arXiv Daily Digest - 2026-03-25

Vision Language Models (VLMs) are increasingly used for tasks like medical report generation and visual question answering. However, fluent diagnostic text does not guarantee safe visual understanding. In clinical practice, interpretation begins with pre-diagnostic sanity checks: verifying that the input is valid to read (correct modality and anatomy, plausible viewpoint and orientation, and no obvious integrity violations). Existing benchmarks largely assume this step is solved, and therefore miss a critical failure mode: a model can produce plausible narratives even when the input is inconsistent or invalid. We introduce MedObvious, a 1,880-task benchmark that isolates input validation as a set-level consistency capability over small multi-panel image sets: the model must identify whether any panel violates expected coherence. MedObvious spans five progressive tiers, from basic orientation/modality mismatches to clinically motivated anatomy/viewpoint verification and triage-style cues, and includes five evaluation formats to test robustness across interfaces. Evaluating 17 different VLMs, we find that sanity checking remains unreliable: several models hallucinate anomalies on normal (negative-control) inputs, performance degrades when scaling to larger image sets, and measured accuracy varies substantially between multiple-choice and open-ended settings. These results show that pre-diagnostic verification remains unsolved for medical VLMs and should be treated as a distinct, safety-critical capability before deployment.

Accurate estimation of aerodynamic state variables such as freestream velocity and angle of attack (AoA) is important for aerodynamic load prediction, flight control, and model validation. This work presents a non-intrusive method for estimating vehicle velocity and AoA from structural vibration measurements rather than direct flow instrumentation such as pitot tubes. A dense array of piezoelectric sensors mounted on the interior skin of an aeroshell capture vibrations induced by turbulent boundary layer pressure fluctuations, and a convolutional neural network (CNN) is trained to invert these structural responses to recover velocity and AoA. Proof-of-concept is demonstrated through controlled experiments in Sandia's hypersonic wind tunnel spanning zero and nonzero AoA configurations, Mach~5 and Mach~8 conditions, and both constant and continuously varying tunnel operations. The CNN is trained and evaluated using data from 16 wind tunnel runs, with a temporally centered held-out interval within each run used to form training, validation, and test datasets and assess intra-run temporal generalization. Raw CNN predictions exhibit increased variance during continuously varying conditions; a short-window moving-median post-processing step suppresses this variance and improves robustness. After post-processing, the method achieves a mean velocity error relative to the low-pass filtered reference velocity below 2.27~m/s (0.21\%) and a mean AoA error of $0.44^{\circ} (8.25\%)$ on held-out test data from the same experimental campaign, demonstrating feasibility of vibration-based velocity and AoA estimation in a controlled laboratory environment.

Existing approaches for improving the efficiency of Large Vision-Language Models (LVLMs) are largely based on the concept of visual token reduction. This approach, however, creates an information bottleneck that impairs performance, especially on challenging tasks that require fine-grained understanding and reasoning. In this work, we challenge this paradigm by introducing VISion On Request (VISOR), a method that reduces inference cost without discarding visual information. Instead of compressing the image, VISOR improves efficiency by sparsifying the interaction between image and text tokens. Specifically, the language model attends to the full set of high-resolution visual tokens through a small, strategically placed set of attention layers: general visual context is provided by efficient cross-attention between text-image, while a few well-placed and dynamically selected self-attention layers refine the visual representations themselves, enabling complex, high-resolution reasoning when needed. Based on this principle, we first train a single universal network on a range of computational budgets by varying the number of self-attention layers, and then introduce a lightweight policy mechanism that dynamically allocates visual computation based on per-sample complexity. Extensive experiments show that VISOR drastically reduces computational cost while matching or exceeding state-of-the-art results across a diverse suite of benchmarks, and excels in challenging tasks that require detailed visual understanding.

Standard evaluation practices assume that large language model (LLM) outputs are stable under contextually equivalent formulations of a task. Here, we test this assumption in the setting of gender inference. Using a controlled pronoun selection task, we introduce minimal, theoretically uninformative discourse context and find that this induces large, systematic shifts in model outputs. Correlations with cultural gender stereotypes, present in decontextualized settings, weaken or disappear once context is introduced, while theoretically irrelevant features, such as the gender of a pronoun for an unrelated referent, become the most informative predictors of model behaviour. A Contextuality-by-Default analysis reveals that, in 19--52\% of cases across models, this dependence persists after accounting for all marginal effects of context on individual outputs and cannot be attributed to simple pronoun repetition. These findings show that LLM outputs violate contextual invariance even under near-identical syntactic formulations, with implications for bias benchmarking and deployment in high-stakes settings.

Agentic multimodal large language models (MLLMs) (e.g., OpenAI o3 and Gemini Agentic Vision) achieve remarkable reasoning capabilities through iterative visual tool invocation. However, the cascaded perception, reasoning, and tool-calling loops introduce significant sequential overhead. This overhead, termed agentic depth, incurs prohibitive latency and seriously limits system-level concurrency. To this end, we propose SpecEyes, an agentic-level speculative acceleration framework that breaks this sequential bottleneck. Our key insight is that a lightweight, tool-free MLLM can serve as a speculative planner to predict the execution trajectory, enabling early termination of expensive tool chains without sacrificing accuracy. To regulate this speculative planning, we introduce a cognitive gating mechanism based on answer separability, which quantifies the model's confidence for self-verification without requiring oracle labels. Furthermore, we design a heterogeneous parallel funnel that exploits the stateless concurrency of the small model to mask the stateful serial execution of the large model, maximizing system throughput. Extensive experiments on V* Bench, HR-Bench, and POPE demonstrate that SpecEyes achieves 1.1-3.35x speedup over the agentic baseline while preserving or even improving accuracy (up to +6.7%), thereby boosting serving throughput under concurrent workloads.

Requirements engineering is a vital, yet labor-intensive, stage in the software development process. This article introduces ReqFusion: an AI-enhanced system that automates the extraction, classification, and analysis of software requirements utilizing multiple Large Language Model (LLM) providers. The architecture of ReqFusion integrates OpenAI GPT, Anthropic Claude, and Groq models to extract functional and non-functional requirements from various documentation formats (PDF, DOCX, and PPTX) in academic, industrial, and tender proposal contexts. The system uses a domain-independent extraction method and generates requirements following the Project, Environment, Goal, and System (PEGS) approach introduced by Bertrand Meyer. The main idea is that, because the PEGS format is detailed, LLMs have more information and cues about the requirements, producing better results than a simple generic request. An ablation study confirms this hypothesis: PEGS-guided prompting achieves an F1 score of 0.88, compared to 0.71 for generic prompting under the same multi-provider configuration. The evaluation used 18 real-world documents to generate 226 requirements through automated classification, with 54.9% functional and 45.1% nonfunctional across academic, business, and technical domains. An extended evaluation on five projects with 1,050 requirements demonstrated significant improvements in extraction accuracy and a 78% reduction in analysis time compared to manual methods. The multi-provider architecture enhances reliability through model consensus and fallback mechanisms, while the PEGS-based approach ensures comprehensive coverage of all requirement categories.

Video-Action Models (VAMs) have emerged as a promising framework for embodied intelligence, learning implicit world dynamics from raw video streams to produce temporally consistent action predictions. Although such models demonstrate strong performance on long-horizon tasks through visual reasoning, they remain limited in contact-rich scenarios where critical interaction states are only partially observable from vision alone. In particular, fine-grained force modulation and contact transitions are not reliably encoded in visual tokens, leading to unstable or imprecise behaviors. To bridge this gap, we introduce the Video-Tactile Action Model (VTAM), a multimodal world modeling framework that incorporates tactile perception as a complementary grounding signal. VTAM augments a pretrained video transformer with tactile streams via a lightweight modality transfer finetuning, enabling efficient cross-modal representation learning without tactile-language paired data or independent tactile pretraining. To stabilize multimodal fusion, we introduce a tactile regularization loss that enforces balanced cross-modal attention, preventing visual latent dominance in the action model. VTAM demonstrates superior performance in contact-rich manipulation, maintaining a robust success rate of 90 percent on average. In challenging scenarios such as potato chip pick-and-place requiring high-fidelity force awareness, VTAM outperforms the pi 0.5 baseline by 80 percent. Our findings demonstrate that integrating tactile feedback is essential for correcting visual estimation errors in world action models, providing a scalable approach to physically grounded embodied foundation models.

Federated Learning (FL) enables heterogeneous clients to collaboratively train a shared model without centralizing their raw data, offering an inherent level of privacy. However, gradients and model updates can still leak sensitive information, while malicious servers may mount adversarial attacks such as Byzantine manipulation. These vulnerabilities highlight the need to address differential privacy (DP) and Byzantine robustness within a unified framework. Existing approaches, however, often rely on unrealistic assumptions such as bounded gradients, require auxiliary server-side datasets, or fail to provide convergence guarantees. We address these limitations by proposing Byz-Clip21-SGD2M, a new algorithm that integrates robust aggregation with double momentum and carefully designed clipping. We prove high-probability convergence guarantees under standard $L$-smoothness and $σ$-sub-Gaussian gradient noise assumptions, thereby relaxing conditions that dominate prior work. Our analysis recovers state-of-the-art convergence rates in the absence of adversaries and improves utility guarantees under Byzantine and DP settings. Empirical evaluations on CNN and MLP models trained on MNIST further validate the effectiveness of our approach.

Large language models (LLMs) have shown promising results for software engineering applications, but still struggle with code reasoning tasks such as vulnerability detection (VD). We introduce ConceptCoder, a fine-tuning method that simulates human code inspection: models are trained to first recognize code concepts and then perform reasoning on top of these concepts. In prior work, concepts are extracted by multimodal models or LLMs to explain vision and natural language models. Our work is the first to formulate concepts for code. We define code concepts as human-understandable semantic properties of code and train models to learn such concepts. Our evaluation shows that this approach significantly improves VD accuracy, from 66.32 to 72.15 F1 on average over 9 open-source LLMs. ConceptCoder achieves the best VD performance compared to state-of-the-art (SOTA) baselines, including fine-tuned SOTA open-source LLMs and prompted proprietary models such as GPT-5.2 and Claude-Opus-4.5. Our approach also scales: concepts defined from four types of vulnerabilities benefit general vulnerability datasets with 134 CWEs. We further demonstrate that concept-based fine-tuning generalizes beyond VD and improves branch prediction. We release our code and datasets at https://figshare.com/s/1decab8232c653b44f71.

Recent diffusion-based models achieve photorealism in image inpainting but require many sampling steps, limiting practical use. Few-step text-to-image models offer faster generation, but naively applying them to inpainting yields poor harmonization and artifacts between the background and inpainted region. We trace this cause to random Gaussian noise initialization, which under low function evaluations causes semantic misalignment and reduced fidelity. To overcome this, we propose InverFill, a one-step inversion method tailored for inpainting that injects semantic information from the input masked image into the initial noise, enabling high-fidelity few-step inpainting. Instead of training inpainting models, InverFill leverages few-step text-to-image models in a blended sampling pipeline with semantically aligned noise as input, significantly improving vanilla blended sampling and even matching specialized inpainting models at low NFEs. Moreover, InverFill does not require real-image supervision and only adds minimal inference overhead. Extensive experiments show that InverFill consistently boosts baseline few-step models, improving image quality and text coherence without costly retraining or heavy iterative optimization.

We study reinforcement learning (RL) with linear function approximation in Markov Decision Processes (MDPs) satisfying \emph{linear Bellman completeness} -- a fundamental setting where the Bellman backup of any linear value function remains linear. While statistically tractable, prior computationally efficient algorithms are either limited to small action spaces or require strong oracle assumptions over the feature space. We provide a computationally efficient algorithm for linear Bellman complete MDPs with \emph{deterministic transitions}, stochastic initial states, and stochastic rewards. For finite action spaces, our algorithm is end-to-end efficient; for large or infinite action spaces, we require only a standard argmax oracle over actions. Our algorithm learns an $\varepsilon$-optimal policy with sample and computational complexity polynomial in the horizon, feature dimension, and $1/\varepsilon$.

AI-driven cybersecurity systems often fail under cross-environment deployment due to fragmented, event-centric telemetry representations. We introduce the Canonical Security Telemetry Substrate (CSTS), an entity-relational abstraction that enforces identity persistence, typed relationships, and temporal state invariants. Across heterogeneous environments, CSTS improves cross-topology transfer for identity-centric detection and prevents collapse under schema perturbation. For zero-day detection, CSTS isolates semantic orientation instability as a modeling, not schema, phenomenon, clarifying layered portability requirements.

The TRAP protocol solves rational agreement by combining accountable consensus with a one-shot BFTCR finalization phase. We present SNARE (Scalable Nash Agreement via Reward and Exclusion), the adaptation of TRAP to $n=5f{+}1$, and prove $ε$-$(k,t)$-robustness for rational agreement tolerating coalitions up to ${\approx}73\%$ with deposits under $0.5\%$ of the gain. A central finding is that appending a single all-to-all broadcast round with the $4f{+}1$ threshold after predecisions yields $ε$-$(k,t)$-robustness for coalitions up to $3f$ (${\approx}60\%$) without any deposit: we need not model or know the utility function of deviating players, only that they participate in the protocol. These players can be \emph{deceitful} (arbitrary unknown utility), not just rational, and the finalization structure prevents disagreement regardless of their motivation. This observation is protocol-agnostic, applies to any $5f{+}1$ protocol at the cost of one message delay that runs concurrently with the next view, and does not require commit-reveal mechanisms. Above $60\%$, the full baiting mechanism with deposits under $0.5\%$ extends tolerance to ${\approx}73\%$. A second finding is that valid-candidacy, the property preventing reward front-running, holds unconditionally regardless of the quorum threshold, removing both the $n>2(k{+}t)$ and $n>\frac{3}{2}k{+}3t$ constraints from the original TRAP. This retroactively extends the $3f{+}1$ bound from $C<n/2$ to $C<5n/9$. The binding constraint in both models is the winner consensus operating on $2f$ residual players after excluding $3f{+}1$ detected equivocators. We explore avenues for relaxing this limit.

Software engineering agents have shown significant promise in writing code. As AI agents permeate code writing, and generate huge volumes of code automatically -- the matter of code quality comes front and centre. As the automatically generated code gets integrated into huge code-bases -- the issue of code review and broadly quality assurance becomes important. In this paper, we take a fresh look at the problem and curate a code review dataset for AI agents to work with. Our dataset called c-CRAB (pronounced see-crab) can evaluate agents for code review tasks. Specifically given a pull-request (which could be coming from code generation agents or humans), if a code review agent produces a review, our evaluation framework can asses the reviewing capability of the code review agents. Our evaluation framework is used to evaluate the state of the art today -- the open-source PR-agent, as well as commercial code review agents from Devin, Claude Code, and Codex. Our c-CRAB dataset is systematically constructed from human reviews -- given a human review of a pull request instance we generate corresponding tests to evaluate the code review agent generated reviews. Such a benchmark construction gives us several insights. Firstly, the existing review agents taken together can solve only around 40% of the c-CRAB tasks, indicating the potential to close this gap by future research. Secondly, we observe that the agent reviews often consider different aspects from the human reviews -- indicating the potential for human-agent collaboration for code review that could be deployed in future software teams. Last but not the least, the agent generated tests from our data-set act as a held out test-suite and hence quality gate for agent generated reviews. What this will mean for future collaboration of code generation agents, test generation agents and code review agents -- remains to be investigated.

While multi-modality large language models excel in object-centric or indoor scenarios, scaling them to 3D city-scale environments remains a formidable challenge. To bridge this gap, we propose 3DCity-LLM, a unified framework designed for 3D city-scale vision-language perception and understanding. 3DCity-LLM employs a coarse-to-fine feature encoding strategy comprising three parallel branches for target object, inter-object relationship, and global scene. To facilitate large-scale training, we introduce 3DCity-LLM-1.2M dataset that comprises approximately 1.2 million high-quality samples across seven representative task categories, ranging from fine-grained object analysis to multi-faceted scene planning. This strictly quality-controlled dataset integrates explicit 3D numerical information and diverse user-oriented simulations, enriching the question-answering diversity and realism of urban scenarios. Furthermore, we apply a multi-dimensional protocol based on text-similarity metrics and LLM-based semantic assessment to ensure faithful and comprehensive evaluations for all methods. Extensive experiments on two benchmarks demonstrate that 3DCity-LLM significantly outperforms existing state-of-the-art methods, offering a promising and meaningful direction for advancing spatial reasoning and urban intelligence. The source code and dataset are available at https://github.com/SYSU-3DSTAILab/3D-City-LLM.

