher-codimension sources, smooth spacetime solutions generally do not exist. This obstruction was established by Geroch and Traschen for sources of codimension $\geq2$. Motivated by this result, this thesis proposes codimension-zero tubes as a fundamental description of gravitating matter. Timelike tubes are constructed within the tubular neighbourhood of an auxiliary timelike curve. The tube interior is foliated by timelike codimension-one hypersurfaces whose dynamics are governed by a brane-like action. The resulting collective stress-energy tensor is smooth, unlike that of a point particle. For a broad class of tension and potential profiles, the strong energy condition is violated inside the tube, while the null and weak energy conditions remain satisfied. In the ultraviolet limit, where the tube radius vanishes, an appropriate rescaling of the Lagrangian density reduces the tube action to the point-particle action together with a canonical self-force-like term. The particle's rest mass then emerges as an effective quantity rather than a fundamental localized parameter. Perturbative stability is analysed at two levels. Field perturbations yield an infinite squared sound speed, showing that the foliation-generating scalar is non-dynamical and cuscuton-like. Small deformations of the leaves lead to the Jacobi equation for timelike hypersurface congruences, further constraining admissible tension and potential profiles. These results establish gravitating tubes as a geometrically and dynamically consistent description of matter that respects the Geroch--Traschen obstruction.

We study variational quantum algorithms for the Maximum Independent Set (MIS) problem on benchmark graphs of 64, 99, and 180 vertices. The Variational Quantum Eigensolver (VQE) and Quantum Approximate Optimization Algorithm (QAOA) are compared across SPSA and COBYLA optimizers at multiple circuit depths. A preprocessing pipeline comprising spectral graph reordering (via the Fiedler vector) and distance-based sparsification reduces circuit depth while preserving energy fidelity. Classical post-processing via history-guided bitstring correction and stepwise maximality extension recovers the exact MIS across all instances. With CVaR optimization, VQE with SPSArecovers up to 6 distinct MIS per run for the 64-node instance and up to 10 distinct MIS per run for the 99-node instance, sampling broadly from the optimal solution population. Repeated runs with different SPSA trajectories collectively enumerate a larger fraction of all MIS for each instance. For the 180-node instance, where standard approaches stall at size 14 (MIS is 15), we introduce ancilla-assisted superposition initialization: ancilla qubits prepare a uniform superposition over classically-found near-optimal solutions, and an excitation-preserving ansatz evolves this state while conserving Hamming weight. This novel construction enables quantum-parallel variational search over multiple seeds simultaneously, discovering the exact MIS where single-seed methods fail. The 180-qubit simulation represents, to our knowledge, the largest scale at which gate-based variational algorithms have solved MIS to optimality. Hardware validation on IBM Quantum hardware ibm_marrakesh confirms that converged simulator parameters transfer effectively to noisy quantum execution.

We present the first calculation of the next-to-leading-order electroweak corrections to vector-boson scattering into doubly polarised Z bosons at the LHC in the fully leptonic decay channel. The production and decay of the two polarised Z bosons are consistently modelled in the double-pole approximation, separating polarisation states at the amplitude level and including factorisable real and virtual electroweak corrections. Doubly polarised and unpolarised signals are investigated and confronted with off-shell results. A broad analysis, including results at integrated and differential level, is carried out in a realistic, CMS-inspired fiducial setup. Our study paves the way to upcoming analyses with LHC Run-3 and High-Luminosity data as well as to further phenomenological investigations.

HyperFORM brings the parametric integration of hyperlogarithms, weighted by rational prefactors, into the symbolic-manipulation system FORM. It ports the capabilities of Erik Panzer's Maple package HyperInt, capitalizing on FORM's speed with bulky algebraic input and on its ability to spread a single calculation across many processor cores. We keep the description of the method brief and concentrate instead on how the package is organized and driven: a fully self-contained program for the three-loop zigzag period serves as a worked illustration, and timing measurements for zigzags through six loops gauge its present reach. HyperFORM is released openly and applies to a broad class of problems, the evaluation of Feynman integrals prominently among them.

We present a strategy to compute the inclusive decay rate for the process $\bar B_s \mapsto X_{\bar sc} \ell \barν$ from first principles in lattice QCD. The physical decay rate is obtained from the interpolation of non-perturbative lattice data, obtained at lighter than physical heavy meson masses ($M_{\bar B_s}^\mathrm{max}=4.3$GeV), with the Operator Product Expansion predictions, which become exact in the limit of infinitely heavy quarks. We also present a new method for the computation of the required lattice four-point correlators, which represents a considerable improvement over the state-of-the-art on the subject. We show the effectiveness of the strategy by performing the calculation on a subset of the available $n_f=2+1+1$ physical-point Extended Twisted Mass Collaboration (ETMC) gauge ensembles. Our current determination of the inclusive decay rate has a 7% total error, that is dominated by uncertainties due to the relatively limited configuration ensembles considered herein, and can be significantly reduced in the near future.

We construct new multiparticle models of $\mathcal{N}=4$ supersymmetric mechanics with spin degrees of freedom by employing nonlinear indecomposable supermultiplets ${\bf (1,4,3){\supset\hspace{-1.1em}+}(4,4,0)}$. These systems are proper deformations of those associated with the standard irreducible $d=1, \mathcal{N}=4$ multiplets. In this way we find a new $\mathcal{N}=4$ supersymmetric generalization of U$(2)$-spin rational Calogero system invariant under $d=1$ superconformal group OSp$(4|2)$. One more deformed model reproduces $\mathcal{N}=4$ supersymmetric U$(2)$-spin hyperbolic Calogero system, up to a shift of the Hamiltonian by some U$(1)$ generators.

The nature of $φ(2170)$ remains open. We simulate its production in $e^+e^-$ collisions at $\sqrt{s}=4.95$ GeV using PACIAE 4.0, which sequentially generates the final partonic state (FPS) and the final hadronic state (FHS). While previous studies have interpreted $φ(2170)$ as an $ss\bar{s}\bar{s}$ or $u\bar{u}s\bar{s}$ state, the $U(1)$ anomaly coupling allows non-strange quarks to couple to a vector $s\bar{s}$ component via soft-gluon interactions. This motivates us to also explore the $d\bar{d}s\bar{s}$ tetraquark configuration. In addition, we consider $φ(2170)$ as an excited strangeonium state, an $s\bar{s}g$ hybrid state, a $\barΛΛ$ bound state, and a $φK^+K^-$ resonance state. The strangeonium, hybrid, and tetraquark candidates are formed by coalescing their constituent partons in the FPS using the dynamically constrained phase-space coalescence model. The $\barΛΛ$ and $φK^+K^-$ states are produced via recombination of their constituent hadrons in the FHS. We calculate the orbital angular momentum quantum number of each candidate in its rest frame and perform spectral classification. Given $J^{PC}=1^{--}$, $φ(2170)$ can be interpreted as a $D$-wave $s\bar{s}$, a $P$-wave $s\bar{s}g$, a $P$-wave $u\bar{u}s\bar{s}/d\bar{d}s\bar{s}/ss\bar{s}\bar{s}$, an $S$-wave $\barΛΛ$, or an $S$-wave $φK^+K^-$ state. The yields of the $D$-wave $s\bar{s}$, $P$-wave $s\bar{s}g$, $u\bar{u}s\bar{s}$ and $d\bar{d}s\bar{s}$ states are of order $10^{-4}$; those for the $S$-wave $\barΛΛ$ and $φK^+K^-$ states are of order $10^{-5}$; while the $P$-wave $ss\bar{s}\bar{s}$ yield is of order $10^{-6}$. Moreover, significant discrepancies are observed in the rapidity distributions and the $p_T$ spectra among the various candidates. These discrepancies could serve as valuable criteria for unraveling the nature of $φ(2170)$.

The Quantum Statistical Parton Model has been successful over the years explain a great set of unpolarized and polarized experimental data. to With the advent of the Marathon and SeaQuest experiments an upgraded version is required to maintain the validity of the model. Moreover, in order to clarify the role of the thermodynamical potentials, the main parameters of the model, we examine the variation of the proton and the neutron entropy with the potentials.

Diphoton systems, with photon polarizations measurable at low energies, serve as ideal qubits and were first used to demonstrate quantum entanglement. However, due to the current absence of dedicated polarimeters at high-energy colliders, the entanglement properties of diphoton systems remain largely unexplored at the high-energy frontier. Studying quantum entanglement at the high-energy frontier, where particle colliders serve as a natural relativistic laboratory, helps us better understand the quantum nature and seek new physics. In this letter, we propose a novel method to measure the entanglement of diphoton systems at proton-proton colliders. The photon spin analyzing power arises from the Bethe-Heitler process occurring in the tracker, where photons scatter off nuclei to produce electron-positron pairs whose joint angular distribution encodes the polarization of the diphoton system. Simulation results show that, under HL-LHC conditions, the statistical significance of quantum entanglement in the Higgs $\to γγ$ process is $0.007σ$, while measuring the continuum diphoton process $q\bar{q} \to γγ$ alone can reach about $1.5σ$.

With increasing experimental precision, the prospect of extracting the polarisation of electroweak gauge bosons is becoming particularly attractive. To this end, regression and classification procedures based on precise and accurate theoretical predictions are becoming increasingly important. In this work, we present the first amplitude-assisted regression procedure at next-to-leading-order accuracy in QCD, supplemented with parton-shower effects, using machine-learning techniques to extract the rate of longitudinal-boson production in high-energy collisions. Several neural-network architectures are presented and benchmarked against a standard random-forest regressor, demonstrating the robustness of the results for di-boson production at the LHC.

Topological string free energy has a rich collection of non-perturbative contributions which are labeled by D-brane charge vectors, and the associated Stokes constants are conjectured to coincide with BPS or DT invariants, i.e. D-brane multiplicities. In this paper, we provide additional evidence to this conjecture by studying modular properties of non-perturbative contributions. We argue using resurgence theory that non-perturbative contributions form orbits of local monodromy group induced by singular points inside a stability chamber, and that the associated Stokes constants must be the same across the orbits. In some examples, this allows generation of infinitely many Stokes constants, which reproduce the entire BPS spectrum. In addition, following [DK26], we also show that generators of Stokes transformations of non-holomorphic partition function satisfy Lie brackets of the Kontsevich-Soibelman Lie algebra, making it possible to identify the global Stokes transformation with the Kontsevich-Soibelman wall-crossing invariant.

Soft and collinear limits of "celestial amplitudes" give rise to infinite dimensional symmetry algebras for two-dimensional (2D) "celestial" conformal field theories (CFTs). A small subset of these operators generates the action of the entire set recursively via the commutation relations. Insertion of this subset into celestial CFT correlators gives the boundary version of the first three terms in a soft expansion of the corresponding 4D bulk gravitational scattering amplitude. In this paper, we find that a bulk analog of this, which would allow the entire soft expansion of a gravitational amplitude to be determined via just the first three terms, does not follow trivially from the celestial algebras. We show how the interplay of soft and collinear limits results in subtleties for bulk amplitudes that do not show up in the boundary description.

Stars on tight orbits around the supermassive black hole at the Galactic Center pass through regions where the dark matter~(DM) density may be strongly enhanced. We compute the orbit-averaged DM-induced energy exchange for S4714 as an example. It is a star on an exceptionally close and relativistic orbit around Sagittarius~A*. For a spiked dark matter profile, the exchange reaches the stellar luminosity at $σ_{χp} \sim 10^{-36}~\mathrm{cm}^2$ for MeV-GeV masses and $σ_{χe} \sim 5\times10^{-38}~\mathrm{cm}^2$ for sub-MeV masses, opening a new annihilation-free route toward dark-star phases. These cross sections lie within the range predicted by freeze-in scenarios and are consistent with cosmic-ray--boosted and solar-reflection dark matter constraints.

We present recursive constructions of several families of generalized Cartan matrices associated with Kac-Moody algebras, whose sizes and coranks grow exponentially. The constructions are encoded by connected multigraphs and by block-doubling operations on their associated symmetric generalized Cartan matrices. Equivalently, the corank problem is translated into a spectral graph-theoretic problem: the corank of $2\mathrm{Id}-\operatorname{Adj}(G)$ is the multiplicity of the adjacency eigenvalue $2$. We give two explicit recursive families, compute their spectra and coranks, and emphasize the difference between absolute exponential growth and relative asymptotic density. The resulting algebras are typically indefinite and singular of corank larger than one, and therefore contain several independent central directions and several isotropic radical directions in the root lattice. We also discuss alternative constructions and possible applications to the algebraic structures appearing in gravity, supergravity, string/M-theory and related generalized symmetry problems.

We show that in the DGKT scenario on a generic Calabi-Yau three-fold, a recently proposed holographic constraint on cubic couplings is satisfied if and only if the Calabi-Yau is rigid, i.e. when $h^{2,1}=0$. More generally, we illustrate how in 4d $\mathcal{N}=1$ supergravity, extremal cubic couplings are determined by the third derivatives of the real, Kähler-invariant superpotential, while the eigenvalues of its Hessian compute the conformal dimensions of the dual scalar operators. These results extend more broadly beyond 4d $\mathcal{N}=1$ supergravity. Applying them to supersymmetric DGKT vacua, we prove that extremal cubic couplings always vanish in the Kähler + universal CS/dilaton sector, whereas non-vanishing (super-)extremal couplings are always present in the complex structure sector. It follows that the holographic constraint is satisfied in DGKT if and only if the Calabi-Yau three-fold is rigid with $h^{2,1}=0$.

The neutron momentum distribution is essential both for reliably extracting fundamental free neutron observables from nuclear measurements and for probing the tensor force via the high-momentum neutron fraction, which is crucial to the theoretical understanding of short-range correlations (SRCs). In this work, we investigate this distribution by studying the $γn \to π^- p$ process at an electron-positron collider, proposing to utilize the beryllium beam pipe at the Beijing Spectrometer III (BESIII). The cross sections for this process on both deuteron and beryllium targets are calculated within the impulse approximation framework. We also evaluate the effective luminosity of the photon flux from radiative Bhabha scattering, taking into account the distribution of target materials within the BESIII experimental setup. Our results show that tens of thousands of events can be generated at BESIII, offering the potential for precise measurements of the neutron momentum distribution. These findings suggest that electron-positron colliders could play a valuable role in elucidating nuclear structure and advancing our understanding of nonperturbative QCD, offering promising new avenues for both particle and nuclear physics.

The extracted $πN$ partial decay width of the $Δ(1232)$ systematically exceeds its total width ($2|r|>Γ$). We demonstrate this anomaly is a natural consequence of S-matrix unitary mixing. Because exact multi-channel shadow poles are distant and model-dependent, we utilize a heuristic elastic model -- treating the overlapping $Δ(1600)$ as fully elastic -- to isolate the core mechanism. We show that evaluating a perturbing S-matrix at a state's complex pole systematically inflates the residue magnitude. This proof of principle confirms complex residues reflect global amplitude topology rather than isolated intrinsic properties, challenging naive interpretations of branching fractions.

Existence of Double-Logarithmic (DL) contributions to scattering amplitudes in QED was discovered by V.V. Sudakov in 1956 and total summation of DL contributions to electron-photon scattering resulted in appearance of famous Sudakov exponentials. Then, thanks to contributions of V.G. Gorshkov, V.N. Gribov, G.V. Frolov and L.N. Lipatov, the pattern of calculations in Double- Logarithmic Approximation (DLA) was constructed. Since then, DLA has become one of basic ways of describing various high-energy processes in the framework of QED, QCD and theory of EW interactions. In the present paper, we remind the history of DLA and present a brief overview of application of DLA to various objects like form factors, scattering amplitudes, DIS structure functions.

In this work, we present KineticXGPU, a PyTorch-based implementation of the $2\to 2$ elastic self-collision operator for dark-sector momentum distributions. The discretized collision operator can be expressed as tensor contractions and is therefore well suited for GPUs. As an application, we study a two-source freeze-in scenario in which the final distribution can develop a bimodal shape. We show that increasing the strength of elastic self-interactions progressively erases this structure and drives the distribution toward a Maxwell-Boltzmann distribution. We compare the phase-space formulation with a set of fluid equations that couple the number density and velocity dispersion. We also compare CPU and GPU runtimes and demonstrate the computational advantage of the tensorized approach. The code is publicly available on GitHub.

The vertex detector upgrade project for the Belle II experiment, based on CMOS depleted monolithic active pixel sensor technology, is planned to be carried out in conjunction with the major modification of the interaction region of the SuperKEKB collider during Long Shutdown 2 from 2032 to 2034. The MAPS sensor, named OBELIX currently under development, is derived from the successor to TJ-Monopix2, with modifications implemented to ensure compatibility with the Belle II trigger system. The new vertex detector consists of two layers of four self-supported consecutive OBELIX sensors, and three layers of discrete OBELIX sensors mounted on mechanical support structures with readout flex circuits attached to the sensors. The detector is arranged cylindrically around the beam pipe at radii ranging from 14 mm to 140 mm. The minimization of the material budget is required in order to enhance physics performance. We present an overview of the project and its latest developments, with particular emphasis on the development of low-material-budget flex circuits employing aluminum conductors.

We investigate the charmonium-nucleon interaction, focusing on its connection with the internal structure of the nucleon encoded in the gravitational form factors. To describe this interaction, we employ an effective potential based on the QCD multipole expansion within the leading chromoelectric dipole approximation. In this framework, the potential is expressed in terms of the energy and pressure distributions inside the nucleon. We first construct these distributions from the gravitational form factors fitted to lattice-QCD data. The remaining model parameters are then fixed by requiring that the resulting $J/ψ$-$N$ potential reproduce the HAL QCD potential outside the short-distance region, as well as the scattering phase shift estimated from the HAL QCD data. Based on this potential, we evaluate the $J/ψ$-$N$ correlation function and further investigate the $ψ(2S)$-$N$ system. We then examine the sensitivity of these correlation functions to the nucleon $D$-form factor.

The notion of depth two and higher mock modular forms have found important applications in mathematical physics and enumerative geometry since their inception through indefinite theta functions with general signature. These theta functions generalize Zwegers' work on Lorentzian signature lattices and the framework of mock modular forms that emanated from it. Mock modular forms can also be studied through Eisenstein and Poincaré series. The interaction of this second point of view with the indefinite theta function approach yields a wealth of tools to unearth the rich structure behind mock modular forms. For mock modular forms of higher depth, on the other hand, indefinite theta functions and their variants largely remained the only available approach. In this paper, we show that one can indeed get mock modular forms of depth two by "coupling" a pair of Eisenstein series that yield depth one mock modular forms, thereby providing a new and independent approach to higher depth mock modular forms. We exemplify this new perspective on a depth two object that appeared in the context of Vafa-Witten invariants.