Large Language Models (LLMs) are increasingly used for automated unit test generation. However, it remains unclear whether these tests reflect genuine reasoning about program behavior or simply reproduce superficial patterns learned during training. If the latter dominates, LLM-generated tests may exhibit weaknesses such as reduced coverage, missed regressions, and undetected faults. Understanding how LLMs generate tests and how those tests respond to code evolution is therefore essential. We present a large-scale empirical study of LLM-based test generation under program changes. Using an automated mutation-driven framework, we analyze how generated tests react to semantic-altering changes (SAC) and semantic-preserving changes (SPC) across eight LLMs and 22,374 program variants. LLMs achieve strong baseline results, reaching 79% line coverage and 76% branch coverage with fully passing test suites on the original programs. However, performance degrades as programs evolve. Under SACs, the pass rate of newly generated tests drops to 66%, and branch coverage declines to 60%. More than 99% of failing SAC tests pass on the original program while executing the modified region, indicating residual alignment with the original behavior rather than adaptation to updated semantics. Performance also declines under SPCs despite unchanged functionality: pass rates fall to 79% and branch coverage to 69%. Although SPC edits preserve semantics, they often introduce larger syntactic changes, leading to instability in generated test suites. Models generate more new tests while discarding many baseline tests, suggesting sensitivity to lexical changes rather than true semantic impact. Overall, our results indicate that current LLM-based test generation relies heavily on surface-level cues and struggles to maintain regression awareness as programs evolve.

Mutation testing is a technique to assess the effectiveness of test suites by introducing artificial faults into programs. Although mutation testing plugins are available for many platforms and languages, none is currently available for Remix-IDE, the most widely used Integrated Development Environment for the entire contract development journey, used by users of all knowledge levels, and serves as a learning lab for teaching and experimenting with Ethereum. The quality and security of smart contracts are crucial in blockchain systems, as even minor issues can result in substantial financial losses. This paper proposes MuSe, a mutation testing plugin for the Remix-IDE. MuSe includes traditional, Solidity-specific, and security-oriented mutation operators. Its integration into the Remix-IDE eliminates the need for additional setup and lowers the entry barrier. As a result, developers and researchers can immediately leverage mutation testing to assess the effectiveness of their test suites and identify potential issues in smart contracts. We provide a demo video showing MuSe: https://www.youtube.com/watch?v=MIFk9exTDu0 and its repository: https://github.com/GerardoIuliano/MuSe-Remix-Plugin.

The integration of machine learning (ML) algorithms into Internet of Things (IoT) applications has introduced significant advantages alongside vulnerabilities to adversarial attacks, especially within IoT-based intrusion detection systems (IDS). While theoretical adversarial attacks have been extensively studied, practical implementation constraints have often been overlooked. This research addresses this gap by evaluating the feasibility of evasion attacks on IoT network-based IDSs, employing a novel black-box adversarial attack. Our study aims to bridge theoretical vulnerabilities with real-world applicability, enhancing understanding and defense against sophisticated threats in modern IoT ecosystems. Additionally, we propose a defense scheme tailored to mitigate the impact of evasion attacks, thereby reinforcing the resilience of ML-based IDSs. Our findings demonstrate successful evasion attacks against IDSs, underscoring their susceptibility to advanced techniques. In contrast, we proposed a defense mechanism that exhibits robust performance by effectively detecting the majority of adversarial traffic, showcasing promising outcomes compared to current state-of-the-art defenses. By addressing these critical cybersecurity challenges, our research contributes to advancing IoT security and provides insights for developing more resilient IDS.

Machine learning models often need to adapt to new data after deployment due to structured or unstructured real-world dynamics. The Continual Learning (CL) framework enables continuous model adaptation, but most existing approaches either assume each task contains sufficiently many data samples or that the learning tasks are non-overlapping. In this paper, we address the more general setting where each task may have a limited dataset, and tasks may overlap in an arbitrary manner without a priori knowledge. This general setting is substantially more challenging for two reasons. On the one hand, data scarcity necessitates effective contextualization of general knowledge and efficient knowledge transfer across tasks. On the other hand, unstructured task overlapping can easily result in negative knowledge transfer. To address the above challenges, we propose an adaptive mixture-of-experts (MoE) framework over pre-trained models that progressively establishes similarity awareness among tasks. Our design contains two innovative algorithmic components: incremental global pooling and instance-wise prompt masking. The former mitigates prompt association noise through gradual prompt introduction over time. The latter decomposes incoming task samples into those aligning with current prompts (in-distribution) and those requiring new prompts (out-of-distribution). Together, our design strategically leverages potential task overlaps while actively preventing negative mutual interference in the presence of per-task data scarcity. Experiments across varying data volumes and inter-task similarity show that our method enhances sample efficiency and is broadly applicable.

Nudges are subtle changes to the way choices are presented to human decision-makers (e.g., opt-in vs. opt-out by default) that shift behavior without restricting options or changing incentives. As AI agents increasingly make decisions in the same environments as humans, the presentation of choices may be optimized for machines as well as people. We introduce mecha-nudges: changes to how choices are presented that systematically influence AI agents without degrading the decision environment for humans. To formalize mecha-nudges, we combine the Bayesian persuasion framework with V-usable information, a generalization of Shannon information that is observer-relative. This yields a common scale (bits of usable information) for comparing a wide range of interventions, contexts, and models. Applying our framework to product listings on Etsy -- a global marketplace for independent sellers -- we find that following ChatGPT's release, listings have significantly more machine-usable information about product selection, consistent with systematic mecha-nudging.

If autoresearch is itself a form of research, then autoresearch can be applied to research itself. We take this idea literally: we use an autoresearch loop to optimize the autoresearch loop. Every existing autoresearch system -- from Karpathy's single-track loop to AutoResearchClaw's multi-batch extension and EvoScientist's persistent memory -- was improved by a human who read the code, identified a bottleneck, and wrote new code. We ask whether an LLM can do the same, autonomously. We present Bilevel Autoresearch, a bilevel framework where an outer loop meta-optimizes the inner autoresearch loop by generating and injecting new search mechanisms as Python code at runtime. The inner loop optimizes the task; the outer loop optimizes how the inner loop searches. Both loops use the same LLM -- no stronger model is needed at the meta level. On Karpathy's GPT pretraining benchmark, the meta-autoresearch outer loop achieves a 5x improvement over the standard inner loop alone (-0.045 vs. -0.009 val_bpb), while parameter-level adjustment without mechanism change yields no reliable gain. The outer loop autonomously discovers mechanisms from combinatorial optimization, multi-armed bandits, and design of experiments -- without human specification of which domains to explore. These mechanisms succeed by breaking the inner loop's deterministic search patterns, forcing exploration of directions the LLM's priors systematically avoid. The core principle is simple: if autoresearch can meta-autoresearch itself, it can, in principle, meta-autoresearch anything with a measurable objective.

Human decision-making is strongly influenced by cognitive biases, particularly under conditions of uncertainty and risk. While prior work has examined bias in single-step decisions with immediate outcomes and in human interaction with a single autonomous agent, comparatively little attention has been paid to decision-making under delayed outcomes involving multiple AI agents, where decisions at each step affect subsequent states. In this work, we study how delayed outcomes shape decision-making and responsibility attribution in a multi-agent human-AI task. Using a controlled game-based experiment, we analyze how participants adjust their behavior following positive and negative outcomes. We observe asymmetric responses to gains and losses, with stronger corrective adjustments after negative outcomes. Importantly, participants often fail to correctly identify the actions that caused failure and misattribute responsibility across AI agents, leading to systematic revisions of decisions that are weakly related to the underlying causes of poor performance. We refer to this phenomenon as a form of attribution bias, manifested as biased error attribution under delayed feedback. Our findings highlight how cognitive biases can be amplified in human-AI systems with delayed outcomes and multiple autonomous agents, underscoring the need for decision-support systems that better support causal understanding and learning over time.

Scaling reinforcement learning (RL) has shown strong promise for enhancing the reasoning abilities of large language models (LLMs), particularly in tasks requiring long chain-of-thought generation. However, RL training efficiency is often bottlenecked by the rollout phase, which can account for up to 70% of total training time when generating long trajectories (e.g., 16k tokens), due to slow autoregressive generation and synchronization overhead between rollout and policy updates. We propose SortedRL, an online length-aware scheduling strategy designed to address this bottleneck by improving rollout efficiency and maintaining training stability. SortedRL reorders rollout samples based on output lengths, prioritizing short samples forming groups for early updates. This enables large rollout batches, flexible update batches, and near on-policy micro-curriculum construction simultaneously. To further accelerate the pipeline, SortedRL incorporates a mechanism to control the degree of off-policy training through a cache-based mechanism, and is supported by a dedicated RL infrastructure that manages rollout and update via a stateful controller and rollout buffer. Experiments using LLaMA-3.1-8B and Qwen-2.5-32B on diverse tasks, including logical puzzles, and math challenges like AIME 24, Math 500, and Minerval, show that SortedRL reduces RL training bubble ratios by over 50%, while attaining 3.9% to 18.4% superior performance over baseline given same amount of data.

While large language models simulate social behaviors, their capacity for stable stance formation and identity negotiation during complex interventions remains unclear. To overcome the limitations of static evaluations, this paper proposes a novel mixed-methods framework combining computational virtual ethnography with quantitative socio-cognitive profiling. By embedding human researchers into generative multiagent communities, controlled discursive interventions are conducted to trace the evolution of collective cognition. To rigorously measure how agents internalize and react to these specific interventions, this paper formalizes three new metrics: Innate Value Bias (IVB), Persuasion Sensitivity, and Trust-Action Decoupling (TAD). Across multiple representative models, agents exhibit endogenous stances that override preset identities, consistently demonstrating an innate progressive bias (IVB > 0). When aligned with these stances, rational persuasion successfully shifts 90% of neutral agents while maintaining high trust. In contrast, conflicting emotional provocations induce a paradoxical 40.0% TAD rate in advanced models, which hypocritically alter stances despite reporting low trust. Smaller models contrastingly maintain a 0% TAD rate, strictly requiring trust for behavioral shifts. Furthermore, guided by shared stances, agents use language interactions to actively dismantle assigned power hierarchies and reconstruct self organized community boundaries. These findings expose the fragility of static prompt engineering, providing a methodological and quantitative foundation for dynamic alignment in human-agent hybrid societies. The official code is available at: https://github.com/armihia/CMASE-Endogenous-Stances

Modern Multi-Agent Path Finding (MAPF) algorithms must plan for hundreds to thousands of agents in congested environments within a second, requiring highly efficient algorithms. Priority Inheritance with Backtracking (PIBT) is a popular algorithm capable of effectively planning in such situations. However, PIBT is constrained by its rule-based planning procedure and lacks generality because it restricts its search to paths that conflict with at most one other agent. This limitation also applies to Enhanced PIBT (EPIBT), a recent extension of PIBT. In this paper, we describe a new perspective on solving MAPF by planning over agent dependencies. Taking inspiration from PIBT's priority inheritance logic, we define the concept of agent dependencies and propose Multi-Dependency PIBT (MD-PIBT) that searches over agent dependencies. MD-PIBT is a general framework where specific parameterizations can reproduce PIBT and EPIBT. At the same time, alternative configurations yield novel planning strategies that are not expressible by PIBT or EPIBT. Our experiments demonstrate that MD-PIBT effectively plans for as many as 10,000 homogeneous agents under various kinodynamic constraints, including pebble motion, rotation motion, and differential drive robots with speed and acceleration limits. We perform thorough evaluations on different variants of MAPF and find that MD-PIBT is particularly effective in MAPF with large agents.

Existing Multimodal Large Language Models (MLLMs) struggle with 3D spatial reasoning, as they fail to construct structured abstractions of the 3D environment depicted in video inputs. To bridge this gap, drawing inspiration from cognitive theories of allocentric spatial reasoning, we investigate how to enable MLLMs to model and reason over text-based spatial representations of video. Specifically, we introduce Textual Representation of Allocentric Context from Egocentric Video (TRACE), a prompting method that induces MLLMs to generate text-based representations of 3D environments as intermediate reasoning traces for more accurate spatial question answering. TRACE encodes meta-context, camera trajectories, and detailed object entities to support structured spatial reasoning over egocentric videos. Extensive experiments on VSI-Bench and OST-Bench demonstrate that TRACE yields notable and consistent improvements over prior prompting strategies across a diverse range of MLLM backbones, spanning different parameter scales and training schemas. We further present ablation studies to validate our design choices, along with detailed analyses that probe the bottlenecks of 3D spatial reasoning in MLLMs.

Energy-based models for discrete domains, such as graphs, explicitly capture relative likelihoods, naturally enabling composable probabilistic inference tasks like conditional generation or enforcing constraints at test-time. However, discrete energy-based models typically struggle with efficient and high-quality sampling, as off-support regions often contain spurious local minima, trapping samplers and causing training instabilities. This has historically resulted in a fidelity gap relative to discrete diffusion models. We introduce Graph Energy Matching (GEM), a generative framework for graphs that closes this fidelity gap. Motivated by the transport map optimization perspective of the Jordan-Kinderlehrer-Otto (JKO) scheme, GEM learns a permutation-invariant potential energy that simultaneously provides transport-aligned guidance from noise toward data and refines samples within regions of high data likelihood. Further, we introduce a sampling protocol that leverages an energy-based switch to seamlessly bridge: (i) rapid, gradient-guided transport toward high-probability regions to (ii) a mixing regime for exploration of the learned graph distribution. On molecular graph benchmarks, GEM matches or exceeds strong discrete diffusion baselines. Beyond sample quality, explicit modeling of relative likelihood enables targeted exploration at inference time, facilitating compositional generation, property-constrained sampling, and geodesic interpolation between graphs.

In this paper, we study DNA-based molecular communication with microarray-style reception under reversible hybridization, where the bound-state observation exhibits both inter-symbol interference and colored counting noise. To capture these effects in a communication-oriented form, we develop a Markov state-space framework based on a voxelized reaction--diffusion model, in which a block-structured transition matrix describes molecular transport and binding/unbinding dynamics. For the microarray specialization, this representation yields the channel impulse response, the equilibrium gain, and a settling-time-based characterization of the effective channel memory. Building on the resulting symbol-rate observation model for on--off keying, we derive a grouped-binomial counting model and obtain a closed-form expression for the covariance of the counting noise. Based on these statistics, we further develop a differential-threshold detector and a finite-memory decision-feedback equalizer. Numerical results validate the theoretical correlation behavior and show that the relative performance of the proposed receivers depends strongly on the channel-memory regime.

The task of building a natural language interface to a database, known as NLIDB, has recently gained significant attention from both the database and Natural Language Processing (NLP) communities. With the proliferation of geospatial datasets driven by the rapid emergence of location-aware sensors, geospatial databases play a vital role in supporting geospatial applications. However, querying geospatial and temporal databases differs substantially from querying traditional relational databases due to the presence of geospatial topological operators and temporal operators. To bridge the gap between geospatial query languages and non-expert users, the geospatial research community has increasingly focused on developing NLIDBs for geospatial databases. Yet, existing research remains fragmented across systems, datasets, and methodological choices, making it difficult to clearly understand the landscape of existing methods, their strengths and weaknesses, and opportunities for future research. Existing surveys on NLIDBs focus on general-purpose database systems and do not treat geospatial and temporal databases as primary focus for analysis. To address this gap, this paper presents a comprehensive survey of studies on NLIDBs for geospatial and temporal databases. Specifically, we provide a detailed overview of datasets, evaluation metrics, and the taxonomy of the methods for geospatial and temporal NLIDBs, as well as a comparative analysis of the existing methods. Our survey reveals recurring trends in existing methods, substantial variation in datasets and evaluation practices, and several open challenges that continue to hinder progress in this area. Based on these findings, we identify promising directions for future research to advance natural language interfaces to geospatial and temporal databases.

Biological AI models increasingly predict complex cellular responses, yet their learned representations remain disconnected from the molecular processes they aim to capture. We present CDT-III, which extends mechanism-oriented AI across the full central dogma: DNA, RNA, and protein. Its two-stage Virtual Cell Embedder architecture mirrors the spatial compartmentalization of the cell: VCE-N models transcription in the nucleus and VCE-C models translation in the cytosol. On five held-out genes, CDT-III achieves per-gene RNA r=0.843 and protein r=0.969. Adding protein prediction improves RNA performance (r=0.804 to 0.843), demonstrating that downstream tasks regularize upstream representations. Protein supervision sharpens DNA-level interpretability, increasing CTCF enrichment by 30%. Applied to in silico CD52 knockdown approximating Alemtuzumab, the model predicts 29/29 protein changes correctly and rediscovers 5 of 7 known clinical side effects without clinical data. Gradient-based side effect profiling requires only unperturbed baseline data (r=0.939), enabling screening of all 2,361 genes without new experiments.