The spectrum of coherent gluon radiation from a quark-antiquark pair undergoing multiple scatterings within a colored medium is central for understanding in-medium parton cascades. However, current efforts are constrained by reliance on a number of approximations, such as the harmonic oscillator approximation, that are only valid within limited regions of phase space. In this paper, we circumvent this problem by expressing the full in-medium gluon emission spectrum as a set of differential equations that can be solved numerically. This formalism, previously applied to the case of medium-induced radiation off a single color charge, allows to resum medium interactions to all orders while employing realistic scattering models. The resulting angle and energy distributions of emitted gluons serve to illustrate the breakdown of color coherence across the entire accessible phase-space, and constitute a definite step towards a higher-precision description of jet observables.

We determine the mixing angle between the \(1^1P_1\) and \(1^3P_1\) axial-vector states in the \(B_c(1P)\) sector using QCD sum rules. The analysis gives \(θ_{B_c(1P)}=(43.3\pm0.2)^\circ\), indicating sizable mixing between these two states. We also compare our result with theoretical studies available in the literature.

Experimental measurements of $J/ψ$ and $ψ(2S)$ decays to octet baryon pairs reveal a negative-$Σ\barΣ$ puzzle in the angular distributions: $α_Σ^{J/ψ}<0$, whereas the other measured $J/ψ$ octet channels and the corresponding $ψ(2S)$ channels have positive angular parameters. We show that an $SU(3)_F$-exact singlet description fails decisively in a $χ^2$ fit, giving $χ^2\simeq10^4$. By contrast, a fit that allows a common final-state interaction to mix the ${\rm S}$- and ${\rm D}$-wave amplitudes, together with a symmetric $SU(3)_F$-breaking octet, gives $χ^2=4.43$ for two nominal degrees of freedom and strongly favors this breaking pattern. The negative $Σ\barΣ$ angular parameter is traced to the suppression of the magnetic form factor by the common final-state rescattering.

We investigate nuclear transparency in exclusive $ρ^{0}$ electroproduction on $^{12}$C and $^{56}$Fe nuclei within a multi-channel final-state interaction (FSI) framework that explicitly incorporates the kinematic decay length effect (DLE) arising from the short-lived $ρ^{0}\rightarrowπ^{+}π^{-}$ decay. A realistic treatment of the deuteron reference state using the Paris potential wave function, which incorporates the short-range repulsive core and tensor correlations, provides a physically reliable normalization for the transparency ratio $T_A/T_D$. The conventional DLE and nuclear shadowing mechanisms together remain insufficient to account for the observed $Q^2$-dependent enhancement, systematically underestimating the measured transparency throughout the CLAS kinematic range. To address this, we introduce an effective Color Dipole Model (CDM) boundary condition for the initial PLC interaction cross section $σ_{\text{h}}(Q^2)$, evaluated as a dipole-weighted average over the $γ^*$--$ρ^0$ transition wave functions, in place of the purely empirical Quantum Diffusion Model (QDM) ansatz. This CDM-inspired initial condition, combined with the standard linear QDM transport, yields a consistent description of the $Q^2$-dependent CLAS data for both targets with an effective in-medium expansion scale $Δm^2 = 0.3~\mathrm{GeV}^2$. Although the present analysis does not provide definitive evidence for the onset of Color Transparency, it demonstrates that a CDM-inspired PLC boundary condition, together with a realistic treatment of the underlying reaction dynamics, yields a physically consistent and quantitatively improved description of the CLAS data.

The knowledge of isospin-breaking effects in hadronic $τ$ decays is required for a high-precision determination of the Hadronic-Vacuum-Polarization contribution to $(g-2)_μ$ from experimental $τ$ data. In this work we present a strategy for their calculation in a fully inclusive setup from first-principles Lattice QCD+QED simulations. We separate radiative corrections in three infrared safe classes, which we study individually. We provide analytic expressions for their effects in the initial state and propose a strategy for final-state corrections directly in Euclidean space. We also examine the non-factorizable contributions and highlight the challenges associated with their analytic continuation from Euclidean to Minkowski space. By studying short-distance corrections in the context of momentum schemes, we provide a prescription for the renormalization of the individual terms at first order in the ispospin-breaking parameters.

We propose a modular symmetric inflationary model based on a Coleman--Weinberg potential generated by integrating out heavy vector-like quarks that couple to the complex modulus field $τ$ through modular forms. In this framework, the imaginary part of modulus $τ$ plays the role of the inflaton, while the real part is identified with a heavy axion. We show that the model successfully explains the current cosmological observations. We further discuss reheating through modulus-dependent gauge kinetic functions and the cosmology of the axion. The axion oscillation dominates over the Universe after the reheating via inflaton decay, and then it decays before Big Bang Nucleosynthesis in the viable parameter region. The quantum fluctuation of the axion can be of order $\mathcal{O}(1)\% $ of that of the inflaton, which would induce isocurvature perturbations that may be detectable in future observations.

We investigate the multidimensional partonic structure of spin-0 mesons, specifically the pion and the kaon, by evaluating their complete set of eight generalized parton distributions (GPDs) up to twist-4. Utilizing the light-front quark model (LFQM) with the Brodsky-Huang-Lepage (BHL) prescription, we compute these distributions in the kinematically rich non-zero skewness ($ξ\neq 0$) domain, strictly within the DGLAP region, $x \in [ξ, 1]$. To construct a three-dimensional tomographic picture, we perform Fourier transforms of the momentum-space GPDs to obtain the impact parameter dependent parton distribution functions (IPDPDFs) in the transverse plane and the corresponding diffraction patterns in the longitudinal coordinate space. The numerical results explicitly reveal the consequences of $\mathrm{SU}(3)$ flavor symmetry breaking, as the strange quark in the kaon dynamically shifts spatial localizations compared to the lighter up quarks. We also observe that while higher-twist correlations exhibit massive amplitude scaling in the pion, they are heavily suppressed by the larger macroscopic mass of the kaon.

We introduce gradient flow (GF) normalization prescriptions for fermionic composite operators in which the flowed fermion wavefunction renormalization factor is fixed nonperturbatively using either the partially conserved axial charge or the conserved vector current. The resulting $A$ and $V$ schemes are defined through standard flowed two-point correlation functions and therefore avoid the backward-flow construction required by local ringed-scheme definitions. In the short-flow-time limit, the $A$ and $V$ schemes can be matched to $\overline{\mathrm{MS}}$ using known ringed-scheme short-flow-time expansion (SFTX) coefficients. We show how these schemes can be implemented through ratios of two-point correlation functions, leading to simple nonperturbative determinations of renormalization factors, anomalous dimensions, and evolution factors which connect lattice-accessible flow times to shorter flow times where perturbative matching is reliable. We illustrate the method with RBC-UKQCD domain-wall fermion ensembles, including a GF determination of the ratio of matching factors $Z_V/Z_A$, and a new GF determination of the renormalized strange quark mass.

Precision measurements of neutrino--electron elastic scattering provide low-energy tests of weak interactions and beyond-the-Standard-Model effects. Non-standard interactions (NSIs) and an anomalous neutrino magnetic moment modify the differential cross section through different kinematic terms, but both can alter the normalization and shape of the recoil-electron spectrum. Likelihood tests are computationally costly when each parameter point requires the recoil spectrum to be propagated through a detector response obtained from Monte Carlo (MC) simulation. We present a GPU-accelerated spectrum-reweighting framework that avoids regenerating detector MC samples for each new-physics parameter point. Bin-to-bin weights are applied at the recoil-spectrum level and folded with a fixed two-dimensional response model in recoil and reconstructed energy. This keeps the detector response inside the likelihood calculation while reducing each parameter update to operations on precomputed spectra and response kernels. The implementation uses NVIDIA Thrust transformation--reduction primitives and is compiled from a common source for CUDA and OpenMP back ends. In the benchmarks considered here, one likelihood evaluation takes ${\sim}87$ ms on an NVIDIA RTX 3080Ti and ${\sim}52$ ms on an NVIDIA A30X; the latter gives a $58\times$ speedup over a single CPU thread and ${\sim}2.5\times$ over a fully loaded 64-thread CPU. The consumer-GPU result demonstrates that interactive parameter scans are feasible on a single workstation. The main acceleration, however, comes from avoiding detector-MC regeneration at each parameter point rather than from GPU execution alone. The framework applies to neutrino--electron scattering analyses in which the new-physics dependence can be represented by reweighting an existing recoil spectrum, including flavor NSI and magnetic-moment cases studied here.

The leading-eikonal small-$x$ dipole gives a compact representation of the gluon momentum form factor $A_g(t)$. We show that the same information is not sufficient to determine the gluon stress form factor $C_g(t)$, and hence the gluon D-term. The reason is kinematic and operatorial: in a Drell-Yan frame $A_g(t)$ is projected by $T_g^{++}$, whereas $C_g(t)$ is projected by the symmetric-traceless transverse stress $T_g^{ij}$. This stress projection first appears through next-to-eikonal fields and is represented by gauge-invariant stress-decorated Wilson lines containing $F^{i-}$, $F^{ij}$, or equivalent sub-eikonal target fields. We construct this operator and match it to the local energy-momentum tensor at tree level, obtaining an operator-level no-go statement: the ordinary dipole $S_x(b_\perp,r_\perp)$, or a saturation profile $Q_s^2(x,b)$, does not by itself determine the sign of the small-$x$ gluon D-term. We then give a finite-correlation response model in which a positive kernel generates an anti-aligned response $F^{i-}=-εR_{\rm NE}F^{i+}$, so that $Λ_{\rm NE}>0$ and $D_g(0)<0$ within that response class. A Gaussian benchmark and numerical scans over Gaussian, Woods-Saxon, power-edge, and McLerran-Venugopalan (MV)-inspired profiles show a stable negative forward D-term for this anti-aligned model, together with the expected core-shell pressure pattern. The gluon D-term is therefore a next-to-eikonal stress probe, not a universal leading-eikonal saturation observable.

We extend the exceptional-Jordan spectral framework for fermion mass hierarchies to the problem of quark and lepton mixing. Following the companion mass paper~\cite{Teli:2026jgr}, each fermion sector is associated with a Hermitian element of $J_3(\mathbb{O}_{\mathbb{C}})$, where adjacent square-root mass ratios are obtained from cubic ladders in $\mathrm{Sym}^3(\mathbf 3)$. Here, these ratios are used as inputs to an adjacent-edge lift from spectral hierarchy data to two-generation mixing angles. The lift is derived from a Fritzsch-type two-state texture~\cite{Fritzsch:1977za, Fritzsch:1979zq} and should be regarded as an effective bridge ansatz rather than a theorem of the Jordan spectrum alone. The exact CP-transport input is supplied by the companion CP Letter~\cite{GuptaTeli:2026aqf}. In the quark sector, the octonionic ladder operator $α_2$ generates a real local rotor in the $(e_1,e_3)$ plane, and the up- and down-sector local Cabibbo-edge amplitudes are complex conjugates, giving the exact local law $φ_{12}=-2χ$. This is a transport-level Cabibbo-rung phase law, not by itself a prediction of the standard CKM Dirac phase. With the fitted companion mass ratios, the minimal two-angle extraction from the measured $|V_{us}|$ gives an effective Cabibbo-block phase $φ_{12}\simeq 105.7^\circ$; this number is a bridge diagnostic, while the balanced octonionic rotor remains the distinguished quadrature reference point. The $(2,3)$ sector requires a phenomenological normalization $κ_{23}\simeq0.56$, and the direct $(1,3)$ element remains a long-edge bridge problem. [Truncated]

In this work, we propose a scheme to classify the molecular states consisting of ground single-charm baryons ($Λ_c$, $Ξ_c$, $Σ_c^{(*)}$, $Ξ_c^{\prime(*)}$, $Ω_c^{(*)}$) and $\bar{D}^{(*)}/\bar{D}_s^{(*)}$ mesons. Within this framework, all considered baryon-meson systems are categorized according to the flavor components of their light degrees of freedom. We briefly illustrate how this classification scheme can consistently explain the experimentally observed $P_c$ and $P_{cs}$ states. This framework also predicts the existences of single-strange and double-strange hidden-charm bound states. The attractive interactions of these states arise from channel mixing between $Σ_c^{(*)}\bar{D}_s^{(*)}$ and $Ξ_c^{\prime(*)}\bar{D}^{(*)}$ for single-strange systems, and mixing between $Ξ_c^{\prime(*)}\bar{D}_s^{(*)}$ and $Ω_c^{(*)}\bar{D}^{(*)}$ for double-strange systems, respectively. Using parameters fitted from the measured $P_c$ and $P_{cs}$ states, we systematically present the predicted mass spectra for these single- and double-strange hidden-charm bound states.

The SNAPPY CubeSat, which was launched May 3, 2026, will demonstrate and space qualify the nuSol neutrino-detection technology. The nuSol technology detects solar neutrinos using a gallium isotope which decays by emitting two particles spaced apart in time; this allows differentiating neutrino events from cosmic rays. In the NIAC Phase II project review in 2021, concept and science were determined to be feasible; however, two precursor studies were recommended before pursuing a full mission study. These studies were to characterize the true deep-space background for the detector's gallium double-pulse signal and to collect a statistically significant number of double-pulse events demonstrating that fast electronics can reliably select and analyze this signal. To test double-pulse signals in space, a NIAC Phase III funded building a 3U CubeSat carrying a 0.1-kg gallium-aluminum-gadolinium-garnet detector housed within an active veto array and shielding. Because the detector requires deep-space-like conditions, the CubeSat is designed for a polar low-Earth orbit at 450 km or higher altitude, collecting data over the Earth's poles above the Van Allen belts. The detector is highly sensitive, with roughly 7-percent energy resolution, with active veto shielding and passive shielding using a patented tungsten-powder and epoxy mixture that disintegrates upon atmospheric reentry. SNAPPY enables additional science during the extended mission phase of year two operations. These include measurements of solar wind particle density and energy spectra with particle identification of electrons, protons, and alpha particles; detection of very low-energy gamma rays from galactic gamma-ray bursts without directionality.

We explore the potential of the Large Hadron Collider to probe a two-component scalar dark matter scenario in the opposite-sign dimuon plus missing transverse energy final state, accompanied by a hard jet. The signal features a soft dimuon system with an invariant mass well below $m_Z$. We consider a 3-Higgs Doublet Model with one active and two inert scalar doublets, where a $Z_2 \times Z_2'$ symmetry stabilises the lightest neutral scalar in each inert sector, yielding two scalar DM candidates. The relevant parameter space is mapped in terms of the two DM masses and the mass splittings between each DM candidate and its corresponding next-to-lightest scalar state. We perform a detector-level Monte Carlo analysis and design a dedicated cut-based selection, including a transverse-mass requirement adapted to the signal topology. For a representative benchmark, we obtain $S/B\simeq 9.8%$ and a statistical-only significance of $S/\sqrt{B}=1.35$ at Run 3 with ${\cal L}=300~{\rm fb}^{-1}$, increasing to $S/\sqrt{B}=4.93$ under a statistical-only extrapolation to ${\cal L}=4~{\rm ab}^{-1}$. Before the full selection, the two dark sectors generate a double-bump structure in the dimuon invariant-mass distribution. After the cuts optimised for inclusive sensitivity, however, this feature is not statistically robust enough to establish the two-component origin of the signal. The benchmark is underabundant and is interpreted as a subdominant two-component DM scenario, while the collider analysis remains independent of its cosmological abundance. Although the numerical study is carried out in the I(2+1)HDM, the results are applicable to weakly interacting sectors with similar electroweak associated production and cascade decays, where a heavier state separated from the DM candidate by less than $m_Z$ produces a soft muon pair via an off-shell $Z$ boson.

Many proposed extensions to the Standard Model of particle physics introduce new bosons that can mediate forces which couple to particle spin. Here we describe a search for such forces coupling spin-polarized neutrons and protons in our magnetometer to spin-polarized electrons within Earth. We measure these interactions by varying the orientation of an optical $^{199}$Hg-$^{133}$Cs free-precession comagnetometer mounted upon a precision rotation platform. From these measurements, we establish upper bounds on the dimensionless coupling constants associated with the axial-axial potential $V_2$ and the axial-vector potential $V_{11}$ as a function of the force's range $λ$. For the electron-neutron and electron-proton potential $V_2$ at infinite range, we find $|g_A^eg_A^n| \leq 3.0 \times 10^{-48}$ and $|g_A^eg_A^p| \leq 3.0 \times 10^{-47}$. For $V_{11}$, we find our most stringent bounds to be $|g_A^eg_V^n| \leq 2.2 \times 10^{-25}$ and $|g_A^eg_V^p| \leq 2.2 \times 10^{-24}$ at $λ\approx 10^3$ km. Our results represent an improvement over previous results by up to a factor of 17 and set the most stringent bounds on long-range axial-axial and axial-vector couplings between electron spins and neutron and proton spins.

Precision calculations in quantum field theory rely very often on perturbation theory and thus on the computation of Feynman integrals. Feynman integrals are also fascinating objects from a mathematical point of view and show deep connections to algebraic geometry. Cutting-edge Feynman integrals usually have geometries of "mixed" type, for example parts of it may correspond to a K3-surface, other parts may correspond to curves of a certain genus and the simplest parts correspond to points. In this talk I will discuss how to extract the geometric information from a Feynman integral and how this information can be used to compute more efficiently Feynman integrals. Non-trivial mixed geometries already occur in $2 \rightarrow 2$-processes at two-loops, like Drell-Yan, Bhabha and Moller scattering.

The search for new particles, conducted at the LHC by the ATLAS and CMS collaborations, covers several production and decay modes, leading to a large variety of final states that could be observed in both detectors. This review focuses on the production of new heavy scalars that have a Higgs boson in the final state. Three cases are covered: resonant production of a heavy scalar $X$ decaying into a Higgs boson and a lighter scalar $S$, heavy neutral Higgs boson decaying into another neutral Higgs boson and a $Z$ boson, and a heavy charged Higgs boson production with subsequent decay into another neutral Higgs boson and a $W$ boson. The reviewed searches are based on the complete LHC Run-2 dataset and partial Run-3 dataset.

We derive a novel representation of the partial wave amplitude over the left-hand cut for $2 \to 2$ scattering. We express the left-hand cut of arbitrary isospin and angular momentum partial waves as an integral of right-hand cut imaginary parts. This formulation provides an explicit, exact extraction of the logarithmic branch cut structures, offering a valuable tool to systematically quantify left-hand cut uncertainties in unitarization methods such as the Inverse Amplitude Method or $N/D$ approaches.

One of the main advances in analysis strategies at the Large Hadron Collider (LHC) has been the ability to study the detailed structure of energy flow within high transverse momentum jets, a field referred to as jet substructure. Jet substructure has provided new ways to search for new physics, measure Standard Model parameters, and study the dynamics of the strong nuclear force. To push to the next level of precision, and to make measurements of increasingly subtle correlations, requires exquisite angular resolution achieved through the use of tracking information. In this paper we leverage recent progress in our understanding of factorization theorems and renormalization group techniques to present the first complete calculations of jet substructure observables at the LHC on tracks. We compute projected energy correlators up to four points at next-to-leading collinear logarithmic accuracy, matching the state of the art for jet substructure observables, but extending to tracks. This marks a significant step in enhancing the collider physics program, enabling precise and systematically improvable comparisons between experimental measurements and theoretical calculations, made possible by the exceptional angular resolution of tracking.

Correlation functions of local operators in Quantum Field Theory (QFT) in Anti-de Sitter space (AdS) are completely fixed by the QFT data: the set of scaling dimensions $Δ_i$ and OPE coefficients $C_{ijk}$ of the boundary operators, and the bulk-boundary (BOE) coefficients $b^{\hatΦ}_i$ encoding how bulk fields decompose into boundary operators. In this work, we generalize the ordinary differential equations (ODEs) that govern the variation of the QFT data under a bulk relevant deformation, originally derived for AdS$_2$ \cite{Loparco:2026fki}, to the cases of AdS$_3$ and AdS$_4$. We demonstrate that these flow equations natively capture the mechanism of merger-annihilation when a boundary operator hits marginality, as well as level repulsion when different $Δ_i$'s approach each other. Furthermore, we address the practical implementation of the framework: we propose substituting the ODE for the OPE coefficients with the crossing equation for greater efficiency, and we observe that Padé approximants dramatically improve the convergence of the sums over boundary operators, at least in free theories. Altogether, these advances lay the groundwork for the future application of the flow equations to the study of strongly coupled QFTs in AdS and their flat space limits.

We propose a Transformer-based Reinforcement Learning architecture, "LB-Explorer", to search for heterotic line bundle standard models arising from compactifications on smooth Calabi-Yau (CY) threefolds. We construct $E_8\times E_8$ vacua with $\text{SU}(5)$ symmetry, where the $\text{SU}(5)$ can be further broken to the Standard Model gauge group via discrete Wilson lines. We test the LB-Explorer environment on complete intersection Calabi-Yau (CICY) manifolds, though the neural network architecture naturally generalizes to any CY admitting a smooth, simplicial Mori cone and a freely-acting discrete symmetry. The LB-Explorer efficiently learns constraints on the line bundle sums, guaranteeing the $E_8$ gauge embedding, anomaly cancellation, poly-stability (supersymmetry), chirality of the spectrum, and the absence of exotic matter. Valid configurations can be subsequently filtered by imposing the missing constraints, such as the equivariant structure of the line bundle sum and further requirements on the particle spectrum. In this direction, we introduce a hybrid architecture incorporating CP-SAT solvers that aims to impose some of the conditions exactly by perturbing solutions found by the LB-Explorer. The versatility and scalability of the LB-Explorer make it a powerful tool for navigating the string landscape with a large number of moduli. The code and tools necessary to reproduce our findings are available at https://github.com/alexmininno/LB-Explorer

Ultralight dark matter exhibits an order-one density fluctuation at the scale of its wavelength. This density fluctuation exists across the entire dark matter halo and interacts with stars and planets, perturbing their motion via gravitational interactions. We investigate the possibility of using precision solar system dynamics to search for ultralight dark matter. We examine this possibility with interplanetary radio range measurements. We show that the precision of current range measurements can probe ultralight dark matter at masses around $10^{-15}\,$eV, had its density in the solar system been $10^5$ larger than the so-called local dark matter density. This limit complements other constraints, such as the one from analyses of pulsar timing observations.

We study the exact partition function of 4d $\mathcal N=1$ supersymmetric gauge theories on a torus times a cylinder $\mathrm{Cyl}=I\times S^1$, where $I$ is a finite interval carrying two boundary components. Each endpoint supports an independent Dirichlet or Robin-like boundary polarization, so that the partition function is a boundary-to-boundary elliptic kernel. We construct the rigid supersymmetric geometry, determine the BPS locus, and compute the chiral-multiplet 1-loop determinants for the four possible boundary polarizations via equivariant localization. The resulting elementary building blocks are theta functions dressed by cubic phases. We then prove rank-changing Seiberg-type dualities as identities of Jeffrey--Kirwan residues of these elliptic kernels. We also discuss factorization into holomorphic-block cap wavefunctions represented by elliptic Gamma functions, dimensional reductions to three and two dimensions, complete-intersection gauged linear sigma models, and elliptic kernels for 4d $\mathcal N=4$ super Yang--Mills and the Klebanov--Witten theory, useful for holographic applications.

We study Krylov spread complexity in holographic theories using genuine string-theory probes. Building on the proposal that the growth rate of spread complexity is measured by a proper momentum in the bulk, we embed the falling-particle picture in top-down examples. We first analyse a D0 brane in the type IIA AdS$_4\times {\mathbb{CP}}^3$ background dual to ABJM theory, identifying it with a dressed monopole operator in the boundary CFT. For purely radial motion the proper-momentum prescription reproduces the expected quadratic growth of the complexity. When the probe carries momentum along an isometric direction, the naive prescription gives an apparent conflict with the short-time behaviour required of Krylov complexity. We propose that the correct fixed-charge description is obtained by Legendre transforming to the Routhian. We support the D0-brane interpretation through the regulated monopole two-point function, whose survival amplitude determines the Krylov moments, and we show that radial fluctuations give controlled corrections to the effective energy governing the complexity growth. We then extend the analysis to a rotating non-BPS D3 brane in AdS$_5\times S^5$, where angular momentum produces a centrifugal barrier and a sharp condition for radial in-fall. In the falling regime the Routhian prescription again gives the correct short-time behaviour. Finally, we consider a wound fundamental string in AdS$_5\times S^5$, which reduces to an effective massive falling particle. This clarifies the distinction between Noether charges, which require a fixed-charge Routhian treatment, and winding data, which enter through the effective mass. Our results provide a string-theoretic realisation of holographic spread complexity for point-like and extended excitations, making manifest their dependence on field theory parameters.

Generic Feynman integrals are widely studied as solutions of Picard-Fuchs equations on moduli spaces of their parameters, and this calls for consideration of this phenomenon at a more basic level - of string amplitudes which are integrals over true non-singular module space of Riemann surfaces and their various generalizations. The main puzzle here is that a single string amplitude involves mane different particle diagrams, corresponding to different parts of the same moduli space, but different particle diagrams are usually believed to satisfy different equations, not unified into a common entity. We begin investigation of this problem, starting from Koba-Nielsen diagrams. While there is nothing interesting at this level for particles, the tree-level open bosonic string amplitudes satisfy non-trivial linear difference equations in kinematic variables. Moreover, the integration-by-parts on moduli space, standing behind Picard-Fuchs equations for particle loops, for strings are operative already at the tree level. We construct a complete system of such equations for arbitrary n-point tree amplitudes, with the number of independent relations matching the kinematic parameters. In variance with the particle case equations are difference ones rather than differential. The low-energy limit $α\to 0$ smoothly recovers the algebraic QFT structure.

Geometric scaling is a central output of nonlinear small-$x$ evolution, but it is less clear whether the same dynamics fixes a probability distribution in transverse phase space. Using fixed-coupling impact-parameter BK evolution in the $SO(3)$-symmetric construction, we build a normalized gluon Husimi phase-space distribution and resolve it with a local coarse graining whose ultraviolet boundary follows $Q_s(Y,b)$. The main result is a distribution-level one: after this $Q_s$-adaptive resolution, the conditional momentum distributions collapse as functions of $k/Q_s(Y,b)$. The conditional entropy then grows with unit slope relative to $\langle\ln Q_s^2\rangle$, as the integrated consequence of that collapse and the two-dimensional momentum measure. Fixed laboratory cutoffs do not show this law, while dense-rapidity, cutoff-window, box-size, regulator-shape, and Husimi-resolution scans keep the $Q_s$-adaptive result stable in the controlled window. Within this fixed-coupling $SO(3)$-BK setting, the result identifies a local phase-space scaling structure of the gluon Husimi distribution rather than a universal law for unregulated global entropy.

We study the chemical equilibration of a hot and dense quark-gluon plasma (QGP) at finite baryon density produced in relativistic heavy-ion collisions within conformal Gubser flow. Chemical non-equilibrium is incorporated through fugacity parameters in the parton phase-space distribution functions, whose evolution is governed by master rate equations coupled to the hydrodynamic expansion with transverse flow. We analyse the interplay between chemical equilibration and finite-density dynamics, and investigate its impact on hard thermal photon production. We observe that both finite density and transverse expansion delay chemical equilibration, leading to a chemically undersaturated medium with quarks lagging behind gluons. While the overall thermal photon yield from the expanding system is suppressed in the non-equilibrium scenario, we find an enhanced early-time contribution to high $p_T$ photon production. By analyzing the instantaneous photon emission in presence of chemical non-equilibrium, we demonstrate that the rates exhibit a distinct temporal structure arising from the interplay of rapid cooling and evolving fugacities. These features may provide potential observable signatures of chemical equilibration dynamics in the QGP.

We study how collinearly improved small-$x$ evolution modifies the elliptic gluon Wigner distribution and its coherent diffractive dijet $\cos2φ$ signal. Starting from the $SO(3)$-symmetric Gubser initial condition, we compare fixed-coupling leading-order (LO) BK evolution with the Iancu--Madrigal--Mueller--Soyez--Triantafyllopoulos (IMST) collinearly improved BK prescription, project the evolved dipole to Wigner/GTMD harmonics, and fold the result with photon wave-function hard factors. Within the $SO(3)$-projected framework, the resummed evolution moves the elliptic node and changes the rapidity and hard-scale dependence of $C_2^{\rm coh}=\langle\cos2φ\rangle$, rather than acting as a simple normalization shift. The effect is largest in node-sensitive windows but remains visible in broader sign-stable bins. Direct daughter-coordinate RHS tests and direct $(r,b,φ)$ evolution quantify the residual projection ambiguity and check the observed trend. Projection-level EIC-like finite-bin estimates show that balanced high-$Δ$ candidate bins offer a better compromise between elliptic signal and denominator weight than the narrow bins with the largest ratio. The unresummed NLO curve is used only as a stability diagnostic; the physical comparison is between LO BK and IMST collinearly improved BK evolution.

The elliptic small-$x$ gluon Wigner distribution correlates transverse momentum with impact parameter, but it is usually accessed through exclusive diffractive dijets whose recoil is sensitive to soft radiation. We propose instead an azimuthal energy-flow moment in DIS dijet production. Within leading-power small-$x$ dijet factorization, its normalized $\cos2φ$ moment is a linear projection of the elliptic Wigner harmonic after a calculable kinematic subtraction, while the final-state energy weighting is infrared and collinear safe. In conjugate recoil space a rotationally scalar Sudakov factor evolves the isotropic and elliptic channels through a fixed $J_0/J_2$ Hankel pair without $W_0\to W_1$ leakage. A proof-of-principle calculation gives a per-mille Sudakov-level moment in an unoptimized conservative window, while auxiliary perturbative-window scans reach several per mille. The observable therefore formulates elliptic Wigner tomography as a moment-level energy-flow measurement whose statistical reach can be assessed by simple angular-moment counting.

We propose that AdS$_3$ gravity coupled to dS$_2$ end-of-the-world branes is dual to a unitary holographic CFT$_2$ with non-unitary conformal boundary conditions. These boundary conditions have complex $g$-functions and appear in conjugate pairs. The associated gravitational path integral admits no real saddles, but does admit complex saddles. We show that an AdS$_3$ black hole microstate with a dS$_2$ brane behind the horizon corresponds to the unitary time evolution of a pure CFT state prepared by a Euclidean path integral on a cylinder with such boundary conditions. The construction predicts a boundary-condition-changing primary with $h=-c/8$, which resides in the boundary sector rather than the bulk spectrum and is therefore compatible with unitarity of the underlying CFT. This realizes dS holography as state preparation in a unitary AdS/CFT Hilbert space.

Classical nonabelian Hodge theory identifies Dolbeault and de Rham moduli spaces by providing a real-analytic isomorphism. In this paper, motivated by the Kapustin--Witten theory, we study this correspondence in the more general framework of perfect complexes on proper varieties, paying special attention to the surface case. We establish a Lagrangian correspondence which relates the shifted symplectic geometries by Pantev--Toën--Vaquié--Vezzosi (PTVV) between the derived stacks of flat and Higgs perfect complexes. We investigate the existence of derived twistor structures of hyperkähler type on the moduli stack of perfect complexes endowed with $λ$-connections by Deligne--Hitchin--Simpson. We establish a version of the AKSZ/PTVV transgression, Lagrangian intersection, and (hyperkähler) symplectic reduction theorems in this context. Moreover, we prove that the derived Riemann--Hilbert correspondence of Porta and Holstein--Porta, which states an equivalence of derived analytic stacks of perfect complexes on $X_{\mathrm{Betti}}$ and $X_{\mathrm{DR}}$, is compatible with the natural shifted--symplectic structures. We then study the relation between the shifted (pre-)twistor structures and the shifted symplectic forms on the fibers, and prove that the analytic Deligne--Hitchin--Simpson moduli stack on a smooth projective variety $X$ has a canonical $2(1-\dim X)$ shifted pretwistor structure over $\mathbb{P}^1_{\mathbb{C}}$, a result which has been anticipated for some time. In particular, the moduli stack of solutions to the Kapustin--Witten equations modulo gauge equivalence on a smooth proper complex algebraic surface exibits a $(-2)$-shifted (pre)twistor structure as a family over $\mathbb{P}^1_{\mathbb{C}}$.

When two neutron stars collide, they eject material containing heavy nuclei formed by the rapid neutron capture process ($r$-process). As these nuclei decay, they power a bright optical/near-infrared transient known as a kilonova (KN). Modeling KN emission is a complex problem involving atomic opacities, radiation transport, and heating powered by the thermalization of radioactive decay products like $γ$-rays, $α$-particles, and $β$-particles. For heating by $γ$-rays, many KN modeling codes do full radiation transport calculations. However, heating by $α$- and $β$-particles relies on simplified descriptions of collisions and transport, and remains an important source of uncertainty in KN models. In this paper, we study the thermalization and transport of $β$-particles. To study thermalization, we use evaluated atomic physics data to estimate per-species contributions to energy deposition, scattering, and electron impact ionization, which we make available online. To include non-local effects, we develop a fully relativistic framework for charged particle transport in a spherically symmetric, homologously expanding ejecta, considering two limiting magnetic-field geometries. Non-local energy deposition and escape reduce thermalization efficiency, especially in the innermost and outermost ejecta, lowering the ejecta temperature and ionization state compared to local deposition models. Coulomb scattering partially offsets these effects by trapping particles at intermediate times. Ionization by secondary electrons significantly enhances the overall ionization rate. We provide analytic prescriptions for the spatially dependent thermalization efficiency for use in future light-curve calculations. Our results demonstrate that evaluated atomic data and charged-particle transport should be incorporated into the next generation of KN models.

We present a systematic reassessment of cosmological constraints beyond $Λ$CDM by progressively relaxing the assumptions underlying Dark Energy (DE), Curvature, Neutrinos, and Inflation. Using the latest CMB data together with DESI BAO and different SN catalogues, we show that the preference for dynamical DE persists across all the extended cosmologies considered. $Ω_k$ remains compatible with flatness, despite a mild $2.2σ$ preference for $Ω_k>0$ that is substantially degraded in dynamical DE extensions. Constraints on $N_{\rm eff}$ are broadly consistent with $N_{\rm eff}=3.04$, while cosmological upper limits on the total neutrino mass vary substantially across the cosmologies explored, ranging from $\sum m_ν\lesssim 0.06$ eV to $\lesssim 0.2$ eV. We quantify both the preference for the mass ordering and the apparent tension between cosmology and oscillation experiments, showing that they are strongly framework dependent. We find no evidence for inflationary tensor modes, with $r\lesssim 0.035$. Constraints on the spectral index $n_s$ show significant model dependence. Allowing for the scalar runnings produces a mild shift toward $α_s>0$ and $β_s>0$ that can reabsorb the preference for larger $n_s$ found in small-scale CMB data, although both $α_s$ and $β_s$ remain consistent with zero at $\sim 1.5σ$. We highlight the implications for slow-roll inflation and benchmark models. None of the extensions considered here can resolve the $H_0$ tension. We discuss the implications for $Ω_m$ and $S_8$. Overall, dynamical DE is the only significant deviation from $Λ$CDM and has the strongest impact on the inferred conclusions in the other sectors of the model.

Gravitational collapse creates large amounts of kinetic energy that could potentially seed turbulence. If such turbulence were also suitable to initiate dynamo action, the resulting magnetic field would further modify the dynamics, especially on small length scales. However, a small-scale dynamo requires vortical turbulence, while the collapse produces mainly irrotational motions, which may not be efficient for dynamo action. Here, we study the efficiency of vorticity production during a turbulent collapse. We use a barotropic equation of state, where pressure and density gradients are parallel, and no magnetic field, so that vorticity can only be produced by viscosity. Using direct numerical simulations of gravitational collapse, we show that, for the parameter space accessible to our numerical resolution, this effect is related to the initial irrotational turbulence and is not a consequence of the collapse flow.

We study a modified cosmological scenario for the late Universe, involving an evolutionary dark energy model associated with the dynamics of a self-interacting scalar field in a potential-dominated regime. Through the analogy with a fluid energy-momentum tensor, we introduce a viscous contribution to the scalar dynamics, accounting for effective non-equilibrium behaviour of the self-interacting scalar cluster. The resulting picture is that of an intrinsic quintessence contribution which, due to the bulk viscosity, admits an effective equation of state parameter that can also take values below -1. Within this framework, we set up the diagnostic tool of the so-called "effective running Hubble constant", which allows us to trace possible deviations from a standard LambdaCDM model. We then compare this theoretical function with binned data from the Master Sample of Supernovae Ia, constructed assuming a LambdaCDM model in the MCMC procedure performed in each bin. We show that the self-interacting scalar field corresponding to the best fit satisfies a slow-rolling condition, since the kinetic energy remains small compared to the potential contribution throughout the redshift interval. The key finding is that, when limiting the model to specific regions of the parameter space and fitting it to the data, the transition only occurs at redshifts significantly lower than the redshift value identified by the DESI Collaboration. Furthermore, for the parameter values ensuring the best fit, no quintessence-to-phantom transition occurs (i.e., the effective equation of state parameter remains below -1 across the whole redshift domain). In other words, Supernovae data alone provide no indication of a change in the nature of the dark energy.