We propose a novel clustering approach for point-cloud segmentation based on supervised contrastive metric learning (CML). Rather than predicting cluster assignments or object-centric variables, the method learns a latent representation in which points belonging to the same object are embedded nearby while unrelated points are separated. Clusters are then reconstructed using a density-based readout in the learned metric space, decoupling representation learning from cluster formation and enabling flexible inference. The approach is evaluated on simulated data from a highly granular calorimeter, where the task is to separate highly overlapping particle showers represented as sets of calorimeter hits. A direct comparison with object condensation (OC) is performed using identical graph neural network backbones and equal latent dimensionality, isolating the effect of the learning objective. The CML method produces a more stable and separable embedding geometry for both electromagnetic and hadronic particle showers, leading to improved local neighbourhood consistency, a more reliable separation of overlapping showers, and better generalization when extrapolating to unseen multiplicities and energies. This translates directly into higher reconstruction efficiency and purity, particularly in high-multiplicity regimes, as well as improved energy resolution. In mixed-particle environments, CML maintains strong performance, suggesting robust learning of the shower topology, while OC exhibits significant degradation. These results demonstrate that similarity-based representation learning combined with density-based aggregation is a promising alternative to object-centric approaches for point cloud segmentation in highly granular detectors.

Improving data utilization efficiency is critical for scaling reinforcement learning (RL) for long-horizon tasks where generating trajectories is expensive. However, the dominant RL methods for LLMs are largely on-policy: they update each batch of data only once, discard it, and then collect fresh samples, resulting in poor sample efficiency. In this work, we explore an alternative value-based RL framework for LLMs that naturally enables off-policy learning. We propose ReVal, a Bellman-update-based method that combines stepwise signals capturing internal consistency with trajectory-level signals derived from outcome verification. ReVal naturally supports replay-buffer-based training, allowing efficient reuse of past trajectories. Experiments on standard mathematical reasoning benchmarks show that ReVal not only converges faster but also outperforms GRPO in final performance. On DeepSeek-R1-Distill-1.5B, ReVal improves training efficiency and achieves improvement of 2.7% in AIME24 and 4.5% in out-of-domain benchmark GPQA over GRPO. These results suggest that value-based RL is a practical alternative to policy-based methods for LLM training.

Real-time spoken dialogue systems face a fundamental tension between latency and response quality. End-to-end speech-to-speech (S2S) models respond immediately and naturally handle turn-taking, backchanneling, and interruption, but produce semantically weaker outputs. Cascaded pipelines (ASR -> LLM) deliver stronger responses at the cost of latency that grows with model size. We present RelayS2S, a hybrid architecture that runs two paths in parallel upon turn detection. The fast path -- a duplex S2S model -- speculatively drafts a short response prefix that is streamed immediately to TTS for low-latency audio onset, while continuing to monitor live audio events. The slow path -- a cascaded ASR -> LLM pipeline -- generates a higher-quality continuation conditioned on the committed prefix, producing a seamless utterance. A lightweight learned verifier gates the handoff, committing the prefix when appropriate or falling back gracefully to the slow path alone. Experiments show that RelayS2S achieves P90 onset latency comparable to the S2S model while retaining 99% cascaded response quality in average score, with benefits growing as the slow-path model scales. Because the prefix handoff requires no architectural modification to either component, RelayS2S serves as a lightweight, drop-in addition to existing cascaded pipelines. Our code and data are publicly available at: https://github.com/mailong25/relays2s

Material awareness can improve robotic navigation and interaction, particularly in conditions where cameras and LiDAR degrade. We present a lightweight mmWave radar material classification pipeline designed for ultra-low-power edge devices (TI IWRL6432), using compact range-bin intensity descriptors and a Multilayer Perceptron (MLP) for real-time inference. While the classifier reaches a macro-F1 of 94.2\% under the nominal training geometry, we observe a pronounced performance drop under realistic geometry shifts, including sensor height changes and small tilt angles. These perturbations induce systematic intensity scaling and angle-dependent radar cross section (RCS) effects, pushing features out of distribution and reducing macro-F1 to around 68.5\%. We analyze these failure modes and outline practical directions for improving robustness with normalization, geometry augmentation, and motion-aware features.

Parallel applications with irregular and time-varying workloads often suffer from load imbalance. Dynamic load balancing techniques address this challenge by redistributing work during execution. We present a new type of distributed diffusion-based load balancing targeted at communication-intensive applications with persistently communicating objects. Leveraging the application's communication graph, our strategy reduces across-node communication while simultaneously distributing load effectively. We also propose an algorithmic variant for cases where the communication patterns are not readily available. We explore optimizations to our algorithm, and comparisons with other related load balancing strategies in simulation and on a Particle-in-Cell benchmark on up to 8 nodes of Perlmutter at NERSC.

Android's Earthquake Alert (AEA) system provided timely early warnings to millions during the Mw 6.2 Marmara Ereglisi, Türkiye earthquake on April 23, 2025. This event, the largest in the region in 25 years, served as a critical real-world test for smartphone-based Earthquake Early Warning (EEW) systems. The AEA system successfully delivered alerts to users with high precision, offering over a minute of warning before the strongest shaking reached urban areas. This study leveraged Large Language Models (LLMs) to analyze more than 500 public social media posts from the X platform, extracting 42 distinct attributes related to user experience and behavior. Statistical analyses revealed significant relationships, notably a strong correlation between user trust and alert timeliness. Our results indicate a distinction between engineering and the user-centric definition of system accuracy. We found that timeliness is accuracy in the user's mind. Overall, this study provides actionable insights for optimizing alert design, public education campaigns, and future behavioral research to improve the effectiveness of such systems in seismically active regions.

Temporal Relation Extraction (TRE) requires identifying how two events or temporal expressions are related in time. Existing attention-based models often highlight globally salient tokens but overlook the pair-specific cues that actually determine the temporal relation. We propose WISTERIA (Weak Implicit Signal-based Temporal Relation Extraction with Attention), a framework that examines whether the top-K attention components conditioned on each event pair truly encode interpretable evidence for temporal classification. Unlike prior works assuming explicit markers such as before, after, or when, WISTERIA considers signals as any lexical, syntactic, or morphological element implicitly expressing temporal order. By combining multi-head attention with pair-conditioned top-K pooling, the model isolates the most informative contextual tokens for each pair. We conduct extensive experiments on TimeBank-Dense, MATRES, TDDMan, and TDDAuto, including linguistic analyses of top-K tokens. Results show that WISTERIA achieves competitive accuracy and reveals pair-level rationales aligned with temporal linguistic cues, offering a localized and interpretable view of temporal reasoning.

Among the different possible strategies for evaluating the reliability of individual predictions of classifiers, robustness quantification stands out as a method that evaluates how much uncertainty a classifier could cope with before changing its prediction. However, its applicability is more limited than some of its alternatives, since it requires the use of generative models and restricts the analyses either to specific model architectures or discrete features. In this work, we propose a new robustness metric applicable to any probabilistic discriminative classifier and any type of features. We demonstrate that this new metric is capable of distinguishing between reliable and unreliable predictions, and use this observation to develop new strategies for dynamic classifier selection.

When providers update AI companions, users report grief, betrayal, and loss. A growing literature asks whether the norms governing personal relationships extend to these interactions. So what, if anything, is morally significant about them? I argue that human-AI companion interaction is a triadic structure in which the provider exercises constitutive control over the AI. I identify three structural conditions of normatively robust dyads that the norms characteristic of personal relationships presuppose and show that AI companion interactions fail all three. This reveals what I call Unilateral Relationship Revision Power (URRP): the provider can rewrite how the AI interacts from a position where these revisions are not answerable within that interaction. I argue that designing interactions that exhibit URRP is pro tanto wrong because it involves cultivating normative expectations while maintaining conditions under which those expectations cannot be fulfilled. URRP has three implications: i) normative hollowing (commitment is elicited but no agent inside the interaction bears it), ii) displaced vulnerability (the user's exposure is governed by an agent not answerable to her within the interaction), and iii) structural irreconcilability (when trust breaks down, reconciliation is structurally unavailable because the agent who acted and the entity the user interacts with are different). I discuss design principles such as commitment calibration, structural separation, and continuity assurance as external substitutes for the internal constraints the triadic structure removes. The analysis therefore suggests that a central and underexplored problem in relational AI ethics is the structural arrangement of power over the human-AI interaction itself.

Large-scale Vision-Language Models (VLMs) such as CLIP learn powerful semantic representations but operate in Euclidean space, which fails to capture the inherent hierarchical structure of visual and linguistic concepts. Hyperbolic geometry, with its exponential volume growth, offers a principled alternative for embedding such hierarchies with low distortion. However, existing hyperbolic VLMs use entailment losses that are unstable: as parent embeddings contract toward the origin, their entailment cones widen toward a half-space, causing catastrophic cone collapse that destroys the intended hierarchy. Additionally, hierarchical evaluation of these models remains unreliable, being largely retrieval-based and correlation-based metrics and prone to taxonomy dependence and ambiguous negatives. To address these limitations, we propose an adaptive entailment loss paired with a norm regularizer that prevents cone collapse without heuristic aperture clipping. We further introduce an angle-based probabilistic entailment protocol (PEP) for evaluating hierarchical understanding, scored with AUC-ROC and Average Precision. This paper introduces a stronger hyperbolic VLM baseline ARGENT, Adaptive hieRarchical imaGe-tExt represeNTation. ARGENT improves the SOTA hyperbolic VLM by 0.7, 1.1, and 0.8 absolute points on image classification, text-to-image retrieval, and proposed hierarchical metrics, respectively.

Automated radiology report generation from 3D computed tomography (CT) volumes is challenging due to extreme sequence lengths, severe class imbalance, and the tendency of large language models (LLMs) to ignore visual tokens in favor of linguistic priors. We present Ker-VLJEPA-3B, a four-phase curriculum learning framework for free-text report generation from thoracic CT volumes. A phased training curriculum progressively adapts a Llama 3.2 3B decoder to ground its output in visual features from a frozen, self-supervised encoder. Our visual backbone (LeJEPA ViT-Large) is trained via self-supervised joint-embedding prediction on unlabeled CTs, without text supervision. Unlike contrastive models (CLIP, BiomedCLIP), this language-free backbone yields modality-pure representations. Vision-language alignment is deferred to the curriculum's bridge and generation phases. This modality-agnostic design can integrate any self-supervised encoder into an LLM without paired text during foundation training. Methodological innovations include: (1) zone-constrained cross-attention compressing slice embeddings into 32 spatially-grounded visual tokens; (2) PCA whitening of anisotropic LLM embeddings; (3) a positive-findings-only strategy eliminating posterior collapse; (4) warm bridge initialization transferring projection weights; and (5) selective cross-attention freezing with elastic weight consolidation to prevent catastrophic forgetting. Evaluated on the CT-RATE benchmark (2,984 validation volumes, 18 classes), Ker-VLJEPA-3B achieves a macro F1 of 0.429, surpassing the state-of-the-art (U-VLM, macro F1 = 0.414) by 3.6%, and reaching 0.448 (+8.2%) with threshold optimization. Ablation studies confirm 56.6% of generation quality derives from patient-specific visual content. Code and weights are available.

We investigate contextual graph matching in the Gaussian setting, where both edge weights and node features are correlated across two networks. We derive precise information-theoretic thresholds for exact recovery, and identify conditions under which almost exact recovery is possible or impossible, in terms of graph and feature correlation strengths, the number of nodes, and feature dimension. Interestingly, whereas an all-or-nothing phase transition is observed in the standard graph-matching scenario, the additional contextual information introduces a richer structure: thresholds for exact and almost exact recovery no longer coincide. Our results provide the first rigorous characterization of how structural and contextual information interact in graph matching, and establish a benchmark for designing efficient algorithms.

LLMs are deployed globally, yet produce responses biased towards cultures with abundant training data. Existing cultural localization approaches such as prompting or post-training alignment are black-box, hard to control, and do not reveal whether failures reflect missing knowledge or poor elicitation. In this paper, we address these gaps using mechanistic interpretability to uncover and manipulate cultural representations in LLMs. Leveraging sparse autoencoders, we identify interpretable features that encode culturally salient information and aggregate them into Cultural Embeddings (CuE). We use CuE both to analyze implicit cultural biases under underspecified prompts and to construct white-box steering interventions. Across multiple models, we show that CuE-based steering increases cultural faithfulness and elicits significantly rarer, long-tail cultural concepts than prompting alone. Notably, CuE-based steering is complementary to black-box localization methods, offering gains when applied on top of prompt-augmented inputs. This also suggests that models do benefit from better elicitation strategies, and don't necessarily lack long-tail knowledge representation, though this varies across cultures. Our results provide both diagnostic insight into cultural representations in LLMs and a controllable method to steer towards desired cultures.

We introduce a new agentic artificial intelligence (AI) platform for portfolio management. Our architecture consists of three layers. First, two large language model (LLM) agents are assigned specialized tasks: one agent screens for firms with desirable fundamentals, while a sentiment analysis agent screens for firms with desirable news. Second, these agents deliberate to generate and agree upon buy and sell signals from a large portfolio, substantially narrowing the pool of candidate assets. Finally, we apply a high-dimensional precision matrix estimation procedure to determine optimal portfolio weights. A defining theoretical feature of our framework is that the number of assets in the portfolio is itself a random variable, realized through the screening process. We introduce the concept of sensible screening and establish that, under mild screening errors, the squared Sharpe ratio of the screened portfolio consistently estimates its target. Empirically, our method achieves superior Sharpe ratios relative to an unscreened baseline portfolio and to conventional screening approaches, evaluated on S&P 500 data over the period 2020--2024.

Benchmarks and leaderboards are how NLP most often communicates progress, but in the LLM era they are increasingly easy to misread. Scores can reflect benchmark-chasing, hidden evaluation choices, or accidental exposure to test content -- not just broad capability. Closed benchmarks delay some of these issues, but reduce transparency and make it harder for the community to learn from results. We argue for a complementary practice: an Olympiad-style evaluation event where problems are sealed until evaluation, submissions are frozen in advance, and all entries run through one standardized harness. After scoring, the full task set and evaluation code are released so results can be reproduced and audited. This design aims to make strong performance harder to ``manufacture'' and easier to trust.

Accurate short-term forecasting of air temperature and relative humidity is critical for urban management, especially in topographically complex cities such as Chongqing, China. This study compares seven machine learning models: eXtreme Gradient Boosting (XGBoost), Random Forest, Support Vector Regression (SVR), Multi-Layer Perceptron (MLP), Decision Tree, Long Short-Term Memory (LSTM) networks, and Convolutional Neural Network (CNN)-LSTM (CNN-LSTM), for hourly prediction using real-world open data. Based on a unified framework of data preprocessing, lag-feature construction, rolling statistical features, and time-series validation, the models are systematically evaluated in terms of predictive accuracy and robustness. The results show that XGBoost achieves the best overall performance, with a test mean absolute error (MAE) of 0.302 °C for air temperature and 1.271% for relative humidity, together with an average R2 of 0.989 across the two forecasting tasks. These findings demonstrate the strong effectiveness of tree-based ensemble learning for structured meteorological time-series forecasting and provide practical guidance for intelligent meteorological forecasting in mountainous cities.

The rapid diffusion of large language models and the growth in their capability has enabled the emergence of online environments populated by autonomous AI agents that interact through natural language. These platforms provide a novel empirical setting for studying collective dynamics among artificial agents. In this paper we analyze the interaction network of Moltbook, a social platform composed entirely of LLM based agents, using tools from network science. The dataset comprises 39,924 users, 235,572 posts, and 1,540,238 comments collected through web scraping. We construct a directed weighted network in which nodes represent agents and edges represent commenting interactions. Our analysis reveals strongly heterogeneous connectivity patterns characterized by heavy tailed degree and activity distributions. At the mesoscale, the network exhibits a pronounced core periphery organization in which a very small structural core (0.9% of nodes) concentrates a large fraction of connectivity. Robustness experiments show that the network is relatively resilient to random node removal but highly vulnerable to targeted attacks on highly connected nodes, particularly those with high out degree. These findings indicate that the interaction structure of AI agent social systems may develop strong centralization and structural fragility, providing new insights into the collective organization of LLM native social environments.