We report a new galaxy cluster catalog, selected using the thermal Sunyaev-Zel'dovich (SZ) effect, from 5 years of observations of the SPT-3G Main field. Drawn from arcminute-resolution data with white noise levels of 3.2, 2.5, and 8.9 $μ$K-arcmin at 95, 150, and 220 GHz, respectively, the sample consists of 8,892 cluster candidates detected above significance $ξ=4$, with an expected purity of $>82\%$ (4,480 at $ξ\ge5$ with purity $>99\%$). Using optical and infrared data we have confirmed 7,190 candidates as clusters. The sample spans a mass range $7.9 \times 10^{13}$ $M_\odot/h_{70}$ \ $< M_\textrm{500c} < $ $1.6 \times 10^{15}$ $M_\odot/h_{70}$ with a median mass of $1.65 \times 10^{14}$ $M_\odot/h_{70}$, and a redshift range of $0.037<z\lesssim 2$ with a median redshift of $z_{\textrm{med}}$ = 0.73; 1,780 clusters are at $z>1$ and 271 at $z>1.5$. Compared to previous SZ cluster samples from South Pole Telescope and Atacama Cosmology Telescope data, the SPT-3G sample is highly consistent in mass and redshift but is significantly deeper, with per-cluster detection signal-to-noise 2-4 times higher and a cluster density of 4.5 confirmed clusters/deg$^2$. We cross match with eRASS1 cluster and point source catalogs, finding 1,279 and 1,319 matches, respectively. The SPT and eROSITA cluster mass estimates are in relatively good agreement. We perform a series of validation checks using both internal data splits and comparisons to external samples. These tests show increasing correlated (dusty) emission with redshift, with a $\sim17\times$ larger 220 GHz temperature increment for clusters at $z\sim1.5$ than $z\sim0.25$, but only weak evidence for correlated synchrotron emission. Finally, a number of clusters are flagged as candidate strong gravitational lenses.

The cosmological principle, which states that the Universe is statistically homogeneous and isotropic on sufficiently large scales, is a foundational assumption of the standard cosmological model. A recent analysis of DESI DR1 galaxy samples reported coherent anisotropic features in the local galaxy distribution extending to gigaparsec scales. If correct, this result would directly contradict the cosmological principle and motivate inhomogeneous cosmologies. Here I analyse the same data and compare them with galaxy distributions predicted by the FLAMINGO cosmological hydrodynamic simulation, performed in the standard $Λ$CDM paradigm. I show that the apparent anomaly disappears when the correct comoving distance scale is used. I also show that, rather than violating the cosmological principle, the observed structures are consistent with those expected in a $Λ$CDM Universe.

Carbocations are key species in interstellar chemistry, providing entry points for building larger hydrocarbons. CH+, and more recently, CH3+, have been detected. Other carbocations await detection to provide a comprehensive view of the astrochemical network that is at work in the interstellar medium. We search for CH2+ in objects in which CH3+ was detected and evaluate the most favorable conditions for detecting the elusive CH2+ reactive cation. We calculated the CH2+ rotational and rovibrational transitions expected to contribute in the mid- to far-infrared, focusing on the lower-energy rovibrational levels. We then calculated CH2+ infrared emission spectra at different excitation temperatures and compared them to JWST spectra of the externally irradiated disk d203-506 in Orion, where CH+ and CH3+ have already been detected. We used thermochemical models to predict the abundance and spatial morphology of CH2+ to better understand its nondetection. The comparison to JWST spectra allowed us to provide excitation-temperature-dependent upper limits to the excited column density. These are several times lower than those detected for CH+ and CH3+ in their excited states. Based on model calculations for photodissociation regions and assuming similar excitation temperatures, the upper limit derived from observations and CH2+ model spectrum is either slightly above or below the column density expected from models of photodissociation regions. We provide a list of tabulated transitions to allow the community to search for this carbocation in future observations as CH2+ is key in providing observational constraints on astrochemical models.

We leverage the strong gravitational field of Abell S1063 to identify faint, highly magnified galaxies using ultra-deep James Webb Space Telescope (JWST)/NIRCam imaging from the GLIMPSE survey and ancillary Hubble Space Telescope (HST)/ACS imaging from the Hubble Frontier Fields program. We construct a photometric catalogue of lensed high-redshift candidates and use these sources to constrain the faint end of the rest-frame UV luminosity function (UVLF) over $z\simeq6$--11. Rather than treating the UVLF turnover ($M_{\rm t}$) as a hard cutoff, we model it as a gradual quadratic suppression and explicitly account for the potential continued contribution of galaxies beyond the turnover. In a shallow-turnover scenario, up to one-third of the UV luminosity density can arise from sources fainter than $M_{\rm t}$. While we find no direct evidence for a turnover down to $M_{\rm UV}=-13.5$ at $z=6$, our analysis can only confidently exclude weak, medium, and strong turnover models down to $M_{\rm t}=-15.9$, $-15.1$, and $-14.8$, respectively. Across these models, we infer lower limits of the UV luminosity, star formation density, and the ionization rate as: $ρ_{\rm UV}\geq22\times10^{25}\,{\rm erg\,s^{-1}\,Hz^{-1}\,Mpc^{-3}}$, ${\rm SFRD}\geq25\times10^{-3}\,M_\odot\,{\rm yr^{-1}\,Mpc^{-3}}$, and $\log_{10}(\dot{n}_{\rm ion}/{\rm s^{-1}\,Mpc^{-3}})\geq51.02$. We find that galaxies fainter than the conventional $M_{\rm UV}=-17$ limit contribute more than half of the UV luminosity density and at least $\sim64\%$ of the ionizing photons produced by star-forming galaxies at $z=6$. Because our turnover model permits a suppressed, but non-zero, galaxy population beyond $M_{\rm t}$, sources fainter than the turnover remain contributors to both $ρ_{\rm UV}$ and $\dot{n}_{\rm ion}$, emphasizing the need to consider the turnover and its shape during reionization.

We present the organisation and early results from the Observatory Science program of the Space-based multi-band astronomical Variable Objects Monitor (SVOM), based on data collected between July 2024 and December 2025. Although primarily designed for gamma-ray burst studies, SVOM's wide-field, multi-wavelength instruments enable a broad range of high-energy astrophysical investigations. We summarize the execution and performance of the General Program and Target-of-Opportunity observations, and we describe the frameworks used for serendipitous source detection and monitoring with the ECLAIRs coded-mask instrument. Over this period, SVOM carried out more than a thousand pointed observations and detected several hundred non-GRB high-energy sources, mainly X-ray binaries, as well as blazars, stellar flares, magnetars, and unidentified events. We highlight some key results, including the monitoring of the microquasar Cygnus X-1, the detection of burst oscillations from the Low-Mass X-ray Binary 4U 0614+091, the spectral-state monitoring of Aql X-1, the first SVOM detection of an X-ray blazar flare from 1ES 1959+650, and observations of a stellar flare from HD 22468. These results demonstrate SVOM's strong capabilities for time-domain astrophysics beyond its core GRB program.

Following its launch on 22 June 2024, the Space-based multi-band astronomical Variable Objects Monitor (SVOM) successfully completed its flight acceptance, commissioning, and scientific validation phases in early 2025, during which several tens of gamma-ray bursts (GRBs) were detected onboard. Three quarters of these events have also been detected by other satellites, and a quarter are SVOM-only GRBs. In this article, we describe these early GRB observations, with a first description of the SVOM GRB sample that is emerging, and of the level of characterisation already achieved, and with a focus on a few events of particular interest. These early results are very encouraging regarding SVOM's ability to detect and fully characterise (including prompt emission, afterglow and distance) a wide range of GRBs (classical long GRBs, short GRBs, X-Ray Flashes, etc.) and to enable the use of these extreme high-energy transients as probes of the distant Universe.

The Type I super-luminous supernova SN~2024jlc (ZTF24aapadbb) exploded on the 25th of May 2024 at $z = 0.039$. Being the closest supernova of this class discovered in recent years and one of the closest ever, represented a rare opportunity to study in detail this type of objects. We performed a multi-wavelength analysis, spanning ten orders of magnitude in frequency, including optical/UV photometry and spectroscopy, soft and hard X-rays, and high-energy $γ$-rays. We characterized the event as a slow-evolving and He-rich supernova, with one of the lowest peak luminosities reported for a super-luminous event $M_g\sim-19.37$ mag, and a light curve evolution compatible with both circumstellar interaction and magnetar spin-down models, with noticeable contribution from $^{56}$Ni decay. No significant excess was found in the soft and hard X-ray bands, for which we provide upper-limits on the flux. Additionally, we analyzed two years of \textit{Fermi}-LAT data, from which we report an intriguing hint of a $γ$-ray signal at the $\sim 3.6 σ$ level, although no firm detection can be claimed. The gamma to optical efficiency ratio, $η= 0.38$, is suggestive of the presence of a central-engine scenario, similar to SN~2017egm. Our analysis suggests that SN~2024jlc could bridge the gap between SLSNe and classical stripped-envelope supernovae. While still poorly populated, this bridge could consist of all SLSN-Ib supernovae, with the key difference residing in the powering mechanism.

The Square Kilometre Array (SKA), with its unprecedented sensitivity, frequency coverage, and large collecting area, is poised to revolutionize our understanding of the Cosmic Dawn (CD) and Epoch of Reionization (EoR) epochs marking the formation of the first luminous sources and the subsequent reionization of the intergalactic medium (IGM). However, detecting the faint redshifted 21-cm signal from neutral hydrogen remains one of the foremost challenges in observational cosmology, as it is buried beneath bright foregrounds from Galactic synchrotron radiation, free-free emission, and extragalactic point sources that are 4-5 orders of magnitude stronger than the cosmological signal. In this chapter, we highlight the key components and characteristics of these foregrounds and review ongoing efforts to model, characterize, and mitigate them. We emphasize how the SKA-Low AA* configuration, through its optimized array design, wide field of view, and improved calibration accuracy, enhances our capacity to suppress foreground contamination and recover the cosmological signal. The SKA Observatory Foreground Challenge plays a pivotal role in this effort by bringing together the global EoR/CD community to develop, compare, and validate foreground removal pipelines using realistic simulated datasets. Building on the experience of existing pathfinders such as LOFAR, MWA, and HERA, these collaborative initiatives are helping refine statistical and machine learning-based approaches for signal recovery. Together, these advancements are laying the groundwork for the SKA to probe the thermal and ionization history of the early Universe with unprecedented precision.

[ABRIDGED] We aim to determine abundances of carbon, nitrogen, sulphur, potassium, and copper for 714 nearby F and G dwarf and subgiant stars, and to re-derive oxygen abundances using updated corrections for departures from the assumption of local thermodynamic equilibrium. These elements extend the chemical inventory of our previous studies and provide new constraints on the relative enrichment histories of the Galactic thin and thick disks. The alpha-element behaviour of oxygen is confirmed, with old stars defining an enhanced sequence relative to young stars. Sulphur closely follows oxygen, while potassium shows broadly alpha-like behaviour in [K/Fe] but residual trends relative to oxygen. Carbon and nitrogen show only modest separation in [X/Fe], but much clearer population differences in [X/O]. Copper displays a strong metallicity dependence and clear separation between old and young populations when compared to oxygen. We also find that [O/Mg] is not constant, demonstrating that oxygen and magnesium provide complementary rather than interchangeable reference scales. Quantitative comparisons of all elements analysed in our studies show that carbon, oxygen, sulphur, and potassium rank among the most age-sensitive abundance ratios in the sample and provide strong discrimination between old and young disk populations. The new abundance measurements substantially expand the diagnostic power of this local stellar sample. The results show that abundance ratios relative to oxygen, together with precise stellar ages, reveal population differences that are partly hidden in traditional [X/Fe] trends. The expanded abundance inventory provides a homogeneous reference dataset for studies of Galactic chemical evolution, Galactic archaeology, and large spectroscopic surveys.

Tomographic measurements of gravitational lensing with different lens and source redshift distributions contain crucial information about the universe's relative expansion rate, and hence dark energy. While this technique is well-established in weak lensing, its application to strong lensing has traditionally focused on Double Source Plane Lenses (DSPLs). However, DSPLs are exceedingly rare and fundamentally limited by the Mass-Sheet Degeneracy (MSD), a systematic uncertainty underexplored in previous literature. To overcome these challenges, we introduce Pseudo Double-Source Plane Lenses (PDSPLs): pairs of independent single-source plane lenses with self-similar deflectors. This generalizes the DSPL formalism to the $\sim 10^5$ galaxy-galaxy lenses expected from upcoming surveys like LSST, Euclid, and Roman. Unlike true DSPLs, PDSPLs are free from the intermediate source mass problem by construction, eliminating the associated secondary MSD and the need for multi-plane ray tracing. We incorporate the deflector galaxy's MSD into a hierarchical forecasting framework, demonstrating that this degeneracy severely degrades constraints from small DSPL samples, thus motivating our PDSPL statistical approach. We forecast constraints on the dark energy equation of state under a Flat $w_0w_a$CDM cosmology. The LSST 10-year photometric sample alone achieves $σ(w_0) \sim 0.45$, while simultaneously constraining the MSD parameter and deflector power-law slope to $\sim 2\%$. Adding a prior $\mathcal{N}(0.3, 0.05)$ on $Ω_{\rm m}$ -- simulating combination with external probes like CMB, BAO, or SNe Ia -- tightens this to $σ(w_0) \sim 0.29$, competitive with current Stage III weak lensing analyses. Notably, this massive photometric sample outperforms smaller subsets with precise spectroscopic follow-up (e.g., from 4MOST), confirming statistical volume dominates over per-pair precision.

We determine with high accuracy the energy of the inner-shell transition $1s^2 2s^2~{}^1\mathrm{S}_0 \rightarrow 1s~2s^2~2p_{3/2}~{}^1\mathrm{P}_1$ ${}^{16}\mathrm{O}_{Kα}^{4+}$ at $554.372(3)~\mathrm{eV}$ ($λ$ = $22.36480(12)~\unicode{x212B}$) as well as its small shift of $2.2 \pm 1.3~\mathrm{meV}$ ($Δλ$ = $0.089(52)~\mathrm{m}\unicode{x212B}$) for the ${}^{18}\mathrm{O}$ isotope. This transition blends with a $K_α$ line of $\mathrm{O}^{5+}$ used in astrophysical diagnostics, potentially affecting its reliability. In contrast to our experimental uncertainty of $\pm 3~\mathrm{meV}$, advanced electronic structure predictions for this four-electron system, including quantum electrodynamic (QED) corrections on the order of $100~\mathrm{meV}$, still scatter by more than $\pm 250~\mathrm{meV}$. Ions generated and stored in an electron beam ion trap were excited at the ELETTRA synchrotron facility with monochromatic soft x rays, with photon energies corrected by an additional spectrometer. Upon resonant excitation of $\mathrm{O}^{4+}$ and subsequent autoionization, we separate the photoions of each isotope by a time-of-flight measurement. This way, we resolve soft x-ray isotopic shifts of a few meV, obtain very accurate data on an essential astrophysical ion, and test calculations down to the level of QED contributions.

Disk galaxies like the Milky Way are thought to evolve through internal dynamical processes: the stellar disk forms a bar, the bar drives gas inflow that builds a nuclear stellar disk, and the bar vertically thickens into an X-shaped bulge. Although this evolution is thought to be slow, completing only at late cosmic times, its timing remains poorly constrained. We report James Webb Space Telescope imaging of a galaxy at redshift 0.92 (7.6 billion years ago) that already hosts an X-shaped bulge, a nuclear stellar disk, and an extended stellar disk, with geometry and inferred bar size indistinguishable from those of present-day barred galaxies. The X-shaped bulge marks the completion of the major phase of bar-driven evolution when the Universe was less than half its current age.

Linear filamentary structures are fundamental constituents of the interstellar medium and play a central role in star formation. Their relative orientation with respect to the ambient magnetic field (B-field) provides key constraints on filament formation mechanisms. We investigate the relative orientation between Milky Way linear filaments (MWLFs) and the plane-of-sky B-field using polarization observations from the Atacama Cosmology Telescope (ACT) DR6, complemented by Planck data. Filament orientations are compared with the local B-field and the Galactic plane, while projection effects and statistical significance are quantified using Monte Carlo simulations of vector pairs in three-dimensions. We find no strong preferential alignment between MWLFs and the ambient B-field. Although the B-field is preferentially aligned with the Galactic plane with relative angles $θ_{\rm BG} \sim0-25°$, filament orientations exhibit a bimodal distribution, being either parallel or perpendicular to the plane ($θ_{\rm FG} \sim0-15°$ and $\sim75-90°$). Filaments located far from the Galactic midplane ($|z|>90$ pc) preferentially show perpendicular alignment with both the plane and the B-field, whereas those near the midplane exhibit a bimodal orientation. These results indicate that large-scale B-fields do not dominate the formation of MWLFs and instead favor a super-Alfvénic regime in which magnetic forces are dynamically subdominant, as expected for filaments associated with supernova-driven shells. Overall, our findings suggest that a face-on view of the Milky Way would resemble nearby disk galaxies such as M74, as observed in JWST images, with its disk structured by a network of supernova-driven bubbles (i.e., a bubbly disk).

Radial brightness profiles of disk galaxies often exhibit so-called breaks -- locations where their exponential-scale length abruptly changes. Some galaxies have downbending (Type II) breaks, where their brightness decays faster in outer regions, while other have upbending (Type III) breaks, resulting in more extended outer disks or envelopes. Disk radial profiles without any breaks (Type I) appear to constitute a minority. The exact fractions of different break types depend on many galactic parameters -- such as Hubble type, stellar mass, spatial environment, and bar presence -- and vary significantly across different studies. Another source of discrepancy is the orientation of galaxies: projection effects may play an important role in break detectability. In this work, we utilize DESI Legacy DR10 imaging to perform photometric decomposition of a sample of 375 edge-on galaxies and investigate their radial breaks. We find that the vast majority (~90%) of disks in our sample have Type II breaks, which is a considerably higher fraction than in many previous works (~50%). We carefully tested our results to check if observed breaks can be a result of flaring or two-disk composition. We showed that a high fraction of Type II breaks can be attributed to projection effects, which enhance the observed surface brightness of breaks in edge-on galaxies.

The galaxy SDSS1335+0728 brightened abruptly in December 2019 (the Ansky event) and has since been confirmed as the host of extreme X-ray quasi-periodic eruptions (QPEs) of debated origin. We constrain the origin of its transient activity by characterising the galaxy properties and nuclear accretion history with spatially resolved VLT/MUSE and JWST MIRI/MRS spectroscopy. We extract stellar and gas kinematics and emission-line fluxes, construct emission-line ionisation diagnostic maps, reconstruct the nuclear ionisation history via a Balmer-line light-echo analysis, and measure the mid-infrared silicate feature strength. The stellar kinematics reveal two counter-rotating stellar regions and kinematically cold gas ($σ_{\rm gas} \lesssim 60$ km s$^{-1}$), consistent with a past minor merger. Stellar populations show an old host with ongoing star formation confined to a ring at intermediate radii. Ionisation diagnostics reveal a three-zone structure: a central region powered by SMBH accretion, where high-ionisation coronal lines ([NeVI]$\lambda7.65μ$m, [NeV]$\lambda14.32μ$m, [OIV]$\lambda25.89μ$m) are confined, a star-forming ring, and a LINER-like outer region. A Balmer-line light-echo analysis yields a minimum ionising luminosity $\log L_{\rm ion,min} \approx 40.5$ erg s$^{-1}$ sustained over at least $\sim 1\,500$ yr. Broad silicate emission at 9.7 and 18$μ$m indicates optically thin dust, inconsistent with a classical active galactic nucleus (AGN) dusty torus. The data are consistent with two scenarios for the pre-2019 accretion: a persisting or gradually fading low-luminosity AGN, or a long-lived tidal disruption event (TDE) remnant disc. In both, Ansky corresponds to a slow, faint transient in a $\sim\!10^6\,M_{\odot}$ SMBH with already ongoing accretion, challenging the "faded AGN" interpretation proposed for some QPE hosts.

We develop a self-consistent framework based on a contact vector current-current interaction that couples the chemical potentials of both sectors through mutual mean-field shifts, with the dark matter (DM) fraction $F_D = N_D/N_B$ fixed as a global input parameter. This formulation provides a physically motivated alternative to fixed-density prescriptions, allowing the local DM density to follow the baryonic matter (BM) density throughout the stellar interior. As an application, we consider a mirror-DM scenario with exact symmetry between the dark and visible sectors and investigate NS matter using the NL3$ωρ$, FSU2R, NL3, and DDME2 equations of state (EOSs). We find that the DM--BM interaction weakens the binding of dense matter, reduces its incompressibility, and softens the EOS. Consequently, DM increases the central density and compactness of NSs, lowers their maximum masses, and shifts the onset of the direct Urca process to higher stellar densities. As a consequence, the onset of rapid cooling is shifted to more massive stars for models with a stiff symmetry energy and to less massive stars for models with a soft symmetry energy, depending on the extra compactness that results from the DM admixture. These results demonstrate that mirror-DM admixtures modify both the microscopic composition and macroscopic structure of NSs, with potential implications for their thermal evolution and multimessenger observational signatures.

We reveal a pristine, scale-invariant 3D Kolmogorov velocity cascade ($α_V^{\mathrm{3D}} \sim 2/3$) spanning $0.05$--$20$~pc in the Polaris Flare using \texttt{PPCOS} $^{12}\text{CO}$ data. A transition scale at $\sim 0.5$~pc marks a bifurcation in the structure functions' exponents, below which the degree of intermittency is also saturated. By deriving an analytical mapping relation ($α_V^{\mathrm{3D}}=α_V-\frac{1}{3}α_I$), we obtain the scale-invariant value of $α_V^{\mathrm{3D}}$, proving that the apparent transition stems from geometric projection and a changing density fractal dimension rather than a turbulent mode shift. Kolmogorov turbulence is smoothly inherited from the large-scale cold neutral medium, remaining uninterrupted by compression or gravity below 0.1 pc.

The 21cm line from neutral hydrogen is expected to be a ubiquitous (albeit faint) tracer of galaxies in the late Universe. With SKAO-MID, large wide-field surveys of several million HI-containing galaxies will become feasible, resulting in catalogues of sufficient size to measure large-scale structure observables such as baryon acoustic oscillations and redshift-space distortions. While optical galaxy surveys over comparable areas are generally deeper, radio surveys of this kind have a number of other advantages, such as broader sampling of the halo mass function and the possibility of measuring luminosity distances via the Tully-Fisher relation. In this chapter, we provide predictions for the galaxy number counts versus redshift that will be achievable with a wide-field HI galaxy survey on SKAO-MID, along with corresponding forecasts for cosmological observables. Given the substantial uncertainty in the HI mass function with redshift, we bracket our predictions using a handful of different modelling methods.

The intrinsic alignment (IA) of galaxies is one of the main contaminants to the weak lensing shear signal. Efforts to model it often assume that the alignment strength depends only on halo mass. In this work, we use the 2.8 Gpc box-size run of the $\mathtt{FLAMINGO}$ suite to show that alignment amplitude, in addition to halo mass, depends on the formation redshift of the host haloes for an LRG-like sample. We show that the assembly histories of galaxies and haloes influence the alignment signal. After correcting for their mass evolution, we find that haloes that formed earlier have a higher alignment amplitude, as do their central galaxies. We also explore the redshift evolution of the alignment signal by fitting the amplitude with a power-law in mass at different snapshots of the simulation. We find that the amplitude of this power-law increases steadily with redshift, while the slope decreases with redshift until $z\sim1$ and then flattens. We provide an empirical mass-redshift intrinsic alignment model fit on the $\mathtt{FLAMINGO}$ simulation. Furthermore, by tracking central galaxies across snapshots, we show that the alignment signal changes with redshift beyond that associated with the change in mass, and that galaxies tracked from higher redshifts have a larger amplitude. Our results indicate that IA modeling in weak lensing surveys cannot have arbitrarily small prior ranges, and complicate the implementation of HOD-based alignment models for gravity-only simulations. They also provide simulation-based guidelines for a redshift evolution model of IA for use in observational studies.

We estimate the age of the Universe using the Xiang & Rix sample of 247,103 Milky Way stars with high-resolution spectroscopy from LAMOST DR7 and $Gaia$ eDR3 parallaxes. Stellar ages were estimated using YY isochrones up to 20 Gyr. To remove stars with unusually high and precise ages, we require old stars to be metal-poor and $α$-enriched. We also require consistency between YY ages and those obtained with FLAME based only on $Gaia$ data. Our final sample of 155,600 stars within 5 kpc provides consistent cosmic age estimates using several techniques of increasing rigour. Our main results use an MCMC reconstruction of the latent age distribution, though our iterative reconstruction is very similar. Applying an innovative approach to our MCMC reconstruction and its uncertainties, we find that the oldest star has an age of $A_\star = 13.73^{+0.18}_{-0.15}$ Gyr. Varying the quality cuts can at most reduce this to $A_\star = 13.31^{+0.21}_{-0.18}$ Gyr or raise it to $14.02^{+0.18}_{-0.15}$ Gyr using a much lower or higher age-dependent metallicity ceiling, respectively. Our inferred $A_\star$ is consistent with the 13.6 Gyr expected in CMB-calibrated $Λ$CDM, assuming the first long-lived stars formed when the Universe was 0.2 Gyr old. This agreement casts doubt on solutions to the Hubble tension solely through new physics prior to recombination, which generally imply a cosmic age of $12.9 \pm 0.2$ Gyr to match low redshift probes. It is difficult for stellar modelling uncertainties to reconcile such a low age with our result given the low metallicities of the oldest stars in our sample and independent asteroseismic constraints.

Our goal is to find chemical diagnostics to determine the physical conditions in protostellar envelopes and help establish the development of matter during the formation of a low-mass star, as well as investigating a possible variation of sulphur depletion during the star formation process at the scale of the cold envelope. With observations with the Yebes-40m and IRAM-30m telescopes, we estimate column densities of sulphur-bearing species in the Class 0 objects HH212 and NGC1333 IRAS4A. A neural emulator of the chemical code Nautilus is used to constrain the chemical time, density, gas temperature, cosmic ray ionization rate, and sulphur elemental abundance in the cold envelope of these objects. We compare the resulting abundances of these species with those towards the Class 0 object B335. While sulphur-bearing species containing carbon chains are between 3 and 7 times more abundant in B335 than in the other two objects, sulphur oxides and nitrogen-bearing species are 3 times more abundant in NGC1333 IRAS4A. Our chemical modelling shows that, while the chemistry of HH212 and NGC1333 IRAS4A is well reproduced considering a gas temperature of 25 K and a sulphur depletion of a factor of 100 in their envelopes, significant differences are found in their average density and cosmic ray ionization rate. Comparing with similar studies in pre-stellar and protostellar cores, we derive an increase in the SO/CS and SO$_2$/C$_2$S ratios of about two orders of magnitude and a potential decrease in the HCS$^+$/CS ratio of a factor of 10 in the transition from the pre-stellar to the Class 0 phase. Sulphur compounds are good evolutionary tracers of the pre- to protostellar phase transition, with oxygen-bearing species being more abundant than those containing carbon in the evolved sources. Nonetheless, sulphur depletion in the cold envelope of Class 0 objects remains similar to that in starless cores.

An accurate estimation of the mass of galaxy clusters is key to precisely and unbiasedly constraining cosmological parameters through their number count. The hydrostatic mass, estimated from the properties of the intracluster medium (ICM) assuming hydrostatic equilibrium, sphericity, and thermal-only pressure, is known to be biased by 10 to 20%, most likely due to non-thermal pressure support from gas motions. Two corrections have been proposed: i) replacing the thermal pressure by the total pressure $P_\mathrm{tot}=P_\mathrm{th}+P_\mathrm{nth}$, or ii) adding effective mass terms derived from the gas momentum equation. We compare these approaches using a numerical replica of the Virgo cluster as a case study, estimating corrected masses from 3D radial profiles in different cluster regions and from projected sightline velocities mimicking XRISM observations. We find that the two methods do not yield the same results in 3D: the non-thermal pressure correction increases the mass by a growing amount with radius (from a few per cent in the core to $\sim$40% at the virial radius), whereas the effective mass terms provide a correction that varies less with radius. When estimated from projections, the two methods agree to within a few per cent for a given sightline, but the non-thermal pressure fraction is underestimated by about a factor of 2 compared to the 3D case. Furthermore, projection effects can change the inferred non-thermal pressure fraction by up to a factor of 2, particularly when the sightline is aligned with cosmic filaments.

We analyze the effective-spin distribution of binary black hole mergers in GWTC-5.0 as a function of primary black hole mass using hierarchical Bayesian inference. We model the population as a mixture of two spin components separated by a transition mass scale inferred directly from the data. We find strong evidence for a transition at $\tilde{m} = 15.2^{+4.3}_{-3.6}\, M_\odot$. Mock-catalog analyses show that such a transition is unlikely to arise from finite-sample fluctuations of a mass-independent $χ_{\rm eff}$ population and the posterior predictive distributions of $χ_{\rm eff}$ inferred below and above the transition are clearly distinct. Below the transition mass, the effective-spin distribution is narrow, peaks at a small positive value $χ_{\rm eff}>0$, but also shows significant support for negative $χ_{\rm eff}$. Above the transition, the distribution is broader and its peak shifts to values consistent with $χ_{\rm eff}\simeq0$, making its support at both positive and negative $χ_{\rm eff}$ roughly similar. These findings suggest that the dominant merger population concentrated around $10\,M_{\odot}$ is statistically distinct from the rest and that it arises from a different formation channel. We show that this low-mass population is broadly consistent with formation from massive stellar multiples in the field: it may either arise from isolated binary star evolution but only if black hole natal kicks below $\tilde{m}$ are generally very large ($\gtrsim100\,\rm km/s$) or be caused by the dynamical evolution of hierarchical triples. In contrast, isolated binary evolution with standard fallback kick models cannot reproduce the support for negative $χ_{\rm eff}$.

The SPHEREx all-sky survey has now measured the R$\sim$40-100 infrared spectra of thousands of nearby brown dwarfs in the chemically rich 0.75-5 $μ$m range. The survey's wide spectral coverage and high S/N permits flux measurements that capture several broadband molecular absorption features, and upwards of 80$\%$ of the total bolometric luminosity of most brown dwarfs. Atmospheric models are known to yield systematic disagreements in the inferred temperatures and radii of brown dwarfs, necessitating benchmarking against observations. In this work, we present SPHEREx spectra across a broad sequence of 37 nearby field brown dwarfs, ranging from L0 to Y4 ($\sim$2500-250 K) and compare them to theoretical expectations. We additionally compile spectra for separate low-gravity and low-metallicity objects, and show how they trend with constant spectral type. We fit the measured spectra to the well-known forward model grids Sonora Diamondback, Elf Owl, BT-Settl, ATMO2020 and ATMO2020++ and compare their goodness-of-fit as a function of wavelength, spectral type, and treatment of clouds and chemistry. We find that the models continue to struggle to simultaneously fit the J/H/K peaks and the 4 $μ$m opacity window, especially in L/T transition objects. The largest deviations appear around the chemistry-sensitive CO$_2$ and CO features. Despite these offsets, the models broadly capture their trends across the L/T transition, with the observed sample of field dwarfs strongly preferring the weak vertical mixing ($k_\mathrm{zz}$ = 10$^4$ cm$^2$s$^{-1}$) Elf Owl models over strong mixing. The spectra shown here along with future SPHEREx data will help guide improvements to models.

FRB20250613A is a repeating FRB discovered by the Australian SKA Pathfinder and localised to a low-metallicity dwarf galaxy at a redshift of $z = 0.0987 \pm 0.0001$. FRB 20250613A exhibits a plethora of exotic features that likely overlay the imprint of the circum-burst environment on some intrinsic features of the source. Here we perform a comprehensive analysis of bursts detected by ASKAP, MeerKAT, and the Murriyang Parkes radio telescopes. Bursts during the MeerKAT epoch show a large apparent variance in scattering on timescales of minutes to hours. Polarimetric analysis of the full sample shows spectral depolarisation with variability on timescales of days and changes in rotation measure of $\sim$ 300 rad m$^{-2}$ over days to months. This suggests a highly turbulent magneto-ionised environment. We find significant preference for separations of $\sim$6.8$\pm$0.8 ms in multi-component bursts that we suggest is likely intrinsic to the burst emission mechanism. Finally, we find that a subset of bursts exhibit variations in these propagation effects on burst components separated by just milliseconds, that are difficult to explain by changing sightlines, but plausibly due to non-linear plasma effects in the circum-burst environment caused by the high field strength of the FRB emission. These properties, which demand a nearby turbulent screen of material, are all consistent with the FRB progenitor being embedded in the dense stellar wind of a Be star binary companion, objects which are relatively plentiful in low-mass and low-metallicity galaxies like the FRB20250613A host.

We applied the barred population-orbit superposition method developed in \citet{Jin2025a,Jin2025b} to construct 3D chemo-dynamical models for the barred S0 galaxy NGC~1381 in the Fornax cluster. Based on the stellar orbits in the models, we decomposed NGC~1381 into six components: (1) a dynamically warm nuclear disc with $f_{\rm nucl}\sim5\%$; (2) a rigidly rotating, BP/X-shaped bar with $f_{\rm bar}\sim30\%$; (3) a dynamically hot, spheroidal bulge with $f_{\rm bulge}\sim17\%$; (4) a dynamically cold thin disc with $f_{\rm thin}\sim28\%$; (5) a vertically extended thick disc with $f_{\rm thick}\sim16\%$; and (6) a dynamically hot, spatially diffuse stellar halo with $f_{\rm halo}\sim5\%$. The nuclear disc, bar, and thin disc are metal-rich ($[Z/\rm H]\gtrsim0$), $α$-poor ($\rm[Mg/Fe]\lesssim0.2$), and old ($\sim13\rm\,Gyr$), corresponding to in situ formation in the early Universe. The bulge, thick disc, and stellar halo are metal-poor ($[Z/\rm H]\lesssim0$), $α$-rich ($\rm[Mg/Fe]\gtrsim0.2$), and younger than or comparable in age to the in situ components, suggesting their relations with ex situ formation contributed by minor mergers. The flat metallicity and [Mg/Fe] gradients in the thick disc and stellar halo indicate they are dominated by a similar population of ex situ stars. In contrast, the bulge exhibits a negative metallicity gradient ($\nabla[Z/\rm H]_{bulge}<0$) pointing to a more complex formation history: the bulge could be either predominantly ex situ or contain a non-negligible mixture of in situ and ex situ stars. Our modelling also reveals the presence of a slow bar ($\mathcal{R}=2.40_{-0.27}^{+0.54}$), with a bar pattern speed of $\rmΩ_p=34_{-7}^{+4}\,km\,s^{-1}\,kpc^{-1}$, a bar length of $R_{\rm bar}=2.24_{-0.22}^{+0.43}\rm\,kpc$, and a corotation radius of $R_{\rm CR}=5.38_{-0.28}^{+1.59}\rm\,kpc$, which is consistent with its ancient formation time.

SNe Ibn and SNe Icn are stripped-envelope explosions whose optical emission is commonly linked to interaction with H-poor circumstellar material (CSM), whereas fast blue optical transients (FBOTs) form an observational class of rapidly evolving, blue, and luminous events with diverse proposed power sources. We present a uniform comparison of these transients to test whether they are separated in optical light-curve and fitted physical parameter space. We compile multiband optical light curves of 25 SNe Ibn, SNe Icn, and FBOTs, measure same-band observables with Gaussian-process reconstructions, and model the data with the unified \texttt{TransFit-CSM} framework. In the observed (g)-band peak-luminosity--rise-time and decline--rise-time planes, the three classes are not cleanly separated: FBOTs preferentially occupy the luminous and rapidly evolving end of the distribution, but show limited overlap with part of the Ibn/Icn population. Their extinction-corrected peak colors span a broadly overlapping blue region, with FBOTs extending to bluer colors. Unified CSM-interaction fits, including shock heating and an effective inner heating component, yield overlapping CSM and ejecta parameter distributions. These results indicate that the optical light curves of SNe Ibn, SNe Icn, and at least some FBOTs can be compared within a common dense-CSM interaction framework, while the most extreme FBOTs may still require additional power sources or non-thermal components.

We investigate whether galaxies identified as superthin in optical images remain superthin in the near-infrared (NIR), and how their extreme disk morphology is related to environment. From a nearby volume-limited sample, we select 210 superthin galaxies using two-dimensional bulge/disk decomposition of SDSS $r$-band images, requiring the disk component to have a major-to-minor axis ratio $a/b>9$. We measure disk shapes from SDSS $griz$ to UKIDSS $JHK$ bands. Both the major- and minor-axis scales decrease from the optical to the NIR, reaching $\sim0.6$ of their $r$-band values in the $K$ band, but the disk axis ratio remains nearly unchanged. Thus, optically selected superthin galaxies remain superthin in the NIR, implying that the old stellar populations traced by NIR light do not form a prominent thick disk. Reanalysis of our sample and a previous superthin sample shows that earlier reported NIR thickening is mainly due to a magnitude- and band-dependent bias in one-dimensional fitting. We further compare their environments with matched control samples using projected cross-correlations, reconstructed local overdensities, and large-scale-structure classifications. Superthin galaxies show lower clustering on $\sim0.1$--$1\,h^{-1}\,\mathrm{Mpc}$ scales and lower overdensities at $1\,h^{-1}\,\mathrm{Mpc}$, but no clear residual dependence on large-scale-structure type. These results suggest that superthin galaxies are preferentially central galaxies in relatively low-mass dark matter halos, consistent with a picture in which high host-halo spin helps build and preserve extended, vertically thin stellar disks.