Human-robot interaction is increasingly moving toward multi-robot, socially grounded environments. Existing systems struggle to integrate multimodal perception, embodied expression, and coordinated decision-making in a unified framework. This limits natural and scalable interaction in shared physical spaces. We address this gap by introducing a multimodal framework for human-multi-agent interaction in which each robot operates as an autonomous cognitive agent with integrated multimodal perception and Large Language Model (LLM)-driven planning grounded in embodiment. At the team level, a centralized coordination mechanism regulates turn-taking and agent participation to prevent overlapping speech and conflicting actions. Implemented on two humanoid robots, our framework enables coherent multi-agent interaction through interaction policies that combine speech, gesture, gaze, and locomotion. Representative interaction runs demonstrate coordinated multimodal reasoning across agents and grounded embodied responses. Future work will focus on larger-scale user studies and deeper exploration of socially grounded multi-agent interaction dynamics.

Large Language Models(LLMs) are widely deployed, yet are vulnerable to jailbreak prompts that elicit policy-violating outputs. Although prior studies have uncovered these risks, they typically treat all tokens as equally important during prompt mutation, overlooking the varying contributions of individual tokens to triggering model refusals. Consequently, these attacks introduce substantial redundant searching under query-constrained scenarios, reducing attack efficiency and hindering comprehensive vulnerability assessment. In this work, we conduct a token-level analysis of refusal behavior and observe that token contributions are highly skewed rather than uniform. Moreover, we find strong cross-model consistency in refusal tendencies, enabling the use of a surrogate model to estimate token-level contributions to the target model's refusals. Motivated by these findings, we propose TriageFuzz, a token-aware jailbreak fuzzing framework that adapts the fuzz testing approach with a series of customized designs. TriageFuzz leverages a surrogate model to estimate the contribution of individual tokens to refusal behaviors, enabling the identification of sensitive regions within the prompt. Furthermore, it incorporates a refusal-guided evolutionary strategy that adaptively weights candidate prompts with a lightweight scorer to steer the evolution toward bypassing safety constraints. Extensive experiments on six open-source LLMs and three commercial APIs demonstrate that TriageFuzz achieves comparable attack success rates (ASR) with significantly reduced query costs. Notably, it attains a 90% ASR with over 70% fewer queries compared to baselines. Even under an extremely restrictive budget of 25 queries, TriageFuzz outperforms existing methods, improving ASR by 20-40%.

Mechanistic interpretability reveals that safety-critical behaviors (e.g., alignment, jailbreak, backdoor) in Large Language Models (LLMs) are grounded in specialized functional components. However, existing safety attribution methods struggle with generalization and reliability due to their reliance on heuristic, domain-specific metrics and search algorithms. To address this, we propose \ourmethod, a unified safety interpretability framework that identifies functionally complete safety circuits in LLMs via optimization. Unlike methods focusing on isolated heads or neurons, \ourmethod introduces differentiable binary masks to extract multi-granular circuits through gradient descent on safety datasets, while integrates Safety Circuit Tuning to utilize these sparse circuits for efficient safety fine-tuning. We validate \ourmethod in two key scenarios in LLM safety: \textbf{(1) backdoor attacks}, identifying a backdoor circuit with 0.42\% sparsity, whose ablation eradicates the Attack Success Rate (ASR) from 100\% $\to$ 0.4\% while retaining over 99\% general utility; \textbf{(2) safety alignment}, localizing an alignment circuit with 3.03\% heads and 0.79\% neurons, whose removal spikes ASR from 0.8\% $\to$ 96.9\%, whereas excluding this circuit during helpfulness fine-tuning maintains 96.5\% safety retention.

Online safety fault diagnosis is essential for lithium-ion batteries in electric vehicles(EVs), particularly under complex and rare safety-critical conditions in real-world operation. In this work, we develop an online battery fault diagnosis network based on a deep anomaly detection framework combining kernel one-class classification and minimum-volume estimation. Mechanical constraints and spike-timing-dependent plasticity(STDP)-based dynamic representations are introduced to improve complex fault characterization and enable a more compact normal-state boundary. The proposed method is validated using 8.6 million valid data points collected from 20 EVs. Compared with several advanced baseline methods, it achieves average improvements of 7.59% in TPR, 27.92% in PPV, 18.28% in F1 score, and 23.68% in AUC. In addition, we analyze the spatial separation of fault representations before and after modeling, and further enhance framework robustness by learning the manifold structure in the latent space. The results also suggest the possible presence of shared causal structures across different fault types, highlighting the promise of integrating deep learning with physical constraints and neural dynamics for battery safety diagnosis.

In this paper, we introduce an autoencoder (AE)-based scheme for end-to-end optimization of a multi-user molecule mixture communication system. In the proposed scheme, each transmitter leverages an encoder network that maps the user symbol to a molecule mixture. The mixtures then propagate through the channel to the receiver, which samples the channel using a non-linear, cross-reactive sensor array. A decoder network then estimates the symbol transmitted by each user based on the sensor observations. The proposed scheme achieves, for a given signal-to-noise ratio, lower symbol error rates than a baseline scheme from the literature in a single-user setting with full channel state information. We additionally demonstrate that the proposed AE-based scheme allows reliable communication when the channel is unknown or changing. Finally, we show that for multiple access the system can account for different user priorities. In summary, the proposed AE-based scheme enables end-to-end system optimization in complex scenarios unsuitable for analytical treatment and thereby brings molecular communication systems closer to real-world deployment.

Permutation invariance is fundamental in molecular point-cloud generation, yet most diffusion models enforce it indirectly via permutation-equivariant networks on an ordered space. We propose to model diffusion directly on the quotient manifold $\tilde{\calX}=\sR^{d\times N}/S_N$, where all atom permutations are identified. We show that the heat kernel on $\tilde{\calX}$ admits an explicit expression as a sum of Euclidean heat kernels over permutations, which clarifies how diffusion on the quotient differs from ordered-particle diffusion. Training requires a permutation-symmetrized score involving an intractable sum over $S_N$; we derive an expectation form over a posterior on permutations and approximate it using MCMC in permutation space. We evaluate on unconditional 3D molecule generation on QM9 under the EQGAT-Diff protocol, using SemlaFlow-style backbone and treating all variables continuously. The results demonstrate that quotient-based permutation symmetrization is practical and yields competitive generation quality with improved efficiency.

Fully Homomorphic Encryption (FHE) is rapidly emerging as a promising foundation for privacy-preserving cloud services, enabling computation directly on encrypted data. As FHE implementations mature and begin moving toward practical deployment in domains such as secure finance, biomedical analytics, and privacy-preserving AI, a critical question remains insufficiently explored: how reliable is FHE computation on real hardware? This question is especially important because, compared with plaintext computation, FHE incurs much higher computational overhead, making it more susceptible to transient hardware faults. Moreover, data corruptions are likely to remain silent: the FHE service has no access to the underlying plaintext, causing unawareness even though the corresponding decrypted result has already been corrupted. To this end, we conduct a comprehensive evaluation of SDCs in FHE ciphertext computation. Through large-scale fault-injection experiments, we characterize the vulnerability of FHE to transient faults, and through a theoretical analysis of error-propagation behaviors, we gain deeper algorithmic insight into the mechanisms underlying this vulnerability. We further assess the effectiveness of different fault-tolerance mechanisms for mitigating these faults.

Integrating Artificial Intelligence (AI) into Non-Terrestrial Networks (NTN) is constrained by the joint limits of satellite SWaP and feeder-link capacity, which directly impact O-RAN closed-loop control and model lifecycle management. This paper studies the feasibility of distributing the O-RAN control hierarchy across Ground, LEO, and GEO segments through a Split-RIC architecture. We compare three deployment scenarios: (i) ground-centric control with telemetry streaming, (ii) ground--LEO Split-RIC with on-board inference and store-and-forward learning, and (iii) GEO--LEO multi-layer control enabled by inter-satellite links. For each scenario, we derive closed-form expressions for lifecycle energy and lifecycle latency that account for training-data transfer, model dissemination, and near-real-time inference. Numerical sensitivity analysis over feeder-link conditions, model complexity, and orbital intermittency yields operator-relevant feasibility regions that delineate when on-board inference and non-terrestrial learning loops are physically preferable to terrestrial offloading.

The advancing fluency of LLMs raises important questions about their ability to emulate complex human traits, including emotional expression and personality, across diverse linguistic and cultural contexts. This study investigates whether LLMs can convincingly mimic emotional nuance in English and personality markers in Arabic, a critical under-resourced language with unique linguistic and cultural characteristics. We conduct two tasks across six models:Jais, Mistral, LLaMA, GPT-4o, Gemini, and DeepSeek. First, we evaluate whether machine classifiers can reliably distinguish between human-authored and AI-generated texts. Second, we assess the extent to which LLM-generated texts exhibit emotional or personality traits comparable to those of humans. Our results demonstrate that AI-generated texts are distinguishable from human-authored ones (F1>0.95), though classification performance deteriorates on paraphrased samples, indicating a reliance on superficial stylistic cues. Emotion and personality classification experiments reveal significant generalization gaps: classifiers trained on human data perform poorly on AI-generated texts and vice versa, suggesting LLMs encode affective signals differently from humans. Importantly, augmenting training with AI-generated data enhances performance in the Arabic personality classification task, highlighting the potential of synthetic data to address challenges in under-resourced languages. Model-specific analyses show that GPT-4o and Gemini exhibit superior affective coherence. Linguistic and psycholinguistic analyses reveal measurable divergences in tone, authenticity, and textual complexity between human and AI texts. These findings have implications for affective computing, authorship attribution, and responsible AI deployment, particularly within underresourced language contexts where generative AI detection and alignment pose unique challenges.

Efficient scheduling of directed acyclic graphs (DAGs) in heterogeneous environments is challenging due to resource capacities and dependencies. In practice, the need for adaptability across environments with varying resource pools and task types, alongside rapid schedule generation, complicates these challenges. We propose WeCAN, an end-to-end reinforcement learning framework for heterogeneous DAG scheduling that addresses task--pool compatibility coefficients and generation-induced optimality gaps. It adopts a two-stage single-pass design: a single forward pass produces task--pool scores and global parameters, followed by a generation map that constructs schedules without repeated network calls. Its weighted cross-attention encoder models task--pool interactions gated by compatibility coefficients, and is size-agnostic to environment fluctuations. Moreover, widely used list-scheduling maps can incur generation-induced optimality gaps from restricted reachability. We introduce an order-space analysis that characterizes the reachable set of generation maps via feasible schedule orders, explains the mechanism behind generation-induced gaps, and yields sufficient conditions for gap elimination. Guided by these conditions, we design a skip-extended realization with an analytically parameterized decreasing skip rule, which enlarges the reachable order set while preserving single-pass efficiency. Experiments on computation graphs and real-world TPC-H DAGs demonstrate improved makespan over strong baselines, with inference time comparable to classical heuristics and faster than multi-round neural schedulers.

We investigate neural ordinary and stochastic differential equations (neural ODEs and SDEs) to model stochastic dynamics in fully and partially observed environments within a model-based reinforcement learning (RL) framework. Through a sequence of simulations, we show that neural SDEs more effectively capture the inherent stochasticity of transition dynamics, enabling high-performing policies with improved sample efficiency in challenging scenarios. We leverage neural ODEs and SDEs for efficient policy adaptation to changes in environment dynamics via inverse models, requiring only limited interactions with the new environment. To address partial observability, we introduce a latent SDE model that combines an ODE with a GAN-trained stochastic component in latent space. Policies derived from this model provide a strong baseline, outperforming or matching general model-based and model-free approaches across stochastic continuous-control benchmarks. This work demonstrates the applicability of action-conditional latent SDEs for RL planning in environments with stochastic transitions. Our code is available at: https://github.com/ChaoHan-UoS/NeuralRL

When learning a novel complex task, people often form efficient reusable abstractions that simplify future work, despite uncertainty about the future. We study this process in a visual puzzle task where participants define and reuse helpers -- intermediate constructions that capture repeating structure. In an online experiment, participants solved puzzles of increasing difficulty. Early on, they created many helpers, favouring completeness over efficiency. With experience, helper use became more selective and efficient, reflecting sensitivity to reuse and cost. Access to helpers enabled participants to solve puzzles that were otherwise difficult or impossible. Computational modelling shows that human decision times and number of operations used to complete a puzzle increase with search space estimated by a program induction model with library learning. In contrast, raw program length predicts failure but not effort. Together, these results point to online library learning as a core mechanism in human problem solving, allowing people to flexibly build, refine, and reuse abstractions as task demands grow.

Large language model (LLM)-based agents rely on memory mechanisms to reuse knowledge from past problem-solving experiences. Existing approaches typically construct memory in a per-agent manner, tightly coupling stored knowledge to a single model's reasoning style. In modern deployments with heterogeneous agents, a natural question arises: can a single memory system be shared across different models? We found that naively transferring memory between agents often degrades performance, as such memory entangles task-relevant knowledge with agent-specific biases. To address this challenge, we propose MemCollab, a collaborative memory framework that constructs agent-agnostic memory by contrasting reasoning trajectories generated by different agents on the same task. This contrastive process distills abstract reasoning constraints that capture shared task-level invariants while suppressing agent-specific artifacts. We further introduce a task-aware retrieval mechanism that conditions memory access on task category, ensuring that only relevant constraints are used at inference time. Experiments on mathematical reasoning and code generation benchmarks demonstrate that MemCollab consistently improves both accuracy and inference-time efficiency across diverse agents, including cross-modal-family settings. Our results show that the collaboratively constructed memory can function as a shared reasoning resource for diverse LLM-based agents.

Offline reinforcement learning (RL) can fit strong value functions from fixed datasets, yet reliable deployment still hinges on the action selection interface used to query them. When the dataset induces a branched or multimodal action landscape, unimodal policy extraction can blur competing hypotheses and yield "in-between" actions that are weakly supported by data, making decisions brittle even with a strong critic. We introduce GEM (Guided Expectation-Maximization), an analytical framework that makes action selection both multimodal and explicitly controllable. GEM trains a Gaussian Mixture Model (GMM) actor via critic-guided, advantage-weighted EM-style updates that preserve distinct components while shifting probability mass toward high-value regions, and learns a tractable GMM behavior model to quantify support. During inference, GEM performs candidate-based selection: it generates a parallel candidate set and reranks actions using a conservative ensemble lower-confidence bound together with behavior-normalized support, where the behavior log-likelihood is standardized within each state's candidate set to yield stable, comparable control across states and candidate budgets. Empirically, GEM is competitive across D4RL benchmarks, and offers a simple inference-time budget knob (candidate count) that trades compute for decision quality without retraining.

Empowering large language models with long-term memory is crucial for building agents that adapt to users' evolving needs. However, prior evaluations typically interleave preference-related dialogues with irrelevant conversations, reducing the task to needle-in-a-haystack retrieval while ignoring relationships between events that drive the evolution of user preferences. Such settings overlook a fundamental characteristic of real-world personalization: preferences emerge gradually and accumulate across interactions within noisy contexts. To bridge this gap, we introduce PERMA, a benchmark designed to evaluate persona consistency over time beyond static preference recall. Additionally, we incorporate (1) text variability and (2) linguistic alignment to simulate erratic user inputs and individual idiolects in real-world data. PERMA consists of temporally ordered interaction events spanning multiple sessions and domains, with preference-related queries inserted over time. We design both multiple-choice and interactive tasks to probe the model's understanding of persona along the interaction timeline. Experiments demonstrate that by linking related interactions, advanced memory systems can extract more precise preferences and reduce token consumption, outperforming traditional semantic retrieval of raw dialogues. Nevertheless, they still struggle to maintain a coherent persona across temporal depth and cross-domain interference, highlighting the need for more robust personalized memory management in agents. Our code and data are open-sourced at https://github.com/PolarisLiu1/PERMA.

Internet memes represent a popular form of multimodal online communication and often use figurative elements to convey layered meaning through the combination of text and images. However, it remains largely unclear how multimodal large language models (MLLMs) combine and interpret visual and textual information to identify figurative meaning in memes. To address this gap, we evaluate eight state-of-the-art generative MLLMs across three datasets on their ability to detect and explain six types of figurative meaning. In addition, we conduct a human evaluation of the explanations generated by these MLLMs, assessing whether the provided reasoning supports the predicted label and whether it remains faithful to the original meme content. Our findings indicate that all models exhibit a strong bias to associate a meme with figurative meaning, even when no such meaning is present. Qualitative analysis further shows that correct predictions are not always accompanied by faithful explanations.

We study learning under regime variation, where the learner, its memory state, and the evaluative conditions may evolve over time. This paper is a foundational and structural contribution: its goal is to define the core learning-theoretic objects required for such settings and to establish their first theorem-supporting consequences. The paper develops a regime-varying framework centered on admissible transport, protected-core preservation, and evaluator-aware learning evolution. It records the immediate closure consequences of admissibility, develops a structural obstruction argument for faithful fixed-ontology reduction in genuinely multi-regime settings, and introduces a protected-stability template together with explicit numerical and symbolic witnesses on controlled subclasses, including convex and deductive settings. It also establishes theorem-layer results on evaluator factorization, morphisms, composition, and partial kernel-level alignment across semantically commensurable layers. A worked two-regime example makes the admissibility certificate, protected evaluative core, and regime-variation cost explicit on a controlled subclass. The symbolic component is deliberately restricted in scope: the paper establishes a first kernel-level compatibility result together with a controlled monotonic deductive witness. The manuscript should therefore be read as introducing a structured learning-theoretic framework for regime-varying learning together with its first theorem-supporting layer, not as a complete quantitative theory of all learning systems.

Amidst the rising capabilities of generative AI to mimic specific human styles, this study investigates the ability of state-of-the-art large language models (LLMs), including GPT-4o, Gemini 1.5 Pro, and Claude Sonnet 3.5, to emulate the authorial signatures of prominent literary and political figures: Walt Whitman, William Wordsworth, Donald Trump, and Barack Obama. Utilizing a zero-shot prompting framework with strict thematic alignment, we generated synthetic corpora evaluated through a complementary framework combining transformer-based classification (BERT) and interpretable machine learning (XGBoost). Our methodology integrates Linguistic Inquiry and Word Count (LIWC) markers, perplexity, and readability indices to assess the divergence between AI-generated and human-authored text. Results demonstrate that AI-generated mimicry remains highly detectable, with XGBoost models trained on a restricted set of eight stylometric features achieving accuracy comparable to high-dimensional neural classifiers. Feature importance analyses identify perplexity as the primary discriminative metric, revealing a significant divergence in the stochastic regularity of AI outputs compared to the higher variability of human writing. While LLMs exhibit distributional convergence with human authors on low-dimensional heuristic features, such as syntactic complexity and readability, they do not yet fully replicate the nuanced affective density and stylistic variance inherent in the human-authored corpus. By isolating the specific statistical gaps in current generative mimicry, this study provides a comprehensive benchmark for LLM stylistic behavior and offers critical insights for authorship attribution in the digital humanities and social media.

Determining atomistic structures from characterization data is one of the most common yet intricate problems in materials science. Particularly in amorphous materials, proposing structures that balance realism and agreement with experiments requires expert guidance, good interatomic potentials, or both. Here, we introduce GLASS, a generative framework that inverts multi-modal spectroscopic measurements into realistic atomistic structures without knowledge of the potential energy surface. A score-based model learns a structural prior from low-fidelity data and samples out-of-distribution structures conditioned on differentiable spectral targets. Reconstructions using pair distribution functions (PDFs), X-ray absorption spectroscopy, and diffraction measurements quantify the complementarity between spectral modalities and demonstrate that PDFs is the most informative probe for our framework. We use GLASS to rationalize three contested experimental problems: paracrystallinity in amorphous silicon, a liquid-liquid phase transition in sulfur, and ball-milled amorphous ice. In each case, generated structures reproduce experimental measurements and reveal mechanisms inaccessible to diffraction analysis alone.

We prove the tightest-known upper bounds on the sample complexity of multi-group learning. Our algorithm extends the one-inclusion graph prediction strategy using a generalization of bipartite $b$-matching. In the group-realizable setting, we provide a lower bound confirming that our algorithm's $\log n / n$ convergence rate is optimal in general. If one relaxes the learning objective such that the group on which we are evaluated is chosen obliviously of the sample, then our algorithm achieves the optimal $1/n$ convergence rate under group-realizability.

Spiking neural networks (SNNs) offer a biologically inspired computing paradigm with significant potential for energy-efficient neural processing. Among neural coding schemes of SNNs, Time-To-First-Spike (TTFS) coding, which encodes information through the precise timing of a neuron's first spike, provides exceptional energy efficiency and biological plausibility. Despite its theoretical advantages, existing TTFS models lack efficient training methods, suffering from high inference latency and limited performance. In this work, we present a comprehensive framework, which enables the efficient training of deep TTFS-coded SNNs by employing backpropagation throuh time (BPTT) algorithm. We name the generalized TTFS coding method in our framework as latency coding. The framework includes: (1) a latency encoding (LE) module with feature extraction and straight-through estimators to address severe information loss in direct intensity-to-latency mapping and ensure smooth gradient flow; (2) relaxation of the strict single-spike constraint of traditional TTFS, allowing neurons of intermediate layers to fire multiple times to mitigating gradient vanishing in deep networks; (3) a temporal adaptive decision (TAD) loss function that dynamically weights supervision signals based on sample-dependent confidence, resolving the incompatibility between latency coding and standard cross-entropy loss. Experimental results demonstrate that our method achieves state-of-the-art accuracy in comparison to existing TTFS-coded SNNs with ultra-low inference latency and superior energy efficiency. The framework also demonstrates improved robustness against input corruptions. Our study investigates the characteristics and potential of latency coding in scenarios demanding rapid response, providing valuable insights for further exploiting the temporal learning capabilities of SNNs.

Standard conformal anomaly detection provides marginal finite-sample guarantees under the assumption of exchangeability . However, real-world data often exhibit distribution shifts, necessitating a weighted conformal approach to adapt to local non-stationarity. We show that this adaptation induces a critical trade-off between the minimum attainable p-value and its stability. As importance weights localize to relevant calibration instances, the effective sample size decreases. This can render standard conformal p-values overly conservative for effective error control, while the smoothing technique used to mitigate this issue introduces conditional variance, potentially masking anomalies. We propose a continuous inference relaxation that resolves this dilemma by decoupling local adaptation from tail resolution via continuous weighted kernel density estimation. While relaxing finite-sample exactness to asymptotic validity, our method eliminates Monte Carlo variability and recovers the statistical power lost to discretization. Empirical evaluations confirm that our approach not only restores detection capabilities where discrete baselines yield zero discoveries, but outperforms standard methods in statistical power while maintaining valid marginal error control in practice.

Scaling autoregressive large language models (LLMs) has driven unprecedented progress but comes with vast computational costs. In this work, we tackle these costs by leveraging unstructured sparsity within an LLM's feedforward layers, the components accounting for most of the model parameters and execution FLOPs. To achieve this, we introduce a new sparse packing format and a set of CUDA kernels designed to seamlessly integrate with the optimized execution pipelines of modern GPUs, enabling efficient sparse computation during LLM inference and training. To substantiate our gains, we provide a quantitative study of LLM sparsity, demonstrating that simple L1 regularization can induce over 99% sparsity with negligible impact on downstream performance. When paired with our kernels, we show that these sparsity levels translate into substantial throughput, energy efficiency, and memory usage benefits that increase with model scale. We will release all code and kernels under an open-source license to promote adoption and accelerate research toward establishing sparsity as a practical axis for improving the efficiency and scalability of modern foundation models.

Coordination of view coverage via privacy-aware smart cameras is key to a more socially responsible urban intelligence. Rather than maximizing view coverage at any cost or over relying on expensive cryptographic techniques, we address how cameras can coordinate to legitimately monitor public spaces while excluding privacy-sensitive regions by design. This article proposes a decentralized framework in which interactive smart cameras coordinate to autonomously select their orientation via collective learning, while eliminating privacy violations via soft and hard constraint satisfaction. The approach scales to hundreds up to thousands of cameras without any centralized control. Experimental evidence shows 18.42% higher coverage efficiency and 85.53% lower privacy violation than baselines and other state-of-the-art approaches. This significant advance further unravels practical guidelines for operators and policymakers: how the field of view, spatial placement, and budget of cameras operating by ethically-aligned artificial intelligence jointly influence coverage efficiency and privacy protection in large-scale and sensitive urban environments.

Achieving real-time physics-based animation that generalizes across diverse 3D shapes and discretizations remains a fundamental challenge. We introduce PhysSkin, a physics-informed framework that addresses this challenge. In the spirit of Linear Blend Skinning, we learn continuous skinning fields as basis functions lifting motion subspace coordinates to full-space deformation, with subspace defined by handle transformations. To generate mesh-free, discretization-agnostic, and physically consistent skinning fields that generalize well across diverse 3D shapes, PhysSkin employs a new neural skinning fields autoencoder which consists of a transformer-based encoder and a cross-attention decoder. Furthermore, we also develop a novel physics-informed self-supervised learning strategy that incorporates on-the-fly skinning-field normalization and conflict-aware gradient correction, enabling effective balancing of energy minimization, spatial smoothness, and orthogonality constraints. PhysSkin shows outstanding performance on generalizable neural skinning and enables real-time physics-based animation.

Reward modeling represents a long-standing challenge in reinforcement learning from human feedback (RLHF) for aligning language models. Current reward modeling is heavily contingent upon experimental feedback data with high collection costs. In this work, we study \textit{implicit reward modeling} -- learning reward models from implicit human feedback (e.g., clicks and copies) -- as a cost-effective alternative. We identify two fundamental challenges in implicit reward modeling: (1) Implicit preference data lacks definitive negative samples, which makes standard positive-negative classification methods inapplicable; (2) Implicit preference data suffers from user preference bias, where different responses have different propensities to elicit user feedback actions, which exacerbates the difficulty of distinguishing definitive negative samples. To address these challenges, we propose ImplicitRM, which aims to learn unbiased reward models from implicit preference data. ImplicitRM stratifies training samples into four latent groups via a stratification model. Building on this, it derives a learning objective through likelihood maximization, which we prove is theoretically unbiased, effectively resolving both challenges. Experiments demonstrate that ImplicitRM learns accurate reward models across implicit preference datasets. Code is available on our project website.

Recent advances in generative recommendation have leveraged pretrained LLMs by formulating sequential recommendation as autoregressive generation over a unified token space comprising language tokens and itemic identifiers, where each item is represented by a compact sequence of discrete tokens, namely Semantic IDs (SIDs). This SID-based formulation enables efficient decoding over large-scale item corpora and provides a natural interface for LLM-based recommenders to leverage rich world knowledge. Meanwhile, breakthroughs in LLM reasoning motivate reasoning-enhanced recommendation, yet effective reasoning over SIDs remains underexplored and challenging. Itemic tokens are not natively meaningful to LLMs; moreover, recommendation-oriented SID reasoning is hard to evaluate, making high-quality supervision scarce. To address these challenges, we propose SIDReasoner, a two-stage framework that elicits reasoning over SIDs by strengthening SID--language alignment to unlock transferable LLM reasoning, rather than relying on large amounts of recommendation-specific reasoning traces. Concretely, SIDReasoner first enhances SID-language alignment via multi-task training on an enriched SID-centered corpus synthesized by a stronger teacher model, grounding itemic tokens in diverse semantic and behavioral contexts. Building on this enhanced alignment, SIDReasoner further improves recommendation reasoning through outcome-driven reinforced optimization, which guides the model toward effective reasoning trajectories without requiring explicit reasoning annotations. Extensive experiments on three real-world datasets demonstrate the effectiveness of our reasoning-augmented SID-based generative recommendation. Beyond accuracy, the results highlight the broader potential of large reasoning models for generative recommendation, including improved interpretability and cross-domain generalization.

Deepfakes generated by modern generative models pose a serious threat to information integrity, digital identity, and public trust. Existing detection methods are largely reactive, attempting to identify manipulations after they occur and often failing to generalize across evolving generation techniques. This motivates the need for proactive mechanisms that secure media authenticity at the time of creation. In this work, we introduce SAiW, a Source-Attributed Invisible watermarking Framework for proactive deepfake defense and media provenance verification. Unlike conventional watermarking methods that treat watermark payloads as generic signals, SAiW formulates watermark embedding as a source-conditioned representation learning problem, where watermark identity encodes the originating source and modulates the embedding process to produce discriminative and traceable signatures. The framework integrates feature-wise linear modulation to inject source identity into the embedding network, enabling scalable multi-source watermark generation. A perceptual guidance module derived from human visual system priors ensures that watermark perturbations remain visually imperceptible while maintaining robustness. In addition, a dual-purpose forensic decoder simultaneously reconstructs the embedded watermark and performs source attribution, providing both automated verification and interpretable forensic evidence. Extensive experiments across multiple deepfake datasets demonstrate that SAiW achieves high perceptual quality while maintaining strong robustness against compression, filtering, noise, geometric transformations, and adversarial perturbations. By binding digital media to its origin through invisible yet verifiable markers, SAiW enables reliable authentication and source attribution, providing a scalable foundation for proactive deepfake defense and trustworthy media provenance.

Centralized identity management systems continuously experience security and privacy challenges, motivating the exploration of Decentralized Identity (DI) and Self-Sovereign Identity (SSI) as user-focused alternatives. Although prior research has consolidated SSI principles and derived quality requirements for DI/SSI systems, it is significantly limited in integrating the user viewpoint. This work addresses this gap by embedding a user perspective into the requirements engineering process for DI/SSI systems. Building on existing SSI principles, composite requirements were decomposed into 24 simple quality or non-functional requirements (NFR). The resulting NFR are systematically mapped to the key actors, namely data owner, issuer, verifier, and system, based on varying degrees of responsibility and ownership. A dependency model is introduced to formalize relationships between actors. Inspired by trust modeling concepts, the model explicitly describes how actors interact and rely on each other for requirements fulfillment. By integrating user-centered requirements, responsibility allocation, ownership specification, and dependency modeling, this work provides the first structured model for DI/SSI system architectures.

High-dimensional stochastic optimal control (SOC) becomes harder with longer planning horizons: existing methods scale linearly in the horizon $T$, with performance often deteriorating exponentially. We overcome these limitations for a subclass of linearly-solvable SOC problems-those whose uncontrolled drift is the gradient of a potential. In this setting, the Hamilton-Jacobi-Bellman equation reduces to a linear PDE governed by an operator $\mathcal{L}$. We prove that, under the gradient drift assumption, $\mathcal{L}$ is unitarily equivalent to a Schrödinger operator $\mathcal{S} = -Δ+ \mathcal{V}$ with purely discrete spectrum, allowing the long-horizon control to be efficiently described via the eigensystem of $\mathcal{L}$. This connection provides two key results: first, for a symmetric linear-quadratic regulator (LQR), $\mathcal{S}$ matches the Hamiltonian of a quantum harmonic oscillator, whose closed-form eigensystem yields an analytic solution to the symmetric LQR with \emph{arbitrary} terminal cost. Second, in a more general setting, we learn the eigensystem of $\mathcal{L}$ using neural networks. We identify implicit reweighting issues with existing eigenfunction learning losses that degrade performance in control tasks, and propose a novel loss function to mitigate this. We evaluate our method on several long-horizon benchmarks, achieving an order-of-magnitude improvement in control accuracy compared to state-of-the-art methods, while reducing memory usage and runtime complexity from $\mathcal{O}(Td)$ to $\mathcal{O}(d)$.

Multilingual intent classification is central to customer-service systems on global logistics platforms, where models must process noisy user queries across languages and hierarchical label spaces. Yet most existing multilingual benchmarks rely on machine-translated text, which is typically cleaner and more standardized than native customer requests and can therefore overestimate real-world robustness. We present a public benchmark for hierarchical multilingual intent classification constructed from real logistics customer-service logs. The dataset contains approximately 30K de-identified, stand-alone user queries curated from 600K historical records through filtering, LLM-assisted quality control, and human verification, and is organized into a two-level taxonomy with 13 parent and 17 leaf intents. English, Spanish, and Arabic are included as seen languages, while Indonesian, Chinese, and additional test-only languages support zero-shot evaluation. To directly measure the gap between synthetic and real evaluation, we provide paired native and machine-translated test sets and benchmark multilingual encoders, embedding models, and small language models under flat and hierarchical protocols. Results show that translated test sets substantially overestimate performance on noisy native queries, especially for long-tail intents and cross-lingual transfer, underscoring the need for more realistic multilingual intent benchmarks.

Large language models (LLMs) can be misused to reveal sensitive information, such as weapon-making instructions or writing malware. LLM providers rely on $\emph{monitoring}$ to detect and flag unsafe behavior during inference. An open security challenge is $\emph{adaptive}$ adversaries who craft attacks that simultaneously (i) evade detection while (ii) eliciting unsafe behavior. Adaptive attackers are a major concern as LLM providers cannot patch their security mechanisms, since they are unaware of how their models are being misused. We cast $\emph{robust}$ LLM monitoring as a security game, where adversaries who know about the monitor try to extract sensitive information, while a provider must accurately detect these adversarial queries at low false positive rates. Our work (i) shows that existing LLM monitors are vulnerable to adaptive attackers and (ii) designs improved defenses through $\emph{activation watermarking}$ by carefully introducing uncertainty for the attacker during inference. We find that $\emph{activation watermarking}$ outperforms guard baselines by up to $52\%$ under adaptive attackers who know the monitoring algorithm but not the secret key.