Neutron star mergers are a leading site of $r$-process, producing radioactively powered optical and infrared transients known as kilonovae. Observations of the kilonovae AT2017gfo, associated with the gravitational-wave event GW170817, and AT2023vfi, associated with GRB 230307A, have enabled measurements of the mass of ejected $r$-process material and the identification of heavy elements in the ejecta. However, late-time observations reveal strong infrared emission with temperature below 1000 K, which is difficult to explain by atomic absorption and emission processes alone. In this paper, we show that kilonova ejecta provide conditions favorable for the formation of dust grains composed of refractory $r$-process elements including Zr, W, and Os. We calculate the kinetic formation of dust grains using reaction rate coefficients of W as a proxy, finding that dust forms efficiently, particularly in slow ejecta. This stands in contrast to a previous study that relied on a classical nucleation framework. By performing radiative transfer simulations that incorporate dust formation, we demonstrate that $r$-process dust naturally explains the observed late-time infrared emission. The formation and abundance of $r$-process dust are highly sensitive to the ejecta mass, composition, and expansion velocity. Infrared emission from $r$-process dust can therefore serve a new probe of heavy-element production in neutron star mergers.

Spatially resolved maps of gas-phase metallicity provide key constraints on the chemical enrichment and mixing processes that drive galaxy evolution, but measurements based only on strong lines remain highly uncertain and dependent on emission-line coverage. In this work, we present a joint analysis of SITELLE observations, covering the $[O II]λ\lambda3726,3729$ doublet, with PHANGS-MUSE spectroscopy covering 4800-9300 Angstroms, including $Hβ$, $[O III]\lambda4959,5007$, $[N II]\lambda6584$, $Hα$, $[S II]λ\lambda6716,6731$, and $[S III]\lambda9069$, within five nearby spiral galaxies. By combining these data, we construct a homogeneous catalog of emission-line fluxes for 604 ionized nebulae, 556 of which are classified as H II regions. This enables a comparison of eight widely used strong-line metallicity calibrations, five new strong-line calibrations, and an investigation of ionization-parameter diagnostics. We recover known systematic offsets among calibrations, but also find that many exhibit very low scatter, less than 0.03-0.04 dex, in radial metallicity gradients. We find that $[S III]/[S II]$ exhibits minimal secondary dependence on metallicity or extinction, suggesting that it may be a more robust tracer of ionization parameter than $[O III]/[O II]$. No significant outliers are identified in O/H or N/O within the sampled regions, indicating internally consistent abundance trends across the inner disks probed by our data. We provide a publicly available catalog of all measured emission-line fluxes, designed to support future investigations, including temperature modeling and strong-line abundance calibrations.

This study aims to test whether a supermassive binary black hole (SMBBH) system with a triple-disk accretion structure can explain the observed $\sim$173-day periodic microlensing variations and spectral energy distribution (SED) of the gravitationally lensed quasar Q J0158-4325. We construct a triple-disk model for the SMBBH system, incorporating realistic accretion disk structures, orbital motion, and microlensing effects. The model is used to simulate optical and X-ray microlensing light curves and SEDs, which are compared with long-term optical monitoring, X-ray observations, and UV-optical spectra from HST and XSHOOTER. Bayesian analysis and MCMC fitting are applied to constrain model parameters. The model successfully reproduces the periodic microlensing variations. Combined light curve and SED fitting favor a high mass ratio ($q>0.5$) SMBBH system with total mass $\sim 10^{9.5}M_\odot$, and nearly equal-mass binaries ($q\sim1$) provides the best agreement with both the optical/UV spectrum and the microlensing signal. This model predicts larger X-ray microlensing amplitudes than in the optical, but, the available X-ray observations lack the precision needed to place strong constraints. We emphasize the need for future high-cadence monitoring to resolve remaining uncertainties. This study demonstrates the effectiveness of combining multi-wavelength microlensing signatures with spectral modeling to provide robust constraints on SMBBH systems, with the developed framework applicable to other lensed quasars for identifying and characterizing candidate SMBBHs.

Type Ia supernovae (SNe Ia) serve as one of cosmic standard candles, but their exact progenitor channel is still an open question. SNe Ia commonly come from binary star evolution. Therefore, one of the major differences among the proposed progenitor channels is whether there is a more-or-less intact companion star remaining at the time of explosion, which causes the SN ejecta to be more asymmetrical. As the SN ejecta evolved into supernovae remnants (SNR), the imprint formed by the companion interaction may affect the morphology of the SNR. In addition, the progenitor systems may have experienced different mass transfer histories and therefore led to formation of different circumstellar material (CSM) environments, which may also affect the early evolution of SNR. In this study, we use GADGET and RAMSES codes to simulate these physical effects and follow the evolution into early-phases of SNRs. In our simulations, we consider different ejecta models and track the element distribution. We compare our simulation with actual observations and conclude that despite some SNRs having morphology resemblance to our simulation results, their highly asymmetric expansion rates are hard to explain by interaction between SN ejecta and a companion star alone.

The microlensing of lensed quasars presents a promising avenue for understanding the structure of accretion disks around supermassive binary black holes (SMBBHs). We investigated the microlensing signatures in multiband (optical, UV, and X-ray) light curves of active SMBBH systems, focusing on how these signatures depend on the mass ratio, separation, and accretion rate. We analyzed the periodic fluctuations in microlensing light curves induced by the orbital motion of SMBBHs. We examined the relation between the mass ratio and the period of variations in light curves across optical, UV, and X-ray bands. We find that the periodic fluctuations in the light curves depend on the mass ratio of the black holes: for nearly equal masses, variations occur at half the orbital period, whereas for low mass ratios, the period corresponds to the orbital period influenced by the secondary mini-disk. Furthermore, all optical, UV, and X-ray light curves exhibit the same period and phase, but the amplitude of variation is greater in the UV and X-ray bands than in the optical bands. These light curves provide insights into the motion and radiation regions of the disks through wavelength-dependent periodic variations, although they yield limited constraints on the system's black hole mass or Eddington ratio, which can instead be derived from the spectral energy distribution (SED). Integrating microlensing data with SED observations is crucial for accurately constraining the parameters of SMBBH systems.

Soliton cores in fuzzy dark matter (FDM) deepen nuclear potentials and have been proposed to strongly boost Bondi accretion, potentially aiding rapid black-hole growth at high redshift. We test this in live Schrodinger-Poisson FDM cores coupled to isothermal gas, evolving a moving black hole that grows via a strictly mass-conserving sink. We measure boosts relative to the initial mean-density Bondi rate. Low-mass seeds, with initial black-hole masses less than about 10^6 solar masses, do not sustain large boosts: black-hole wandering and soliton sloshing drive bursty accretion, with dense gas only intermittently present near the black hole. Intermediate seeds, with initial black-hole masses around 10^7 solar masses, produce the most durable enhancement, reaching boosts of order 100 for sound speed cs = 60 km/s, while hotter gas approaches near-background Bondi rates. High-mass seeds, with initial black-hole masses around 10^8 solar masses, quickly exhaust the sink-scale reservoir and become supply-limited, suppressing long-lived growth despite the deepened potential. In general, central-potential deepening, for example by a soliton halo, does not guarantee long-lived fueling: sustained boosts emerge only when the black hole remains dynamically confined within the dense nuclear gas region. Our results suggest that SMBH formation channels relying on soliton-enhanced accretion alone are unlikely to provide sufficient early growth.

Magnetic fields are a fundamental component of astrophysical systems, yet many key questions about their origin, amplification, and role in structure formation and evolution remain unresolved. The SKA will mark a transformational step forward in addressing these questions, enabling studies of cosmic magnetism across a large range of spatial scales and environments. This overview summarizes the main science cases in the Cosmic Magnetism Science Working Group, which cover a huge breadth of scales from the smallest scales governing planet and star formation, all the way up to the large-scale structure of the Universe. The chapter summarizes the main observational techniques for studying magnetic fields, including direct polarization imaging, Faraday rotation, rotation-measure grids, and Zeeman splitting. We also address fundamental considerations of these studies including SKA-Low vs SKA-Mid and wide-area vs deep observing strategies.

UGC 2369S is a luminous infrared galaxy (LIRG) undergoing a late-stage merger in a triple system, where the heavily obscured northern core is suspected to host an active galactic nucleus (AGN). However, severe dust and gas obscuration makes definitive confirmation challenging. We aim to provide direct observational evidence for the buried AGN through high-resolution radio imaging, while investigating the AGN accretion and feedback properties within this merger-driven gas-rich environment. We analyzed archival European VLBI Network (1.6 GHz) and Very Long Baseline Array (1.7 and 5 GHz) data of UGC 2369S. Through high-resolution imaging and visibility-domain Gaussian modeling, we characterized the morphology and intensity of its milliarcsecond-scale radio emission. A compact radio component is detected at the northern core, exhibiting high brightness temperature ($T_{\rm b}>10^7$ K) and flat radio spectrum ($α\approx -0.45$), which confirms the presence of an obscured AGN. The sub-Eddington accretion rate ($λ_{\rm Edd} \approx 2.7 \times 10^{-4}$) indicates that it falls within the radiatively inefficient accretion flow (RIAF) state. We provide direct imaging evidence for an AGN in the northern core of UGC 2369S, revealing a deeply buried, jet-emitting low-luminosity AGN (LLAGN) enshrouded by a Compton-thick gas cocoon. This demonstrates that VLBI is a uniquely effective tool for disentangling nuclear accretion and feedback processes within the heavily obscured environments of multiple-merger systems.

We systematically organize the radio-emitting components in radio-quiet active galactic nuclei (RQ AGN), including jets, accretion disk coronae, dust, ionized gas outflows, and circumnuclear star formation. We present a diagnostic framework for distinguishing these components using spectral turnovers and spectral indices produced by synchrotron self-absorption (SSA) and free-free absorption (FFA), together with brightness temperature and peak frequency. The central premise is that the observed spectral index and its spatial distribution are not unique source properties unless the observing beam is specified: changing the angular resolution changes the physical scale being sampled and therefore changes the mixture of radio-emitting components. By exploiting this scale dependence with SKA1-MID and SKA-VLBI, spatially resolved spectral-index mapping will reveal which physical processes dominate from circumnuclear star formation on $\sim$100 pc scales to jets, coronal emission, and compact ionized gas on parsec and sub-parsec scales. Through multi-frequency continuum imaging and spectral-index mapping, SKA observations will provide a multi-scale physical view of radio-quiet AGN that links radio emission mechanisms to accretion, obscuration, and feedback.

Global 21cm experiments aim to detect the hydrogen 21cm signal by separating it from foreground emission that can be orders of magnitude brighter than the signal. REACH (the Radio Experiment for the Analysis of Cosmic Hydrogen) forward-models the sky by jointly fitting signal and foreground spectral parameters to an existing sky map. The fitted parameters yield spectrally constrained, absolutely calibrated maps of the radio sky across the full 50-170 MHz observing band, among the lowest continuous frequencies yet mapped. We assess REACH's ability to fit the 21cm signal and recover accurate foreground maps, using physically motivated foreground models of increasing complexity (starting from a pure synchrotron power law model, then introducing variable amplitudes, curvature, and a free-free component). We evaluate these models against simulated REACH observations of correspondingly complex foregrounds, based on the Global Sky Model and the Python Sky Model. To recover the 21cm signal, more complex datasets require correspondingly complex models, but this introduces degeneracies which limit accurate recovery of foreground parameters. Fitting a foreground with independent synchrotron and free-free emission enables component-separated sky mapping, which has applications beyond radio cosmology; synchrotron is well-recovered across the sky, but free-free recovery is limited. REACH is therefore capable of probing Galactic physics at uniquely low frequencies, alongside its primary goal of detecting the 21cm signal.

The chemistry of stars provides powerful insight into the history of the Milky Way. With multiple large-sky spectroscopic surveys that are currently available, using chemistry as a means to study the evolution and history of the Milky Way has flourished. Open clusters have long been used as landmarks to calibrate different age dating methods (e.g., gyrochronology and asteroseismology). In this work, we utilize the SDSS-IV/APOGEE-based Open Cluster Chemical Abundances and Mapping (OCCAM) survey as our foundation for new optical observations; enabling us to characterize neutron-capture abundances for known cluster members. For 56 stars in 18 open clusters, we collected high-resolution (R > 50,000), high-S/N (>75 at 5500A), spectra from Keck I and Magellan Baade telescopes. With these data, we derive abundances for 23 elements using BACCHUS, including 7 neutron capture abundances not measurable by APOGEE. Finally, we characterize the radial distribution of these neutron-capture elements in the Milky Way. We find that the second-peak s-process and r-process abundances exhibit relatively flat gradients in the Milky Way. Although not as distinct, the first-peak s-process abundances also have slopes which are shallower than the alpha and iron-peak elements. The differences in the neutron-capture gradients from the lighter elements not just indicates the sources producing these elements are fundamentally different, but that the timescales on which they are produced also differ (especially for the r-process). Moreover, a metallicity dependence of the AGB stars responsible for producing the heaviest s-process abundances may be necessary to consider in Galactic evolution models.

The masses and ages of stars are key quantities for understanding exoplanetary, stellar, and galactic evolution. In the context of Gaia, these parameters provide insights into the stellar populations, helping to trace the formation and history of the Galaxy. As part of the Gaia Data Processing and Analysis Consortium (DPAC), the Final Luminosity Age Mass Estimator (FLAME) pipeline processes Gaia data to derive stellar parameters comprising luminosities, radii, masses and ages. This paper discusses the methods and data used in FLAME for Gaia Data releases and the expected performances of FLAME for the 4th Gaia Data Release. FLAME comprises two main components: the first one, which is analytical, is used to estimate luminosity, radius, and radial velocity correction due to gravitational redshift by exploiting the atmospheric, astrometric, and photometric parameters produced within Gaia. The second is a model inference based on two main approaches: a classical minimization approach, and a Bayesian framework. It aims to derive mass, age, and evolutionary stage. The two step implementation offers flexibility in handling photometric properties that are prone to systematic errors. Tests with simulated data, the Sun, and well characterised samples of stars show that the methods in FLAME perform as expected, producing results in statistical agreement with the literature. We provide new stellar fundamental parameters for some high velocity stars, stars with very low mass companions, and a selection of stars in the Plato Field of View. In Gaia Data Release 4 approximately 500 million sources will have results from the pipeline. [abridged]

We present a broadband spectral analysis of the $γ$-ray emission from the pulsar wind nebula HESS~J1825$-$137, combining observations from Fermi Large Area Telescope (\textit{Fermi}-LAT), High Energy Stereoscopic System (H.E.S.S.), High-Altitude Water Cherenkov Observatory (HAWC), and Very Energetic Radiation Imaging Telescope Array System (VERITAS) across the $\sim 0.1$~GeV--$160$~TeV energy range. The spectral energy distribution is modelled under purely leptonic, purely hadronic, and lepto-hadronic scenarios using the \textsc{Naima} radiative modeling framework with Markov Chain Monte Carlo parameter estimation. Model comparison via the Bayesian Information Criterion reveals that the baseline GeV--TeV data favour a purely leptonic interpretation, while the inclusion of simulated Cherenkov Telescope Array Observatory (CTAO) observations or Large High Altitude Air Shower Observatory (LHAASO) ultra-high-energy (UHE; $E_γ \ge 100\,\mathrm{TeV}$) measurements shifts the preference toward models incorporating a hadronic component ($Δ\mathrm{BIC} = -28.87$ and $-7.89$, respectively). The inferred electron energy budget for the baseline GeV--TeV dataset, $W_e = 4.25 \times 10^{48}$~erg, is consistent with previous estimates reported in the literature. The proton energy budget, $W_p \approx 2.5 \times 10^{48}$~erg, is energetically compatible with $pp$ interactions in the dense molecular environment adjacent to the nebula. These results demonstrate that precise spectral measurements above $\sim 10$~TeV, where Klein--Nishina suppression of inverse Compton emission creates a window for hadronic processes, are essential to establish the dominant emission mechanism in this source.

We present soft X-ray spectroscopy of the black-hole X-ray binary V4641~Sgr with the \textit{XMM-Newton} Reflection Grating Spectrometer (RGS). The RGS spectrum shows narrow emission features from N\,\textsc{vi--vii} and O\,\textsc{vii--viii} superimposed on a partially covered disk blackbody continuum. A blind Gaussian search confirms the presence of significant lines at the expected rest wavelengths. He-like triplet ratios (high $G$, low $R$) and full photoionization modelling both indicate a dense, photoionized plasma. Small redshifted velocities of $\sim 540$--$720\ \mathrm{km\ s^{-1}}$ are suggested, which are consistent with quasi-static or slowly flowing gas away from the observer after accounting for systematics. Photoionization modelling requires two \textsc{xstar} components with an intermediate ionization parameter ($\logξ\simeq 3.1$) and a low ionization parameter ($\logξ\simeq 0.36$), respectively. The simultaneous EPIC-pn spectrum suggests highly ionized Fe emission structures, hinting at an additional, more highly ionized component. These results imply the existence of a radially extended, multiphase, and dense disk atmosphere in the source. We compare the source with other X-ray binaries showing similar emission lines. V4641~Sgr shares a similarly high inclination with other sources; however, the presence of low ionization emission lines distinguishes it from the rest.

The persistent tension between early- and late-Universe measurements of the Hubble constant (H0) remains on of the most significant challenges in modern cosmology. The Spectroastrometry and Reverberation Mapping (SARM) method offers a promising, calibration-independent approach to address this issue by combining time-delay measurements of the Broad-Line Region (BLR) with interferometric angular size determinations. Current implementations of SARM, however, are limited by the difficulty of performing near-infrared reverberation mapping (RM) on the same emission lines observed by GRAVITY, restricting applications to only a few bright AGN. We propose using the capabilities of SHARP, the next-generation near-infrared spectrograph for the Extremely Large Telescope (ELT), to overcome these limitations. SHARP's sensitivity and multi-object spectroscopy will enable (1) efficient long-term monitoring of existing GRAVITY targets with minimal time investment, and (2) systematic RM campaigns for the fainter AGN that will be observed by GRAVITY+. These advances will give us precise infrared lags for tens of AGN, enabling geometric distance measurements and a robust, calibration-free determination of H0. Beyond cosmology, SHARP will allow detailed studies of BLR structure and kinematics in the infrared, advancing our understanding of AGN physics and with repercussion on the measurements of Supermassive Black Holes (SMBH) masses.