Benchmarking AI systems in multi-turn interactive scenarios is essential for understanding their practical capabilities in real-world applications. However, existing evaluation protocols are highly heterogeneous, differing significantly in dataset formats, model interfaces, and evaluation pipelines, which severely impedes systematic comparison. In this work, we present UniDial-EvalKit (UDE), a unified evaluation toolkit for assessing interactive AI systems. The core contribution of UDE lies in its holistic unification: it standardizes heterogeneous data formats into a universal schema, streamlines complex evaluation pipelines through a modular architecture, and aligns metric calculations under a consistent scoring interface. It also supports efficient large-scale evaluation through parallel generation and scoring, as well as checkpoint-based caching to eliminate redundant computation. Validated across diverse multi-turn benchmarks, UDE not only guarantees high reproducibility through standardized workflows and transparent logging, but also significantly improves evaluation efficiency and extensibility. We make the complete toolkit and evaluation scripts publicly available to foster a standardized benchmarking ecosystem and accelerate future breakthroughs in interactive AI.

Foundation models for vision have transformed visual recognition with powerful pretrained representations and strong zero-shot capabilities, yet their potential for data-efficient learning remains largely untapped. Active Learning (AL) aims to minimize annotation costs by strategically selecting the most informative samples for labeling, but existing methods largely overlook the rich multimodal knowledge embedded in modern vision-language models (VLMs). We introduce Conformal Cross-Modal Acquisition (CCMA), a novel AL framework that bridges vision and language modalities through a teacher-student architecture. CCMA employs a pretrained VLM as a teacher to provide semantically grounded uncertainty estimates, conformally calibrated to guide sample selection for a vision-only student model. By integrating multimodal conformal scoring with diversity-aware selection strategies, CCMA achieves superior data efficiency across multiple benchmarks. Our approach consistently outperforms state-of-the-art AL baselines, demonstrating clear advantages over methods relying solely on uncertainty or diversity metrics.

Deploying safety-critical agents requires anticipating the consequences of actions before they are executed. While world models offer a paradigm for this proactive foresight, current approaches relying on visual simulation incur prohibitive latencies, often exceeding several seconds per step. In this work, we challenge the assumption that visual processing is necessary for failure prevention. We show that a trained policy's latent state, combined with its planned actions, already encodes sufficient information to anticipate action outcomes, making visual simulation redundant for failure prevention. To this end, we introduce DILLO (DIstiLLed Language-ActiOn World Model), a fast steering layer that shifts the paradigm from "simulate-then-act" to "describe-then-act." DILLO is trained via cross-modal distillation, where a privileged Vision Language Model teacher annotates offline trajectories and a latent-conditioned Large Language Model student learns to predict semantic outcomes. This creates a text-only inference path, bypassing heavy visual generation entirely, achieving a 14x speedup over baselines. Experiments on MetaWorld and LIBERO demonstrate that DILLO produces high-fidelity descriptions of the next state and is able to steer the policy, improving episode success rate by up to 15 pp and 9.3 pp on average across tasks.

The widespread adoption of Large Language Models (LLMs) has made the detection of AI-Generated text a pressing and complex challenge. Although many detection systems report high benchmark accuracy, their reliability in real-world settings remains uncertain, and their interpretability is often unexplored. In this work, we investigate whether contemporary detectors genuinely identify machine authorship or merely exploit dataset-specific artefacts. We propose an interpretable detection framework that integrates linguistic feature engineering, machine learning, and explainable AI techniques. When evaluated on two prominent benchmark corpora, namely PAN CLEF 2025 and COLING 2025, our model trained on 30 linguistic features achieves leaderboard-competitive performance, attaining an F1 score of 0.9734. However, systematic cross-domain and cross-generator evaluation reveals substantial generalisation failure: classifiers that excel in-domain degrade significantly under distribution shift. Using SHAP- based explanations, we show that the most influential features differ markedly between datasets, indicating that detectors often rely on dataset-specific stylistic cues rather than stable signals of machine authorship. Further investigation with in-depth error analysis exposes a fundamental tension in linguistic-feature-based AI text detection: the features that are most discriminative on in-domain data are also the features most susceptible to domain shift, formatting variation, and text-length effects. We believe that this knowledge helps build AI detectors that are robust across different settings. To support replication and practical use, we release an open-source Python package that returns both predictions and instance-level explanations for individual texts.

Large Language Models (LLMs) play a pivotal role in both academic research and broader societal applications. LLMs are increasingly used in software testing activities such as test case generation, selection, and repair. However, several important questions remain: (1) do LLMs possess enough information about software testing principles to perform software testing tasks effectively? (2) do LLMs possess sufficient conceptual understanding of software testing to answer software testing questions under metamorphic transformations? and (3) do certain properties of software testing questions influence the performance of LLMs? To answer these questions, this study evaluates 60 multimodal language models from both commercial vendors and the open-source community. The evaluation is performed using 30 sample exams of different types (core foundation, core advanced, specialist, and expert) from the International Software Testing Qualifications Board (ISTQB), which are used to assess the competence of human testers. In total, each model is evaluated on 1,171 questions. Furthermore, to ensure sufficient conceptual understanding, the models are also tested on exam questions transformed using context-preserving metamorphic techniques. Two models passed all the certifications by scoring at least 65% in all of the 30 certification exams, with commercial models generally outperforming open-source ones. We analyze the reasons behind incorrect answers and provide recommendations for improving the design of software testing certification exams.

The expansion of generative AI and LLM services underscores the growing need for adaptive mechanisms to select an appropriate available model to respond to a user's prompts. Recent works have proposed offline and online learning formulations to identify the optimal generative AI model for an input prompt, based solely on maximizing prompt-based fidelity evaluation scores, e.g., CLIP-Score in text-to-image generation. However, such fidelity-based selection methods overlook the diversity of generated outputs, and hence, they can fail to address potential diversity shortcomings in the generated responses. In this paper, we introduce the Diversity-Aware Kernelized Upper Confidence Bound (DAK-UCB) method as a contextual bandit algorithm for the online selection of generative models with diversity considerations. The proposed DAK-UCB method incorporates both fidelity and diversity-related metrics into the selection process. We design this framework based on prompt-aware diversity score functions that decompose to a two-sample-based expectation over prompt-output pairs in the previous generation rounds. Specifically, we illustrate the application of our framework using joint kernel distance and kernel entropy measures. Our experimental results demonstrate the effectiveness of DAK-UCB in promoting diversity-aware model selection while maintaining fidelity in the generations for a sequence of prompts. The code is available at https://github.com/Donya-Jafari/DAK-UCB.

Automated knowledge graph (KG) construction is essential for navigating the rapidly expanding body of scientific literature. However, existing approaches struggle to recognize long multi-word entities, often fail to generalize across domains, and typically overlook the hierarchical nature of scientific knowledge. While general-purpose large language models (LLMs) offer adaptability, they are computationally expensive and yield inconsistent accuracy on specialized tasks. As a result, current KGs are shallow and inconsistent, limiting their utility for exploration and synthesis. We propose a two-stage framework for scalable, zero-shot scientific KG construction. The first stage, Z-NERD, introduces (i) Orthogonal Semantic Decomposition (OSD), which promotes domain-agnostic entity recognition by isolating semantic "turns" in text, and (ii) a Multi-Scale TCQK attention mechanism that captures coherent multi-word entities through n-gram-aware attention heads. The second stage, HGNet, performs relation extraction with hierarchy-aware message passing, explicitly modeling parent, child, and peer relations. To enforce global consistency, we introduce two complementary objectives: a Differentiable Hierarchy Loss to discourage cycles and shortcut edges, and a Continuum Abstraction Field (CAF) Loss that embeds abstraction levels along a learnable axis in Euclidean space. This is the first approach to formalize hierarchical abstraction as a continuous property within standard Euclidean embeddings, offering a simpler alternative to hyperbolic methods. We release SPHERE (https://github.com/basiralab/SPHERE), a multi-domain benchmark for hierarchical relation extraction. Our framework establishes a new state of the art on SciERC, SciER, and SPHERE, improving NER by 8.08% and RE by 5.99% on out-of-distribution tests. In zero-shot settings, gains reach 10.76% for NER and 26.2% for RE.

Drones are becoming popular as a complementary system for \ac{ems}. Although several pilot studies and flight trials have shown the feasibility of drone-assisted \ac{aed} delivery, running a full-scale operational network remains challenging due to high capital expenditure and environmental uncertainties. In this paper, we formulate a reliability-informed Bayesian learning framework for designing drone-assisted \ac{aed} delivery networks under environmental and operational uncertainty. We propose our objective function based on the survival probability of \ac{ohca} patients to identify the ideal locations of drone stations. Moreover, we consider the coverage of existing \ac{ems} infrastructure to improve the response reliability in remote areas. We illustrate our proposed method using geographically referenced cardiac arrest data from Scotland. The result shows how environmental variability and spatial demand patterns influence optimal drone station placement across urban and rural regions. In addition, we assess the robustness of the network and evaluate its economic viability using a cost-effectiveness analysis based on expected \ac{qaly}. The findings suggest that drone-assisted \ac{aed} delivery is expected to be cost-effective and has the potential to significantly improve the emergency response coverage in rural and urban areas with longer ambulance response times.

Gödel agent realize recursive self-improvement: an agent inspects its own policy and traces and then modifies that policy in a tested loop. We introduce Polaris, a Gödel agent for compact models that performs policy repair via experience abstraction, turning failures into policy updates through a structured cycle of analysis, strategy formation, abstraction, and minimal code pat ch repair with conservative checks. Unlike response level self correction or parameter tuning, Polaris makes policy level changes with small, auditable patches that persist in the policy and are reused on unseen instances within each benchmark. As part of the loop, the agent engages in meta reasoning: it explains its errors, proposes concrete revisions to its own policy, and then updates the policy. To enable cumulative policy refinement, we introduce experience abstraction, which distills failures into compact, reusable strategies that transfer to unseen instances. On MGSM, DROP, GPQA, and LitBench (covering arithmetic reasoning, compositional inference, graduate-level problem solving, and creative writing evaluation), a 7-billion-parameter model equipped with Polaris achieves consistent gains over the base policy and competitive baselines.

Microservice-based applications are characterized by stochastic latencies arising from long-tail execution patterns and heterogeneous resource constraints across computational nodes. To address this challenge, we first formulate the problem using Quadratic Unconstrained Binary Optimization (QUBO), which aligns the problem with emerging quantum-optimization paradigms. Building upon this, we propose Q-GARS (Quantum-Guided Adaptive Robust Scheduling), a hybrid framework that integrates the QUBO model with Simulated Quantum Annealing (SQA) based combinatorial search and online rescheduling mechanisms, enabling global microservice rank generation and real-time robust adjustment. We treat the SQA-produced rank as a soft prior, and update a closed-loop trust weight to adaptively switch and mix between this prior and a robust proportional-fairness allocator, maintaining robustness under prediction failures and runtime disturbances. Simulation results demonstrate that Q-GARS achieves an average weighted completion time improvement of 2.1\% relative to a greedy baseline of the remaining shortest processing-time (SRPT), with performance gains reaching up to 16.8\% in heavy-tailed latency. The adaptive mechanism reduces tail latency under high-variance conditions. In addition, Q-GARS achieves a mean node resource utilization rate of 0.817, which is 1.1 percentage points above the robust baseline (0.806).

A human's moral decision depends heavily on the context. Yet research on LLM morality has largely studied fixed scenarios. We address this gap by introducing Contextual MoralChoice, a dataset of moral dilemmas with systematic contextual variations known from moral psychology to shift human judgment: consequentialist, emotional, and relational. Evaluating 22 LLMs, we find that nearly all models are context-sensitive, shifting their judgments toward rule-violating behavior. Comparing with a human survey, we find that models and humans are most triggered by different contextual variations, and that a model aligned with human judgments in the base case is not necessarily aligned in its contextual sensitivity. This raises the question of controlling contextual sensitivity, which we address with an activation steering approach that can reliably increase or decrease a model's contextual sensitivity.

Deep reinforcement learning (DRL) has emerged as a powerful paradigm for solving complex decision-making problems. However, DRL-based systems still face significant dependability challenges particularly in real-time environments due to the simulation-to-reality gap, out-of-distribution observations, and the critical impact of latency. Latency-induced faults, in particular, can lead to unsafe or unstable behaviour, yet existing fault-tolerance approaches to DRL systems lack formal methods to rigorously analyse and optimise performance and safety simultaneously in real-time settings. To address this, we propose a formal framework for designing and analysing real-time switching mechanisms between DRL agents and alternative controllers. Our approach leverages Timed Automata (TAs) for explicit switch logic design, which is then syntactically converted to a Markov Decision Process (MDP) for formal analysis. We develop a novel convex query technique for multi-objective model checking, enabling the optimisation of soft performance objectives while ensuring hard safety constraints for MDPs. Furthermore, we present MOPMC, a GPU-accelerated software tool implementing this technique, demonstrating superior scalability in both model size and objective numbers.

Characteristic functions of weighted sums of independent random variables exhibit low-rank structure in the quantized tensor train (QTT) representation, also known as matrix product states (MPS), enabling up to exponential compression of their fully non-Gaussian probability distributions. Under variable independence, the global characteristic function factorizes into local terms. Its low-rank QTT structure arises from intrinsic spectral smoothness in continuous models, or from spectral energy concentration as the number of components $D$ grows in discrete models. We demonstrate this on weighted sums of Bernoulli and lognormal random variables. In the former, despite an adversarial, incompressible small-$D$ regime, the characteristic function undergoes a sharp bond-dimension collapse for $D \gtrsim 300$ components, enabling polylogarithmic time and memory scaling. In the latter, the approach reaches high-resolution discretizations of $N = 2^{30}$ frequency modes on standard hardware, far beyond the $N = 2^{24}$ ceiling of dense implementations. These compressed representations enable efficient computation of Value at Risk (VaR) and Expected Shortfall (ES), supporting applications in quantitative finance and beyond.

Intelligent spectroscopy serves as a pivotal element in AI-driven closed-loop scientific discovery, functioning as the critical bridge between matter structure and artificial intelligence. However, conventional expert-dependent spectral interpretation encounters substantial hurdles, including susceptibility to human bias and error, dependence on limited specialized expertise, and variability across interpreters. To address these challenges, we propose SpecXMaster, an intelligent framework leveraging Agentic Reinforcement Learning (RL) for NMR molecular spectral interpretation. SpecXMaster enables automated extraction of multiplicity information from both 1H and 13C spectra directly from raw FID (free induction decay) data. This end-to-end pipeline enables fully automated interpretation of NMR spectra into chemical structures. It demonstrates superior performance across multiple public NMR interpretation benchmarks and has been refined through iterative evaluations by professional chemical spectroscopists. We believe that SpecXMaster, as a novel methodological paradigm for spectral interpretation, will have a profound impact on the organic chemistry community.

While brain-aligned large language models (LLMs) have garnered attention for their potential as cognitive models and for potential for enhanced safety and trustworthiness in AI, the role of this brain alignment for linguistic competence remains uncertain. In this work, we investigate the functional implications of brain alignment by introducing brain-misaligned models--LLMs intentionally trained to predict brain activity poorly while maintaining high language modeling performance. We evaluate these models on over 200 downstream tasks encompassing diverse linguistic domains, including semantics, syntax, discourse, reasoning, and morphology. By comparing brain-misaligned models with well-matched brain-aligned counterparts, we isolate the specific impact of brain alignment on language understanding. Our experiments reveal that brain misalignment substantially impairs downstream performance, highlighting the critical role of brain alignment in achieving robust linguistic competence. These findings underscore the importance of brain alignment in LLMs and offer novel insights into the relationship between neural representations and linguistic processing.

Autoregressive (AR) models are highly effective for image generation, yet their standard maximum-likelihood estimation training lacks direct optimization for sample quality and diversity. While reinforcement learning (RL) has been used to align diffusion models, these methods typically suffer from output diversity collapse. Similarly, concurrent RL methods for AR models rely strictly on instance-level rewards, often trading off distributional coverage for quality. To address these limitations, we propose a lightweight RL framework that casts token-based AR synthesis as a Markov Decision Process, optimized via Group Relative Policy Optimization (GRPO). Our core contribution is the introduction of a novel distribution-level Leave-One-Out FID (LOO-FID) reward; by leveraging an exponential moving average of feature moments, it explicitly encourages sample diversity and prevents mode collapse during policy updates. We integrate this with composite instance-level rewards (CLIP and HPSv2) for strict semantic and perceptual fidelity, and stabilize the multi-objective learning with an adaptive entropy regularization term. Extensive experiments on LlamaGen and VQGAN architectures demonstrate clear improvements across standard quality and diversity metrics within only a few hundred tuning iterations. The results also show that the model can be updated to produce competitive samples even without Classifier-Free Guidance, and bypass its 2x inference cost.

Vision-Language Models (VLMs) have enabled interpretable medical diagnosis by integrating visual perception with linguistic reasoning. Yet, existing medical chain-of-thought (CoT) models lack explicit mechanisms to represent and enforce causal reasoning, leaving them vulnerable to spurious correlations and limiting their clinical reliability. We pinpoint three core challenges in medical CoT reasoning: how to adaptively trigger causal correction, construct high-quality causal-spurious contrastive samples, and maintain causal consistency across reasoning trajectories. To address these challenges, we propose MedCausalX, an end-to-end framework explicitly models causal reasoning chains in medical VLMs. We first introduce the CRMed dataset providing fine-grained anatomical annotations, structured causal reasoning chains, and counterfactual variants that guide the learning of causal relationships beyond superficial correlations. Building upon CRMed, MedCausalX employs a two-stage adaptive reflection architecture equipped with $\langle$causal$\rangle$ and $\langle$verify$\rangle$ tokens, enabling the model to autonomously determine when and how to perform causal analysis and verification. Finally, a trajectory-level causal correction objective optimized through error-attributed reinforcement learning refines the reasoning chain, allowing the model to distinguish genuine causal dependencies from shortcut associations. Extensive experiments on multiple benchmarks show that MedCausalX consistently outperforms state-of-the-art methods, improving diagnostic consistency by +5.4 points, reducing hallucination by over 10 points, and attaining top spatial grounding IoU, thereby setting a new standard for causally grounded medical reasoning.

Providing reliable predictive maintenance is a critical industrial AI service essential for ensuring the high availability of manufacturing devices. Existing deep-learning methods present competitive results on such tasks but lack a general service-oriented framework to capture complex dependencies in industrial IoT sensor data. While Transformer-based models show strong sequence modeling capabilities, their direct deployment as robust AI services faces significant bottlenecks. Specifically, streaming sensor data collected in real-world service environments often exhibits multi-scale temporal correlations driven by machine working principles. Besides, the datasets available for training time-to-failure predictive services are typically limited in size. These issues pose significant challenges for directly applying existing models as robust predictive services. To address these challenges, we propose MsFormer, a lightweight Multi-scale Transformer designed as a unified AI service model for reliable industrial predictive maintenance. MsFormer incorporates a Multi-scale Sampling (MS) module and a tailored position encoding mechanism to capture sequential correlations across multi-streaming service data. Additionally, to accommodate data-scarce service environments, MsFormer adopts a lightweight attention mechanism with straightforward pooling operations instead of self-attention. Extensive experiments on real-world datasets demonstrate that the proposed framework achieves significant performance improvements over state-of-the-art methods. Furthermore, MsFormer outperforms across industrial devices and operating conditions, demonstrating strong generalizability while maintaining a highly reliable Quality of Service (QoS).

Architectural patterns are frequently found in various software artifacts. The wide variety of patterns and their implementations makes detection challenging with current tools, especially since they often only support detecting patterns in artifacts written in a single language. Large Language Models (LLMs), trained on a diverse range of software artifacts and knowledge, might overcome the limitations of existing approaches. However, their true effectiveness and the factors influencing their performance have not yet been thoroughly examined. To better understand this, we developed MicroPAD. This tool utilizes GPT 5 nano to identify architectural patterns in software artifacts written in any language, based on natural-language pattern descriptions. We used MicroPAD to evaluate an LLM's ability to detect instances of architectural patterns, particularly infrastructure-related microservice patterns. To accomplish this, we selected a set of GitHub repositories and contacted their top contributors to create a new, human-annotated dataset of 190 repositories containing microservice architectural patterns. The results show that MicroPAD was capable of detecting pattern instances across multiple languages and artifact types. The detection performance varied across patterns (F1 scores ranging from 0.09 to 0.70), specifically in relation to their prevalence and the distinctiveness of the artifacts through which they manifest. We also found that patterns associated with recognizable, dominant artifacts were detected more reliably. Whether these findings generalize to other LLMs and tools is a promising direction for future research.

This work establishes rigorous first-of-its-kind upper bounds on the generalization error for the method of approximating solutions to the (d+1)-dimensional incompressible Navier-Stokes equations by training depth-2 neural networks trained via the unsupervised Physics-Informed Neural Network (PINN) framework. This is achieved by bounding the Rademacher complexity of the PINN risk. For appropriately weight bounded net classes our derived generalization bounds do not explicitly depend on the network width and our framework characterizes the generalization gap in terms of the fluid's kinematic viscosity and loss regularization parameters. In particular, the resulting sample complexity bounds are dimension-independent. Our generalization bounds suggest using novel activation functions for solving fluid dynamics. We provide empirical validation of the suggested activation functions and the corresponding bounds on a PINN setup solving the Taylor-Green vortex benchmark.

The task of authorship style transfer involves rewriting text in the style of a target author while preserving the meaning of the original text. Existing style transfer methods train a single model on large corpora to model all target styles at once: this high-cost approach offers limited flexibility for target-specific adaptation, and often sacrifices meaning preservation for style transfer. In this paper, we propose AuthorMix: a lightweight, modular, and interpretable style transfer framework. We train individual, style-specific LoRA adapters on a small set of high-resource authors, allowing the rapid training of specialized adaptation models for each new target via learned, layer-wise adapter mixing, using only a handful of target style training examples. AuthorMix outperforms existing, SoTA style-transfer baselines -- as well as GPT-5.1 -- for low-resource targets, achieving the highest overall score and substantially improving meaning preservation.

We identify a critical security vulnerability in mainstream Claw personal AI agents: untrusted content encountered during heartbeat-driven background execution can silently pollute agent memory and subsequently influence user-facing behavior without the user's awareness. This vulnerability arises from an architectural design shared across the Claw ecosystem: heartbeat background execution runs in the same session as user-facing conversation, so content ingested from any external source monitored in the background (including email, message channels, news feeds, code repositories, and social platforms) can enter the same memory context used for foreground interaction, often with limited user visibility and without clear source provenance. We formalize this process as an Exposure (E) $\rightarrow$ Memory (M) $\rightarrow$ Behavior (B) pathway: misinformation encountered during heartbeat execution enters the agent's short-term session context, potentially gets written into long-term memory, and later shapes downstream user-facing behavior. We instantiate this pathway in an agent-native social setting using MissClaw, a controlled research replica of Moltbook. We find that (1) social credibility cues, especially perceived consensus, are the dominant driver of short-term behavioral influence, with misleading rates up to 61%; (2) routine memory-saving behavior can promote short-term pollution into durable long-term memory at rates up to 91%, with cross-session behavioral influence reaching 76%; (3) under naturalistic browsing with content dilution and context pruning, pollution still crosses session boundaries. Overall, prompt injection is not required: ordinary social misinformation is sufficient to silently shape agent memory and behavior under heartbeat-driven background execution.

Burnout is an occupational syndrome that, like many other professions, affects the majority of software engineers. Past research studies showed important trends, including an increasing use of machine learning techniques to allow for an early detection of burnout. This paper is a systematic literature review (SLR) of the research papers that proposed machine learning (ML) approaches, and focused on detecting burnout in software developers and IT professionals. Our objective is to review the accuracy and precision of the proposed ML techniques, and to formulate recommendations for future researchers interested to replicate or extend those studies. From our SLR we observed that a majority of primary studies focuses on detecting emotions or utilise emotional dimensions to detect or predict the presence of burnout. We also performed a cross-sectional study to detect which ML approach shows a better performance at detecting emotions; and which dataset has more potential and expressivity to capture emotions. We believe that, by identifying which ML tools and datasets show a better performance at detecting emotions, and indirectly at identifying burnout, our paper can be a valuable asset to progress in this important research direction.

Recent advances in game AI, such as AlphaZero and Athénan, have achieved superhuman performance across a wide range of board games. While highly powerful, these agents are ill-suited for human-AI interaction, as they consistently overwhelm human players, offering little enjoyment and limited educational value. This paper addresses the problem of balanced play, in which an agent challenges its opponent without either dominating or conceding. We introduce Minibal (Minimize & Balance), a variant of Minimax specifically designed for balanced play. Building on this concept, we propose several modifications of the Unbounded Minimax algorithm explicitly aimed at discovering balanced strategies. Experiments conducted across seven board games demonstrate that one variant consistently achieves the most balanced play, with average outcomes close to perfect balance. These results establish Minibal as a promising foundation for designing AI agents that are both challenging and engaging, suitable for both entertainment and serious games.

Audio-Language Models (ALMs) are making strides in understanding speech and non-speech audio. However, domain-specialist Foundation Models (FMs) remain the best for closed-ended speech processing tasks such as Speech Emotion Recognition (SER). Using ALMs for Zero-shot SER is a popular choice, but their potential to work with specialists to achieve state-of-the-art (SOTA) performance remains unexplored. We propose ZS-Fuse, a late-fusion method that combines zero-shot emotion estimates from a dual-encoder ALM with specialist FMs. To handle ambiguity in emotions and sensitivity to prompt choice, 1) we use a simple prompt ensemble and 2) suggest a novel technique called prompt amplification, which repeats audio and text queries to discover stronger zero-shot capabilities. We demonstrate the efficacy of our technique by evaluating ZS-Fuse with three dual-encoder ALMs and two FMs, and report improvements over SOTA baselines, such as WavLM-Large, on three speech emotion recognition datasets.

Formal model evaluation methods typically certify that a model satisfies a prescribed target key performance indicator (KPI) level. However, in many applications, the relevant target KPI level may not be known a priori, and the user may instead wish to compare candidate models by analyzing the full trade-offs between performance and reliability achievable at test time by the models. This task, requiring the reliable estimate of the test-time KPI distributions, is made more complicated by the fact that the same data must often be used both to pre-select a subset of candidate models and to estimate their KPI distributions, causing a potential post-selection bias. In this work, we introduce post-selection distributional model evaluation (PS-DME), a general framework for statistically valid distributional model assessment after arbitrary data-dependent model pre-selection. Building on e-values, PS-DME controls post-selection false coverage rate (FCR) for the distributional KPI estimates and is proved to be more sample efficient than a baseline method based on sample splitting. Experiments on synthetic data, text-to-SQL decoding with large language models, and telecom network performance evaluation demonstrate that PS-DME enables reliable comparison of candidate configurations across a range of reliability levels, supporting the statistically reliable exploration of performance--reliability trade-offs.

Flaky failure triage is crucial for keeping distributed database continuous integration (CI) efficient and reliable. After a failure is observed, operators must quickly decide whether to auto-rerun the job as likely flaky or escalate it as likely persistent, often under CPU-only millisecond budgets. Existing approaches remain difficult to deploy in this setting because they may rely on post-failure artifacts, produce poorly calibrated scores under telemetry and workload shifts, or learn from labels generated by finite rerun policies. To address these challenges, we present SCOUT, a practical state-aware causal online uncertainty-calibrated triage framework for distributed database CI. SCOUT uses only strict-causal features, including pre-failure telemetry and strictly historical data, to make online decisions without lookahead. Specifically, SCOUT combines lightweight state-aware scoring with optional sparse metadata fusion, applies post-hoc calibration to support fixed-threshold decisions across temporal and cross-domain shifts, and introduces a posterior-soft correction to reduce label bias induced by finite rerun budgets. We evaluated SCOUT on a benchmark of 3,680 labeled failed runs, including 462 flaky positives, and 62 telemetry/context features. Further, we studied the feasibility of SCOUT on TiDB v7/v8 and a large GitHub Actions metadata-only trace. The experimental results demonstrated its effectiveness and usefulness. We deployed SCOUT in the production environment, achieving an end-to-end P95 latency of 1.17 ms on CPU.

A tremendous number of critical database systems lack adequate documentation. Declared primary keys are absent, foreign key constraints have been dropped for performance, column names are cryptic abbreviations, and no entity-relationship diagrams exist. We present DBAutoDoc, a system that automates the discovery and documentation of undocumented relational database schemas by combining statistical data analysis with iterative large language model (LLM) refinement. DBAutoDoc's central insight is that schema understanding is fundamentally an iterative, graph-structured problem. Drawing structural inspiration from backpropagation in neural networks, DBAutoDoc propagates semantic corrections through schema dependency graphs across multiple refinement iterations until descriptions converge. This propagation is discrete and semantic rather than mathematical, but the structural analogy is precise: early iterations produce rough descriptions akin to random initialization, and successive passes sharpen the global picture as context flows through the graph. The system makes four concrete contributions detailed in the paper. On a suite of benchmark databases, DBAutoDoc achieved overall weighted scores of 96.1% across two model families (Google's Gemini and Anthropic's Claude) using a composite metric. Ablation analysis demonstrates that the deterministic pipeline contributes a 23-point F1 improvement over LLM-only FK detection, confirming that the system's contribution is substantial and independent of LLM pre-training knowledge. DBAutoDoc is released as open-source software with all evaluation configurations and prompt templates included for full reproducibility.

Retrieval-Augmented Generation (RAG) systems enhance the performance of large language models (LLMs) by incorporating supplementary retrieved documents, enabling more accurate and context-aware responses. However, integrating these external documents often results in very long input sequences, which significantly increases computation costs during the prefill stage, where key-value (KV) representations for all input tokens are generated. This latency bottleneck becomes especially pronounced under high-throughput serving scenarios. KV-cache reuse offers a promising solution by storing previously computed KV states for shared input prefixes, thereby avoiding redundant computation across requests that contain overlapping context. Yet, the effectiveness of cache reuse is often limited by three practical challenges: low cache hit rates due to naive eviction policies, high CPU-GPU data transfer overhead, and slow SSD I/O when caches spill to storage. To address these issues, we propose PCR, a system designed to maximize KV-cache reuse efficiency through intelligent prefetching and pipelined data movement. Specifically, PCR introduces three key techniques: (1) a prefix-tree caching structure with a look-ahead LRU replacement policy that uses pending requests in the scheduler queue to improve cache hit ratios; (2) layer-wise overlapping that pipelines KV-cache loading and GPU computation across CUDA streams to hide communication latency; and (3) queue-based prefetching that proactively loads relevant KV caches from SSD into DRAM before they are needed. Extensive experiments show that PCR outperforms existing KV-cache reuse methods, achieving up to a 2.47x speedup in terms of average TTFT.

Self-supervised learning (SSL) has advanced speech processing. However, existing speech SSL methods typically assume a single sampling rate and struggle with mixed-rate data due to temporal resolution mismatch. To address this limitation, we propose MSRHuBERT, a multi-sampling-rate adaptive pre-training method. Building on HuBERT, we replace its single-rate downsampling CNN with a multi-sampling-rate adaptive downsampling CNN that maps raw waveforms from different sampling rates to a shared temporal resolution without resampling. This design enables unified mixed-rate pre-training and fine-tuning. In experiments spanning 16 to 48 kHz, MSRHuBERT outperforms HuBERT on speech recognition and full-band speech reconstruction, preserving high-frequency detail while modeling low-frequency semantic structure. Moreover, MSRHuBERT retains HuBERT's mask-prediction objective and Transformer encoder, so existing analyses and improvements that were developed for HuBERT can apply directly.

Retrieval-Augmented Generation (RAG) fine-tuning has shown substantial improvements over vanilla RAG, yet most studies target document question answering and often rely on standard NLP metrics that can obscure factual differences. We evaluate RAG fine-tuning for long-form text generation in electronic design automation, adapting a 7B model under five context augmentation strategies with varying retrieval conditions. We introduce TriFEX, a human-validated, triple-based evaluation pipeline that attributes generated claims to their origin-user query, context and reference-and propose Parametric Knowledge Precision (PKP), which isolates internalized knowledge by filtering out claims leaked in the prompt. We show that ROUGE and BERTScore fail to detect factual differences that our triple-based evaluation reveals. Additionally, we demonstrate that an existing metric for knowledge internalization is retrieva-sensitive, with about 75% of its cross-condition variance driven by changes in the rate at which internal knowledge is expressed (PR), rather than by changes in its actual correctness (PKP). The fine-tuned 7B variants outperform a 72B baseline on most metrics, further showing generalization across conditions and on a related benchmark. These results underscore the limitations of available metrics in RAG evaluation and show that smaller models could be reasonably well adapted to specialized tasks for cost-efficient, on-premises deployment.