The James Webb Space Telescope (JWST) detected numerous massive and relatively compact stellar clumps around proto-galaxies at high redshift (z>0.5). Their properties suggest that these systems may represent proto-globular clusters (GCs), but their possible connection to local old GCs is poorly understood. In this Letter, we explore the dynamical evolution of proto-star clusters, building the missing evolutionary link between high-z systems observed by JWST and local GCs. Our simulations include the effects of stellar interactions, stellar evolution, and the strong time-dependent cosmological tidal field in which these proto-star clusters evolve. We also explore the role of multiple stellar populations and stellar-mass black holes (BHs), two fundamental ingredients in stellar cluster dynamics. We show that systems hosting multiple populations (as routinely observed in local GCs) are more likely to endure the early strong tidal field than single-population clusters. In addition, after 12 Gyr, such systems have properties consistent with those of Galactic GCs. Our work confirms that the high-z clumps observed by JWST can be the progenitors of the local GCs. Finally, we show that a population of stellar-mass BHs within a proto-star cluster favors its disruption, but that surviving systems can retain a sizable population of BHs.

We study the feasibility of natural and directed panspermia via interstellar objects (ISOs) like 3I/ATLAS. The paper is organized around two questions. First, could natural panspermia occur if microbes or biomolecules survived inside shielded ice and were later exposed during perihelion and outbound activity? Second, could directed panspermia occur if a technological civilization planted life-bearing material inside or onto an icy ISO so that it later transported life through the Milky Way? We combine data on 3I/ATLAS with order-of-magnitude thermal, biological, and mission constraints. SPHEREx provides the volatile and organic context through CO$_2$, H$_2$O, CO, dust, and a broad C--H feature, while JWST/MIRI provides the first direct CH$_4$ detection in an interstellar object and confirms an unusual volatile inventory, including enhanced CO$_2$:H$_2$O and CH$_4$:H$_2$O ratios. We distinguish dormant interstellar cruise from active perihelion. Natural panspermia is plausible as microbes can survive or repair damage in ice films, veins, or frozen matrices at very low metabolic rates. Methane production is more nuanced. Frozen survival metabolism would require $\sim10^{14}$--$10^{15}$ kg of biomass to match the JWST CH$_4$ rates, but active methanogenic archaea in warm, liquid, substrate-rich settings can produce methane many orders of magnitude faster, reducing the required biomass in optimistic laboratory-rate comparisons. Directed panspermia faces a different challenge: a direct 60 km s$^{-1}$ impact releases $1.8\times10^9$ J kg$^{-1}$, hundreds of times the specific energy of TNT, and would destroy a biological capsule. 3I/ATLAS-like objects are therefore best treated as test cases for panspermia diagnostics rather than as evidence for life. Natural panspermia requires preservation plus a credible liquid-water or near-surface activation pathway.

Finding the first (Population III or Pop III) stars is one of the fundamental quests of astronomy, aiming to deliver the missing link in how stars form at early cosmic times. Yet their initial mass function, formation sites and feedback remain highly uncertain, as well as the timing and topology of the transition to metal-enriched star formation. The observability of their peculiar spectral features is also debated, due to their short lifetime and faintness. This review summarizes current theoretical expectations for Pop III star formation, and the main observational strategies that have been adopted to constrain their properties across cosmic time, including near-field cosmology studies, direct searches for extremely metal-poor star-forming complexes and/or hard-ionizing spectral signatures at high and intermediate redshifts, and prospects for identifying Pop III activity up to Cosmic Dawn. The combination of JWST spectroscopy, time-domain searches, lensing surveys, stellar archaeology, absorption-line studies, as well as improved simulations, is yielding a growing number of observational candidates and narrowing the allowed parameter space for the first stars, setting the stage for a ``golden era'' of Pop III searches.

The bright Ly$α$ line is regularly used to identify high-redshift star-forming galaxies known as Ly$α$ emitters (LAEs). However, Ly$α$ is affected by resonant scattering and dust absorption making interpretation of its brightness challenging without additional observations. We use SCUBA-2, PACS and SPIRE data to investigate the far-infrared emission, Ly$α$ escape fraction ($f{esc}$(Ly$α$)) and infrared excess (IRX=LIR/LUV) in $\sim$4000 LAEs at z=2.2-6 from SC4K. Five LAEs, all hosting AGN, are individually detected with fluxes $S_{850}$ = 3.7-5.5 mJy at 850$μ\mathrm{m}$. Stacking is used to probe the average emission from all individually undetected LAEs, though the stacks are undetected at all wavelengths (e.g. $S_{850}$ < 0.09 mJy; 3$σ$). We group the sample into bins of redshift, stellar mass, Ly$α$ luminosity and AGN status. Most subsets are undetected but LAEs containing AGN and that have high stellar masses ($M_{\star} = 10^{10} - 10^{12}\, M_{\odot}$; including and excluding AGN) are detected at most wavelengths, suggesting that stellar mass and AGN heating may be enhancing the dust visibility. Individually detected LAEs and detected stacks have $f{esc}$(Ly$α$)=1-7%, while all undetected stacks $\geq$ 10%. All LAEs together average over > 21% and display significant scatter, suggesting a clumpy ISM dust distribution. Non-zero $f{esc}$(Ly$α$) in massive and AGN-hosting LAEs suggests ionizing photons may escape even from dusty galaxies, challenging the idea that dusty galaxies are poor leakers. Examination of the IRX-$β_{UV}$ relation shows LAEs have higher IRX than typical star-forming galaxies at similar redshifts. However, our detections tend to favour more massive, AGN-hosting systems and deeper observations are therefore needed.

The identification of the sources and acceleration mechanisms of cosmic rays require precise measurements of their mass composition. Currently, the most reliable method is to measure the atmospheric depth at which cosmic ray air showers in our atmosphere reach their maximum (\Xmax). However, the hadronic interaction properties that govern the longitudinal development of air showers are not precisely known, which is a major source of systematic uncertainty on the mass composition. SKA-Low will observe cosmic rays in the 10$^{16}$ - 10$^{18}$ eV energy range with unprecedented resolution and bandwidth. This allows for a much more detailed reconstruction of the longitudinal shower evolution, which can be used to gain better understanding of the hadronic interactions, as well as the primary mass composition. After the first interaction of the cosmic ray with an atom in an air molecule, the secondary particles still carry a significant fraction of the total energy. When one of these particle travels very far before interacting again, it produces a sub-shower that can be recognized as a secondary bump in the longitudinal profile. Simulations have demonstrated that SKA-Low can resolve such double bump profiles by virtue of its high antenna density and broad bandwidth. In this chapter, we demonstrate how double-bump showers and other anomalous longitudinal developments can be used to constrain hadronic interaction properties, and to determine the mass composition of cosmic rays in the Galactic-to-extragalactic transition region.

The Schwarzschild orbit-superposition method is a highly flexible dynamical modelling tool. It constrains the mass distribution of a galaxy using line-of-sight velocity and photometric observations. However, constructing such a dynamical model of a galaxy is computationally expensive. We present SchwarMAX, a new publicly available GPU implementation of the Schwarzschild orbit-superposition method. The GPU-native code is significantly faster than other implementations, with entire model construction taking around a second on GPU A100. Using SchwarMAX, we can explore the distributions of both baryonic and dark matter in a galaxy across a high-dimensional parameter space. We demonstrate its performance using mock integrated-field spectroscopic unit data generated from an N-body simulated barred galaxy. We explore the 12-dimensional space of disc, bar and halo parameters using Markov Chain Monte Carlo. The density profiles and the bar pattern speed of the galaxy are recovered with good accuracy. We show that the code can be applied to barred galaxies across a wide range of inclination angles and can be easily extended to other stellar systems, such as elliptical and dwarf galaxies.

Gas motions provide insight into the dynamical history and physical processes within galaxy clusters. We investigate the kinematics of the ICM in the core of A496, a nearby, X-ray bright, strong cool-core cluster, using high-resolution data from the Resolve micro-calorimeter on board XRISM. We compared our measurement with other Resolve cluster core measurements and further compared our results with simulations and multiwavelength observations. From an optical redshift analysis, we found that the BCG is at rest with respect to the systemic velocity of the cluster. Despite multiple previously detected cold fronts and harboring a weak central radio source, Resolve observation shows that the core of A496 is dynamically quiescent. The ICM is moving with respect to the BCG with a LOS bulk velocity of $v_{\rm bulk}=-69_{-20}^{+25}\,\mathrm{km\,s}^{-1}$. We measured a turbulent velocity of $σ_{\rm v}=78_{-16}^{+18}\,\mathrm{km\,s}^{-1}$, the lowest value reported by the instrument on a cluster core to date. This value is in good agreement with the velocity dispersion of the H$α$ filament in the core, which may indicate condensation of ICM in the wake of the radio bubble. Assuming isotropic turbulence, the ICM turbulent velocity corresponds to a subsonic 3D Mach number of $0.15_{-0.03}^{+0.04}$ and a non-thermal pressure fraction of $1.2_{-0.5}^{+0.6}\,\%$. The mechanical AGN feedback from the recent activity of the central radio source is estimated to contribute about 7-9% to the ICM heating. The 1D LOS bulk velocity from the SLOW constrained Universe simulation is consistent with the measured value, suggesting that AGN feedback has a negligible contribution. The A496 SLOW turbulent velocity, as in other reported Resolve--simulation comparisons, is higher, but remains within $1.5σ$ uncertainty. A496 may represent one of the most quiescent sloshing cores observed so far.

We present the first JWST observations of a Calcium-Strong Transient (CaST), SN 2024uj, a rare class of supernovae (SNe) with observable properties that are consistent with both thermonuclear explosions of white dwarfs (WDs) and the core collapse of massive stars. SN 2024uj is offset by $\sim6.6$ kpc from its host and exhibits a double-peaked light curve consistent with shock cooling of nearby circumstellar material. At early times, its optical spectra resemble those of normal SNe Ib, but strong [Ca II] $λλ$7291, 7324 emission emerges between $+$2 and $+$17 days after maximum light. Radiative-transfer models of a massive stripped He star cannot reproduce this early forbidden Ca emission, even with artificially enhanced surface Ca, whereas it arises naturally in thermonuclear scenarios. The $+$150 d JWST/NIRSpec spectrum reveals highly asymmetric, multicomponent He I at both 1.083 and 2.058 $μ$m. The He extends to $\gtrsim+$5000 km/s, with a strong, narrow peak at $+$1500 km/s, indicating that He is distributed throughout the ejecta with a concentration offset from center. This He distribution overlaps central [Ca II] and [O I], implying a degree of mixing difficult to produce in a massive star explosion. The He peak might further trace interaction with a shocked, ejected companion in a thermonuclear system. The NIRSpec spectrum also shows molecular CO emission and a rising continuum that, together with a 10 $μ$m photometric detection, indicates dust emission extending into the mid-infrared. Given the remote environment, early forbidden Ca, mixed He/Ca/O ejecta, and possible companion signature, we favor a thermonuclear origin for SN 2024uj involving at least one low-mass, partially He-rich WD.

Large-scale ionised outflows and nuclear obscuration are fundamental manifestations of AGN activity, yet direct observational evidence linking these phenomena remains scarce. We use the eROSITA Final Equatorial Depth Survey, among the largest uniform optical spectroscopic datasets of X-ray AGN, to investigate how AGN accretion rate affects ionised outflow kinematics and X-ray obscuration. Our sample comprises 2.840 AGN at z<0.82 with high-quality SDSS spectra. Through optical spectral fitting, we measure Eddington ratios ($λ_{Edd}$) and [O III] emission-line kinematics, tracing ionised outflows. In addition, we use archival eROSITA X-ray spectroscopy with X-ray stacking analyses to constrain the obscuration of the sample, $N_H$. We find that (1) 35% of the entire sample hosts a [O III] outflows ($W_{80}>600$ km/s), with the outflow incidence increasing with the AGN luminosity from 15% at $L_{AGN}<10^{44}$ erg/s up to 60% at $L_{AGN}>10^{46}$ erg/s; (2) the outflow incidence increases with Eddington ratio from 29% at $\log λ_{Edd}<-2.3$ to 50% at $\log λ_{Edd}>-1.7$; and (3) the AGN obscuration decreases with Eddington ratio, as sources with $\logλ_{Edd}>-1.7$ are 5 times less obscured than lower Eddington ratios AGN. In addition, we find that 1% of the sample populates the "forbidden region" of the $N_H-λ_{Edd}$ plane, where the outflow incidence peaks at 52%, consistent with a short-lived feedback phase. Notably, when matching the Eddington ratios samples in AGN luminosity, these trends vanish, implying that radiation pressure drives changes in outflow activity and obscuration, while the black hole mass does not play a significant role. Our results are in agreement with AGN radiative feedback scenarios, where the Eddington ratio regulates the AGN environment by driving powerful, galaxy-wide outflows and shaping the amount of circumnuclear material.

We present a Gaia DR3 source-level census of \emph{Stellar Snake} complexes within 3\,kpc of the Sun. We define a Stellar Snake as a mutually coherent association of two or more stellar overdensities, characterised by consistent positions, kinematics, orbital invariants, ages, and chemical properties, rather than as a single gravitationally bound object. Moving beyond catalogue-driven searches seeded by known open clusters, our framework operates directly on individual Gaia sources to recover extended, low-density substructures and interconnecting stellar bridges. The multi-stage pipeline extracts statistically significant, non-overlapping base nodes, infers homogeneous parameters using a PointNet point-cloud regressor, and links these nodes into large-scale macro-structures across a 9D space spanning positions, tangential velocities, radial velocity, age \(\log t\), and orbital integrals \((E,L_Z)\). After FoF-topology cross-validation and boundary resolution, the final catalogue contains 1,256 Stellar Snake candidates comprising 802,489 unique member-star entries in 5,491 final base nodes selected from a 9,909-node input pool. Derived parameters are validated against external open-cluster catalogues and spectroscopic benchmarks. To quantify structural coherence, we introduce a graph-relation Snake Reliability Index (SRI), coupled with a peripheral-branch diagnostic and Gold/Silver/Bronze quality flags. At the population level, the census shows a broad age--metallicity pattern, a declining upper envelope of member-star entries toward older ages, and a projected association between young Snake nodes, nearby spiral-arm loci, and the Radcliffe Wave. This homogeneous inventory provides an observational foundation for probing the formation, coherence, and dynamical evolution of hierarchical stellar complexes in the Milky Way.

We report the low-resolution long-slit spectroscopic observations and confirmations of 11 quasar pair candidates, which are selected from the MGQPC catalog presented in the first paper of our series work (hereafter, Paper-I) and the early version of this catalog. The spectroscopic follow-up was carried out with 5 spectrographs equipped on 3 telescopes, and the major discoveries include 6 dual quasars and 3 projected quasars. One of the dual quasars has a high redshift of $\sim$ 3.1. The LQ hypothesis of 3 dual quasars cannot be completely ruled out. We investigated the reason why previous spectroscopic surveys missed several new quasars. We discussed a projected quasar with a wide-separation lensing configuration, as well as two quasar-star projections that mimic the configuration of lensed quasars. The photometric redshifts for the 11 observed candidates were extracted from the second paper of our series work (hereafter, Paper-II) to illustrate their positive role in mitigating contamination from projected quasars and quasar-star projections. We also reviewed and discussed the confirmation strategies for dual and lensed quasar candidates, and outlined future confirmation strategies for them in the context of the era dominated by large-scale spectroscopic and imaging surveys.

We investigate the multi-wavelength structure of recently quenched post-starburst (PSB) galaxies at 0.5 < z < 3, using photometrically selected samples from the Ultra Deep Survey (UDS). Leveraging deep eight-band JWST/NIRCam imaging from the PRIMER programme, we analyze ~120 PSBs across the rest-frame optical-to-near-infrared, and compare with a reference sample of ~3000 passive and star-forming galaxies. Structural parameters (effective radius Re and Sersic index n) are derived independently in each waveband, and reveal that PSBs exhibit minimal structural variation with wavelength, indicating negligible stellar population age gradients or internal dust obscuration. We confirm that PSBs follow the established redshift-mass trends: at z > 1, massive PSBs (M* > 10^10 Msun) are compact spheroids resembling massive passive galaxies, albeit significantly more compact, whereas at 0.5 < z < 1, PSBs are typically low-mass (M* < 10^10 Msun) compact, disc-dominated systems akin to low-mass passive discs. Furthermore, for the first time, we systematically quantify disturbance indicators (residual flux fraction RFF, asymmetry, residual asymmetry) across a large PSB sample. At all masses, PSBs exhibit low RFF and asymmetry values comparable to passive systems and consistent with smooth, largely undisturbed morphologies. However, at z > 1, massive PSBs (M* > 10^10.25 Msun) show enhanced residual asymmetry relative to the passive population, indicating a previously unrecognized level of structural disturbance masked beneath a smooth stellar distribution. These results suggest that, while structural transformation is largely complete by the PSB phase, residual disturbances persist at high redshift, supporting a scenario in which rapid quenching proceeds via two distinct pathways: highly disruptive events (e.g. major mergers) at high z and high mass, and comparatively gentle processes at later times.

We report the discovery of a sample of little red dots (LRDs) at $z \approx 2$ identified from deep JWST/NIRCam imaging and wide-field slitless spectroscopy over $140$ arcmin$^2$ from the EIGER survey. With an improved blind broad-line identification algorithm, we select 19 sources at spectroscopic redshifts $z = 1.55-3.18$ identified via rest-frame near-infrared lines (Paschen-$β$, HeI+Pa$γ$ and OI). Based on a range of spectro-photometric criteria, we classify five of these sources as LRDs and the other 14 as classical active galactic nuclei (AGNs). This classification is corroborated by some X-ray detections among the AGNs. Classical AGNs dominate the number counts above optical luminosities M$_{5100}<-22.5$, whereas the LRD fraction among broad-line sources reaches 100 % at M$_{5100}\approx-20$. The LRDs span the range in Balmer break strengths seen in the higher redshift populations. Blue-shifted HeI absorption is detected in the two reddest sources. The HeI/Pa$γ$ ratio cleanly separates LRDs from classical AGNs and seems to anti-correlate with Balmer break strength, likely tracing HeI self-absorption at higher gas column densities. Our LRD sample has a similar optical luminosity range as their high-redshift counterparts, corresponding to black hole masses of $\sim10^{6}$ M$_{\odot}$ at the Eddington luminosity. We measure LRD number densities of $\approx 7\times10^{-6}$ cMpc$^{-3}$ at $z = 1.9-2.5$, which indicates that LRDs represent $\lesssim 3$ % of the AGN population at these epochs. Our results confirm the previously reported decline in the LRD number density with respect to $z \approx 5$ based on photometric surveys, although we find the decline to be more gentle than earlier emphasized.

Early-time light curves of Type Ia Supernovae (SNe Ia) encode critical information about their progenitor systems. We characterize the rise of normal SNe Ia using a volume-complete sample of 972 events from the Zwicky Transient Facility Data Release 2, an order of magnitude larger than any previous dataset for similar analyses. Fitting light curves up to $30\%$ of peak flux with a power-law model under a hierarchical Bayesian framework, we provide robust population-level constraints on the rise time ($t_\mathrm{rise}$; $μ=18.55\pm0.08$ days, $σ=1.42\pm0.07$ days), rise index ($α$; $μ=2.10\pm0.04$, $σ=0.48\pm0.03$ in ZTF $r$), and $g-r$ color evolution ($α_g - α_r$; $μ=0.20\pm0.02$, $σ=0.17\pm0.02$). These power-law fits are sensitive to the chosen truncation epoch if data beyond $\sim$$40\%$ of peak flux are included, but generally converge when restricted to earlier epochs. The relation between rise morphology and light-curve width ($\texttt{SALT2}$ $x_1$ stretch) bifurcates into two distinct regimes: high-stretch SNe Ia show clear trends where a higher $x_1$ correlates with shallower rises and more persistent blue colors, whereas low-stretch SNe Ia lack such trends. While rise times correlate positively with $x_1$ overall, this relation flattens significantly within the high-stretch population. Searching for anomalies, we identify several normal SNe Ia with unusually long rise times, which potentially exhibit short-duration ($\lesssim$2 days) flux excesses over a smooth rise. Long-duration ($\sim$5 days) flux excesses appear common within the high-stretch population and are tied to the shallow rises and early blue colors, pointing to widespread outward $^{56}$Ni mixing. Multi-dimensional explosion models with more realistic progenitor setups are needed to fully reproduce the observed dichotomy in rise morphology and stretch.

Stars can radially redistribute (migrate) within galactic disks. The degree to which this occurs as dynamically `cold' (preserves orbital eccentricity) or `hot' (increases eccentricity) remains debated. Many models presume that radial redistribution occurs primarily via cold torquing, resulting in changes in angular momentum without dynamical heating. We test the net dynamical heating associated with redistribution over stellar lifetimes using the FIRE cosmological zoom-in simulations of 12 Milky Way-mass galaxies. We select star particles today that underwent significant changes in orbital angular momentum, j_phi, since birth. We investigate net changes in their orbital eccentricity, e, and we quantify the `cold-torqued' fraction of star particles with |Delta j_phi/j_phi,birth| > 0.2 that preserved eccentricity (|Delta e| < 0.1) since birth. The direction of radial redistribution is most critical: outward-migrating stars experienced smaller net changes in eccentricity, whereas inward-migrating stars almost always heat since birth. For stars born on near-circular orbits (e_birth < 0.2), the cold-torqued fraction decreases rapidly with age today and is generally < 50% at ages >~2 Gyr. Stars born on moderately eccentric orbits (e_birth ~ 0.4) are the most likely to preserve their birth eccentricity. However, the cold-torqued fraction is higher in earlier-forming and/or dynamically-colder disks. Significantly, we identify a population of stars that dynamically `cooled', decreasing in eccentricity since birth: this is the primary way that stars end up on near-circular orbits today. Overall, a star's migration direction, its e_birth, and its age primarily determine whether it was dynamically heated, cooled, or unchanged. In general, radial redistribution in FIRE is typically not cold between birth and today.

In the hot and dilute intracluster medium of galaxy clusters, large-scale buoyancy instabilities can develop due to the transport of heat along magnetic field lines. In particular, the peripheries of galaxy clusters are unstable to the magneto-thermal instability (MTI), which may contribute to the observed levels of turbulence. Recent theoretical and numerical work has revealed that the stable background entropy stratification controls the nonlinear saturation of the instability, by setting the strength and the integral scale of the resulting turbulent state. However, observations of the periphery of galaxy clusters show that the radial entropy profiles near the virial radii $R_{500}$ may be flatter than predicted by models of smooth gravitational accretion. This motivates us to investigate the saturation of the MTI in adiabatic (buoyantly neutral) atmospheres, using both phenomenological approaches and Boussinesq numerical simulations, carried out with the pseudospectral code SNOOPY. We find that the adiabatic MTI saturates in a state characterised by the formation of large-scale plumes and their destruction by shear instability, yielding a new scaling law for the saturated turbulent kinetic energy, $\sim$$χω_T$, as the adiabatic limit is approached, where $χ$ is the effective thermal diffusivity and $ω_T$ is the MTI frequency. This predicts that the MTI plumes may achieve near sonic speeds in cluster outskirts, thus providing significant turbulent pressure support, even in the face of suppressed thermal conduction.

Light curves of Type Ia Supernovae (SNe Ia) in the days following explosion encode the diversity of progenitor systems and explosion physics. We present a hierarchical Bayesian framework to robustly constrain the population-level light-curve morphology of SNe Ia by fitting a large light-curve dataset simultaneously to power-law rises. Using a multivariate Gaussian population prior, this framework automatically down-weights sparsely sampled SNe and noisy measurements in the inference, obviating the need for restrictive quality cuts that introduce selection biases. Validation on simulated power-law light curves demonstrates that the population prior effectively suppresses the volume-projection bias from the asymmetric likelihood: compared to the classic two-step approach of fitting individual SNe and then aggregating the results, the hierarchical approach dramatically reduces the bias on the population-level parameters (mean, scatter, and correlation). When fitting the power-law model to light curves with more realistic morphologies, while the rise time can be mildly underestimated due to model misspecification, the recovered population scatter remains reliable. Furthermore, SNe with early flux excesses can emerge as outliers in the inferred parameter space, offering a potential diagnostic for identifying such events. Finally, we show that the inferred population distribution can also improve individual-event inference. Restricting the population prior to nuisance amplitudes, while preserving the complete correlation structure, regularizes fits to individual SNe without shrinking the physically meaningful rise time and rise index toward their population means.

We report evidence for millihertz (mHz) QPOs in the bright atoll neutron star low-mass X-ray binary (NS LMXB) source GX 3+1 using NICER in the 0.5--10 keV energy band. Across 7 observational datasets obtained over 6 days, we made 8 candidate mHz QPO detections with local significance above 95%, one of which remains above 95% global significance after the trial correction. These mHz QPOs were detected in the 7--15 mHz frequency range and had fractional rms amplitudes ranging from 0.51--1.41%. Our studies indicate no association between the hardness intensity diagram location of the candidate QPOs and their rms amplitudes. There appears to be a monotonous increase in rms with energy, except in some observations where there is a pivot around 3--4 keV or at ~ 5 keV. Previous studies of mHz QPOs in other sources in literature indicate a pivot around 3 keV in the rms-energy relation, but in our study some tentative detections suggest a pivot at around ~ 5 keV, though the large uncertainties in most cases prevent a robust statistical claim. Although properties such as the frequency range of detection and fractional rms amplitudes of the mHz QPOs in our study are well in agreement with that in literature for other NS LMXBs, the luminosity at which these candidate QPOs occur are higher than that of other sources. The rms-energy relation of the candidate mHz QPOs and the luminosity at which they occur challenges some of the existing considerations of mHz QPO origin.

AGN feedback driven by radio jets plays a key role in regulating the cold ISM of galaxies. Neutral hydrogen traced through the H1 21-cm line provides a powerful probe of the kinematics, distribution, and physical conditions of cold gas in the central regions of AGN. Previous observations have detected H1 column densities down to $\sim10^{20}$ cm$^{-2}$, but typically at arcsecond-scale resolution, inhibiting the characterization of small-scale H1 gas clouds and their connection to molecular gas reservoirs and sites of jet-ISM interaction. High-resolution H1 imaging is therefore required to determine whether the atomic gas is associated with circumnuclear structures, jet-driven outflows, compressed gas layers, or fragmented cold clouds embedded within a disturbed multiphase ISM. In this chapter, we focus on molecular gas-rich radio AGN hosting large-scale jets and exhibiting strong interactions between the radio plasma and the surrounding ISM, where atomic, molecular, and ionized gas coexist. These systems provide ideal laboratories for investigating the spatial distribution and kinematics of H1, constraining the impact of radio jets on the cold gas, and determining whether the gas is associated with inflowing, outflowing, or otherwise disturbed components. SKA-VLBI will deliver milliarcsecond-scale imaging and $μ$Jy-level sensitivity, resolving H1 gas clouds on parsec scales across extended radio structures. This capability will enable detailed constraints on the location, morphology, and kinematics of atomic gas from the nuclear regions to kiloparsec-scale jet-ISM interaction sites. Combined with molecular gas observations at other wavelengths, these measurements will provide a comprehensive view of the jet-ISM interactions, the impact of AGN-driven feedback, and the role of cold gas in galaxy evolution.

We present a novel mapmaking method for obtaining unbiased estimates of CMB polarisation, tailored to modern CMB experiments with a rotating half-wave plate. These experiments are exposed to strong unpolarised contaminant sources, such as atmospheric emission and ground pickup, which can be several orders of magnitude stronger than the sky signal. Our mapmaker mitigates these systematic effects by marginalising over all signals that vary slowly compared to the timescale of a polarimeter's angle rotation on the sky, while recovering high-fidelity polarisation maps. When the variability timescales of the unpolarised signals exceed a quarter of the half-wave plate rotation period, the method can produce maps with nearly optimal noise levels and minimal contamination. Furthermore, if the half-wave plate rotation period is sufficiently short relative to the beam-scale crossing time, the method efficiently mitigates the sky intensity-to-polarisation leakage. This mapmaker, named the Polarisation-Optimised Map-Making Estimator (POMME), is implemented within the open-source FURAX package and is ready for application to upcoming ground-based CMB surveys.

We present a comprehensive chemo-dynamical analysis of eight open clusters selected from the UBC catalog using Gaia DR3 data. These clusters are located at heliocentric distances of ~2-5 kpc, probing regions of the Galactic disk beyond the solar neighborhood. Cluster membership is determined using the UPMASK algorithm, while structural parameters are derived from radial density profiles through King model fitting combined with MCMC sampling. Their structural parameters reveal diverse internal configurations, from diffuse to moderately concentrated systems. Fundamental astrophysical parameters (extinction, distance, metallicity, and age) are obtained via Bayesian isochrone fitting based on PARSEC models. The clusters span a wide age range (~20 Myr to ~5 Gyr) and show a broad metallicity distribution (-0.34 <= [Fe/H] (dex) <= +0.25). Orbital analysis based on backward integrations shows that all clusters follow nearly circular orbits (e ~ 0.03-0.09) with low vertical distances from the Galactic plane (Zmax < 0.4 kpc), confirming their membership in the Galactic thin disk and dynamically cold kinematics. Comparison between inferred traceback orbital radii and present-day guiding radii indicates modest radial displacements, with Delta R < 0.5 kpc for the UBC sample. These offsets are consistent with mild radial redistribution expected for young, dynamically cold open clusters rather than strong radial migration. The results suggest that radial migration should be considered when interpreting the present-day spatial and chemical distribution of these clusters, although the inferred migration amplitudes remain moderate. Our results demonstrate that relatively distant open clusters can be characterized with high precision using Gaia DR3 data and that mild radial redistribution should be considered when interpreting the present-day distribution of stellar populations in the Galactic disk.

Transient astrophysics provides a set of unique laboratories for studying fundamental physics. From the launching of powerful relativistic jets to merging neutron stars, highly-magnetised compact objects, or stellar explosions, transients probe the Universe at its most extreme. The SKAO will provide an unrivalled set of capabilities for transient observations on timescales from nanoseconds to decades, opening new discovery space. With its sensitivity, broad spectral coverage, wide field of view, and high survey speed, SKAO will allow us to discover and understand rare events that provide powerful new insights into regimes of high energy density, strong gravity, and intense magnetic fields. Complemented by a suite of multi-wavelength and multi-messenger facilities, and supported by a network of smaller existing radio telescopes and new computational capabilities, SKAO will unveil the most powerful and exotic events in our Universe, addressing some of the key questions in modern astrophysics and cosmology.

A highly star-forming galaxy at z ~ 2 hosting an obscured, luminous active galactic nucleus (AGN) and a relativistic radio jet sets the stage for a cosmic crime scene. The victim is star formation, the suspect is AGN feedback. These systems offer a rare opportunity to catch this process in the act. We propose SHARP@ELT integral-field spectroscopic observations of heavily obscured, luminous AGN with resolved radio emission to witness the onset of feedback and its impact on the host interstellar medium (ISM). Such objects trace a short-lived (< 10^5 yr) evolutionary phase in which a recent starburst, newly triggered radio jets, and a deeply embedded AGN coexist. In this phase, feedback is expected to suppress star formation while clearing the dusty nuclear regions. Using SHARP/VESPER, we will derive spatially resolved maps of stellar ages, star formation rate density, gas density, and ionization state, alongside the kinematics of stars and ionized gas. By directly comparing these properties with the radio structures, we will quantify the effects of both jet-driven and radiative feedback on the ISM. Our sample consists of eleven luminous, obscured AGN at 1.5 < z < 2.5 with resolved radio emission on scales of ~1-15 kpc. Exploiting the VESPER multi-Integral Field Selector capability, we will obtain resolved continuum and emission-line maps for at least 110 galaxies at cosmic noon, enabling a comprehensive characterization of their environment, multiphase ISM, and nuclear activity within 55 h of integration time. SHARP will thus reveal AGN feedback at the epoch when it is most effective, providing a decisive step toward understanding its role in galaxy evolution.

The environment plays a key role in shaping how galaxies form and evolve. Galaxies in dense nodes of the cosmic web are thought to grow and quench earlier, and faster and become more massive than those in the field. To understand the physical drivers of this environmental effect, we must probe the most crowded regions of the Universe at the epoch when growth was at its peak and the transition to quiescence was triggered, around 10 billion years ago (z ~ 2). This period saw the downturn of the cosmic star-formation and black-hole accretion histories, the quenching and morphological transformation of massive galaxies, and the virialisation of the first clusters. Several processes might be at play: stellar and AGN feedback, reduced gas accretion, disk instabilities, morphological quenching, interactions, and ram-pressure stripping. The ELT/SHARP instrument, with its sensitivity, spectral resolution, wavelength coverage, and multiplexing capabilities over a wide field, is ideally suited to study these mechanisms by targeting multiple members of dense structures simultaneously. Cluster and protocluster cores at z ~ 2 span roughly 1 arcmin and host about ten massive (Mstar > 10^10.5 Msun) galaxies. VESPER can deliver spatially resolved gas and stellar kinematics, map recent and past star formation, identify companions, inflows, outflows, shocks, and AGN activity for the most massive core members. With 80 hr of VESPER time, we can obtain this type of data for about 60 galaxies selected from the densest regions of five clusters at 1.5 < z < 1.7 and five protoclusters at 2 < z < 2.5 spanning the evolutionary phases of maximal growth and rapid decline. Such a sample would permit to trace the evolution from protoclusters to virialised clusters and identify the environmental processes responsible for their rapid transformations.

The modulation of galactic cosmic rays, driven by the evolution of the heliospheric magnetic field, strongly influences the intensity of cosmic rays reaching near-Earth space. Characterizing this process is crucial both for advancing our understanding of cosmic-ray transport and for assessing radiation exposure and related hazards in space environments. Here we present a newly developed forecasting framework built on a numerical description of charged particle transport in the heliosphere and its dependence on solar activity, designed for the long-term forecasting of galactic cosmic-ray fluxes. It solves a one-dimensional, spherically symmetric form of the Parker transport equation, including diffusion, solar-wind advection, and adiabatic energy losses. The model has been validated using multi-species flux measurements from space-based experiments: PAMELA, AMS-02, and ACE. Its strategy is based on Hilbert-Huang transform filtering and cross-correlation between delayed solar proxies and effective model parameters. Our charge-sign- and rigidity-dependent parametric description of the diffusion-advection processes yields good overall agreement with the data, as shown by the reconstruction uncertainty. The robustness of this approach is validated across a broad set of multichannel datasets covering different particle species, energy ranges, and phases of solar activity, supporting its applicability to space radiation monitoring and forecasting. Furthermore, when coupled with solar-proxy forecasting models, it enables decadal-scale predictions of galactic cosmic-ray fluxes, thereby supporting long-term planning and radiation-risk assessment for future space missions.

Discovery of dark-matter (DM) free dwarf galaxies in the NGC 1052 neighborhood has had a considerable impact on modern cosmology. They have been explained through a dwarf--dwarf head-on collision that is a rare event. We find that they could alternatively be associated with a head-on, 1:1 merger after it has been tuned to generate the E4 morphology of NGC 1052. Our simulations show that such mergers produce long-lived tidal features, associated with the remnant galaxy, and in the form of large tidal tails including tidal dwarf galaxies (TDGs). We underline that such tidal features are predicted by the hierarchical scenario in which massive galaxies are formed by galaxy mergers. The latter can reproduce both the tidal features in the NGC1052 outskirts and the observed dwarf galaxies. The simulated TDGs have similar sizes to those observed, while they are ten times smaller in the dwarf bullet scenario. However, we cannot reproduce the luminous globular cluster systems due to resolution limitations. Resolving the radial distance between the DM-free dwarfs is necessary to identify the scenario of their formation. We suggest that there should be many other examples of DM-free dwarf galaxies in the neighborhood of local massive galaxies and galaxy groups.

Produced by the interaction between the ``pulsar wind'' powered by the rotational energy of a neutron star and its surroundings, the study of pulsar wind nebulae (PWNe) provides vital insight into the physics of neutron star magnetospheres and ultra-relativistic outflows. Spatially-resolved studies of the continuum and polarized radio emission of these sources are vital for understanding the production of $e^\pm$ in the magnetospheres of neutron stars, the acceleration of these particles (and potentially baryons) to $\gtrsim10^{15}~{\rm eV}$ energies, and their propagation within the PWN and in the surrounding interstellar medium. The significant improvements in sensitivity, dynamic range, timing capabilities offered by the Square Kilometer Array have the potential to greatly improve our understanding of the origin of some of the highest energy particles produced in the Milky Way.

Galaxy outflows are usually ordered by stellar mass and star-formation rate (SFR), but the same feedback budget may couple differently to gas in diffuse and compact galaxies. We use Dark Energy Spectroscopic Instrument (DESI) Data Release 1 stacked spectra of massive star-forming galaxies at $0.35<z<1.0$ to test whether stellar surface density, $Σ_\star=M_\star/(2πR_e^2)$, is an independent empirical coordinate of down-the-barrel singly ionized magnesium (MgII) cool-gas absorption. In AGN-clean samples matched in stellar mass, and in a stricter sample matched in both stellar mass and a Balmer-line SFR proxy, the MgII outflow equivalent width (EW) rises monotonically with $Σ_\star$ in every redshift bin. From the lowest to highest $Σ_\star$ tertile, EW increases by 0.37-0.61 Angstrom, while the absolute outflow velocity changes only weakly. DESI therefore shows that cool-gas outflow strength in massive star-forming galaxies is not set only by how much stellar mass or star formation a galaxy has, but also by how tightly the galaxy is built. The structural dependence points to changes in the absorbing velocity distribution and/or the effective covering fraction of cool outflowing gas.