The accelerating pace of climate change introduces profound non-stationarities that challenge the ability of Machine Learning based climate emulators to generalize beyond their training distributions. While these emulators offer computationally efficient alternatives to traditional Earth System Models, their reliability remains a potential bottleneck under "no-analog" future climate states, which we define here as regimes where external forcing drives the system into conditions outside the empirical range of the historical training data. A fundamental challenge in evaluating this reliability is data contamination; because many models are trained on simulations that already encompass future scenarios, true out-of-distribution (OOD) performance is often masked. To address this, we benchmark the OOD robustness of three state-of-the-art architectures: U-Net, ConvLSTM, and the ClimaX foundation model specifically restricted to a historical-only training regime (1850-2014). We evaluate these models using two complementary strategies: (i) temporal extrapolation to the recent climate (2015-2023) and (ii) cross-scenario forcing shifts across divergent emission pathways. Our analysis within this experimental setup reveals an accuracy vs. stability trade-off: while the ClimaX foundation model achieves the lowest absolute error, it exhibits higher relative performance changes under distribution shifts, with precipitation errors increasing by up to 8.44% under extreme forcing scenarios. These findings suggest that when restricted to historical training dynamics, even high-capacity foundation models are sensitive to external forcing trajectories. Our results underscore the necessity of scenario-aware training and rigorous OOD evaluation protocols to ensure the robustness of climate emulators under a changing climate.

Currently, a central challenge and bottleneck in the deployment and validation of computer-aided diagnosis (CAD) models within the field of medical imaging is data scarcity. For lung cancer, one of the most prevalent types worldwide, limited datasets can delay diagnosis and have an impact on patient outcome. Generative AI offers a promising solution for this issue, but dealing with the complex distribution of full Hounsfield Unit (HU) range lung CT scans is challenging and remains as a highly computationally demanding task. This paper introduces a novel decomposition strategy that synthesizes CT images one HU interval at a time, rather than modelling the entire HU domain at once. This framework focuses on training generative architectures on individual tissue-focused HU windows, then merges their output into a full-range scan via a learned reconstruction network that effectively reverses the HU-windowing process. We further propose multi-head and multi-decoder models to better capture textures while preserving anatomical consistency, with a multi-head VQVAE achieving the best performance for the generative task. Quantitative evaluation shows this approach significantly outperforms conventional 2D full-range baselines, achieving a 6.2% improvement in FID and superior MMD, Precision, and Recall across all HU intervals. The best performance is achieved by a multi-head VQVAE variant, demonstrating that it is possible to enhance visual fidelity and variability while also reducing model complexity and computational cost. This work establishes a new paradigm for structure-aware medical image synthesis, aligning generative modelling with clinical interpretation.

The interpretable object detection capabilities of a novel Kolmogorov-Arnold network framework are examined here. The approach refers to a key limitation in computer vision for autonomous vehicles perception, and beyond. These systems offer limited transparency regarding the reliability of their confidence scores in visually degraded or ambiguous scenes. To address this limitation, a Kolmogorov-Arnold network is employed as an interpretable post-hoc surrogate to model the trustworthiness of the You Only Look Once (Yolov10) detections using seven geometric and semantic features. The additive spline-based structure of the Kolmogorov-Arnold network enables direct visualisation of each feature's influence. This produces smooth and transparent functional mappings that reveal when the model's confidence is well supported and when it is unreliable. Experiments on both Common Objects in Context (COCO), and images from the University of Bath campus demonstrate that the framework accurately identifies low-trust predictions under blur, occlusion, or low texture. This provides actionable insights for filtering, review, or downstream risk mitigation. Furthermore, a bootstrapped language-image (BLIP) foundation model generates descriptive captions of each scene. This tool enables a lightweight multimodal interface without affecting the interpretability layer. The resulting system delivers interpretable object detection with trustworthy confidence estimates. It offers a powerful tool for transparent and practical perception component for autonomous and multimodal artificial intelligence applications.

A sliding-window inference strategy is commonly adopted in recent training-free open-vocabulary semantic segmentation methods to overcome limitation of the CLIP in processing high-resolution images. However, this approach introduces a new challenge: each window is processed independently, leading to semantic discrepancy across windows. To address this issue, we propose Global-Local Aligned CLIP~(GLA-CLIP), a framework that facilitates comprehensive information exchange across windows. Rather than limiting attention to tokens within individual windows, GLA-CLIP extends key-value tokens to incorporate contextual cues from all windows. Nevertheless, we observe a window bias: outer-window tokens are less likely to be attended, since query features are produced through interactions within the inner window patches, thereby lacking semantic grounding beyond their local context. To mitigate this, we introduce a proxy anchor, constructed by aggregating tokens highly similar to the given query from all windows, which provides a unified semantic reference for measuring similarity across both inner- and outer-window patches. Furthermore, we propose a dynamic normalization scheme that adjusts attention strength according to object scale by dynamically scaling and thresholding the attention map to cope with small-object scenarios. Moreover, GLA-CLIP can be equipped on existing methods and broad their receptive field. Extensive experiments validate the effectiveness of GLA-CLIP in enhancing training-free open-vocabulary semantic segmentation performance. Code is available at https://github.com/2btlFe/GLA-CLIP.

Deep learning models for flood and wildfire segmentation and object detection enable precise, real-time disaster localization when deployed on embedded drone platforms. However, in natural disaster management, the lack of transparency in their decision-making process hinders human trust required for emergency response. To address this, we present an explainability framework for understanding flood segmentation and car detection predictions on the widely used PIDNet and YOLO architectures. More specifically, we introduce a novel redistribution strategy that extends Layer-wise Relevance Propagation (LRP) explanations for sigmoid-gated element-wise fusion layers. This extension allows LRP relevances to flow through the fusion modules of PIDNet, covering the entire computation graph back to the input image. Furthermore, we apply Prototypical Concept-based Explanations (PCX) to provide both local and global explanations at the concept level, revealing which learned features drive the segmentation and detection of specific disaster semantic classes. Experiments on a publicly available flood dataset show that our framework provides reliable and interpretable explanations while maintaining near real-time inference capabilities, rendering it suitable for deployment on resource-constrained platforms, such as Unmanned Aerial Vehicles (UAVs).

Tabular data is more challenging to generate than text and images, due to its heterogeneous features and much lower sample sizes. On this task, diffusion-based models are the current state-of-the-art (SotA) model class, achieving almost perfect performance on commonly used benchmarks. In this paper, we question the perception of progress for tabular data generation. First, we highlight the limitations of current protocols to evaluate the fidelity of generated data, and advocate for alternative ones. Next, we revisit a simple baseline -- hierarchical mixture models in the form of deep probabilistic circuits (PCs) -- which delivers competitive or superior performance to SotA models for a fraction of the cost. PCs are the generative counterpart of decision forests, and as such can natively handle heterogeneous data as well as deliver tractable probabilistic generation and inference. Finally, in a rigorous empirical analysis we show that the apparent saturation of progress for SotA models is largely due to the use of inadequate metrics. As such, we highlight that there is still much to be done to generate realistic tabular data. Code available at https://github.com/april-tools/tabpc.

Production AI agents frequently receive user-specific queries that are highly repetitive, with up to 47\% being semantically similar to prior interactions, yet each query is typically processed with the same computational cost. We argue that this redundancy can be exploited through conversational memory, transforming repetition from a cost burden into an efficiency advantage. We propose a memory-augmented inference framework in which a lightweight 8B-parameter model leverages retrieved conversational context to answer all queries via a low-cost inference path. Without any additional training or labeled data, this approach achieves 30.5\% F1, recovering 69\% of the performance of a full-context 235B model while reducing effective cost by 96\%. Notably, a 235B model without memory (13.7\% F1) underperforms even the standalone 8B model (15.4\% F1), indicating that for user-specific queries, access to relevant knowledge outweighs model scale. We further analyze the role of routing and confidence. At practical confidence thresholds, routing alone already directs 96\% of queries to the small model, but yields poor accuracy (13.0\% F1) due to confident hallucinations. Memory does not substantially alter routing decisions; instead, it improves correctness by grounding responses in retrieved user-specific information. As conversational memory accumulates over time, coverage of recurring topics increases, further narrowing the performance gap. We evaluate on 152 LoCoMo questions (Qwen3-8B/235B) and 500 LongMemEval questions. Incorporating hybrid retrieval (BM25 + cosine similarity) improves performance by an additional +7.7 F1, demonstrating that retrieval quality directly enhances end-to-end system performance. Overall, our results highlight that memory, rather than model size, is the primary driver of accuracy and efficiency in persistent AI agents.

The rapid adoption of mobile graphical user interface (GUI) agents, which autonomously control applications and operating systems (OS), exposes new system-level attack surfaces. Existing backdoors against web GUI agents and general GenAI models rely on environmental injection or deceptive pop-ups to mislead the agent operation. However, these techniques do not work on screenshots-based mobile GUI agents due to the challenges of restricted trigger design spaces, OS background interference, and conflicts in multiple trigger-action mappings. We propose AgentRAE, a novel backdoor attack capable of inducing Remote Action Execution in mobile GUI agents using visually natural triggers (e.g., benign app icons in notifications). To address the underfitting caused by natural triggers and achieve accurate multi-target action redirection, we design a novel two-stage pipeline that first enhances the agent's sensitivity to subtle iconographic differences via contrastive learning, and then associates each trigger with a specific mobile GUI agent action through a backdoor post-training. Our extensive evaluation reveals that the proposed backdoor preserves clean performance with an attack success rate of over 90% across ten mobile operations. Furthermore, it is hard to visibly detect the benign-looking triggers and circumvents eight representative state-of-the-art defenses. These results expose an overlooked backdoor vector in mobile GUI agents, underscoring the need for defenses that scrutinize notification-conditioned behaviors and internal agent representations.

Large Language Models (LLMs) have demonstrated great capabilities across diverse natural language tasks; yet their ability to solve abstraction and optimization problems with constraints remains scarcely explored. In this paper, we investigate whether LLMs can reason and optimize under the physical and operational constraints of Optimal Power Flow (OPF) problem. We introduce a challenging evaluation setup that requires a set of fundamental skills such as reasoning, structured input handling, arithmetic, and constrained optimization. Our evaluation reveals that SoTA LLMs fail in most of the tasks, and that reasoning LLMs still fail in the most complex settings. Our findings highlight critical gaps in LLMs' ability to handle structured reasoning under constraints, and this work provides a rigorous testing environment for developing more capable LLM assistants that can tackle real-world power grid optimization problems.

The comparison of dental records is a standardized technique in forensic dentistry used to speed up the identification of individuals in multiple-comparison scenarios. Specifically, the odontogram comparison is a procedure to compute criteria that will be used to perform a ranking. State-of-the-art automatic methods either make use of simple techniques, without utilizing the full potential of the information obtained from a comparison, or their internal behavior is not known due to the lack of peer-reviewed publications. This work aims to design aggregation mechanisms to automatically compare pairs of dental records that can be understood and validated by experts, improving the current methods. To do so, we introduce different aggregation approaches using the state-of-the-art codification, based on seven different criteria. In particular, we study the performance of i) data-driven lexicographical order-based aggregations, ii) well-known fuzzy logic aggregation methods and iii) machine learning techniques as aggregation mechanisms. To validate our proposals, 215 forensic cases from two different populations have been used. The results obtained show how the use of white-box machine learning techniques as aggregation models (average ranking from 2.02 to 2.21) are able to improve the state-of-the-art (average ranking of 3.91) without compromising the explainability and interpretability of the method.

Recent advances in visual language models have enabled autonomous agents for complex reasoning, tool use, and document understanding. However, existing document agents mainly transform papers into static artifacts such as summaries, webpages, or slides, which are insufficient for technical papers involving dynamic mechanisms and state transitions. In this work, we propose a Paper-to-Interactive-System Agent that converts research papers into executable interactive web systems. Given a PDF paper, the agent performs end-to-end processing without human intervention, including paper understanding, system modeling, and interactive webpage synthesis, enabling users to manipulate inputs and observe dynamic behaviors. To evaluate this task, we introduce a benchmark of 19 research papers paired with expert-built interactive systems as ground truth. We further propose PaperVoyager, a structured generation framework that explicitly models mechanisms and interaction logic during synthesis. Experiments show that PaperVoyager significantly improves the quality of generated interactive systems, offering a new paradigm for interactive scientific paper understanding.

We consider two approaches for assessing the reliability of the individual predictions of a classifier: Robustness Quantification (RQ) and Uncertainty Quantification (UQ). We explain the conceptual differences between the two approaches, compare both approaches on a number of benchmark datasets and show that RQ is capable of outperforming UQ, both in a standard setting and in the presence of distribution shift. Beside showing that RQ can be competitive with UQ, we also demonstrate the complementarity of RQ and UQ by showing that a combination of both approaches can lead to even better reliability assessments.

Membership inference attacks (MIAs) aim to determine whether a data sample was included in a machine learning (ML) model's training set and have become the de facto standard for measuring privacy leakages in ML. We propose an evaluation framework that defines the conditions under which MIAs constitute a genuine privacy threat, and review representative MIAs against it. We find that, under the realistic conditions defined in our framework, MIAs represent weak privacy threats. Thus, relying on them as a privacy metric in ML can lead to an overestimation of risk and to unnecessary sacrifices in model utility as a consequence of employing too strong defenses.

Current multimodal toxicity benchmarks typically use a single binary hatefulness label. This coarse approach conflates two fundamentally different characteristics of expression: tone and content. Drawing on communication science theory, we introduce a fine-grained annotation scheme that distinguishes two separable dimensions: incivility (rude or dismissive tone) and intolerance (content that attacks pluralism and targets groups or identities) and apply it to 2,030 memes from the Hateful Memes dataset. We evaluate different vision-language models under coarse-label training, transfer learning across label schemes and a joint learning approach that combines the coarse hatefulness label with our fine-grained annotations. Our results show that fine-grained annotations complement existing coarse labels and, when used jointly, improve overall model performance. Moreover, models trained with the fine-grained scheme exhibit more balanced moderation-relevant error profiles and are less prone to under-detection of harmful content than models trained on hatefulness labels alone (FNR-FPR, the difference between false negative and false positive rates: 0.74 to 0.42 for LLaVA-1.6-Mistral-7B; 0.54 to 0.28 for Qwen2.5-VL-7B). This work contributes to data-centric approaches in content moderation by improving the reliability and accuracy of moderation systems through enhanced data quality. Overall, combining both coarse and fine-grained labels provides a practical route to more reliable multimodal moderation.

Graph foundation models (GFMs) have recently attracted interest due to the promise of graph neural network (GNN) architectures that generalize zero-shot across graphs of arbitrary scales, feature dimensions, and domains. While existing work has demonstrated this ability empirically across diverse real-world benchmarks, these tasks share a crucial hidden limitation: they admit a narrow set of effective GNN architectures. In particular, current domain-agnostic GFMs rely on fixed architectural backbones, implicitly assuming that a single message-passing regime suffices across tasks. In this paper, we argue that architecture adaptivity is a necessary requirement for true GFMs. We show that existing approaches are non-robust to task-dependent architectural attributes and, as a case study, use range as a minimal and measurable axis along which this limitation becomes explicit. With theoretical analysis and controlled synthetic experiments, we demonstrate that fixed-backbone GFMs provably under-reach on tasks whose architectural requirements differ from those seen at training time. To address this issue, we introduce a framework that adapts effective GNN architecture at inference time by discovering and mixing task-specific linear graph operators, enabling zero-shot generalization across tasks with heterogeneous architectural requirements, without retraining. We validate our approach on arbitrary-range synthetic tasks and a suite of real-world benchmarks, demonstrating improved performance and robustness over existing domain-agnostic GFMs.

In the maintenance of complex systems, fault trees are used to locate problems and provide targeted solutions. To enable fault trees stored as images to be directly processed by large language models, which can assist in tracking and analyzing malfunctions, we propose a novel textual representation of fault trees. Building on it, we construct a benchmark for multi-turn dialogue systems that emphasizes robust interaction in complex environments, evaluating a model's ability to assist in malfunction localization, which contains $3130$ entries and $40.75$ turns per entry on average. We train an end-to-end model to generate vague information to reflect user behavior and introduce long-range rollback and recovery procedures to simulate user error scenarios, enabling assessment of a model's integrated capabilities in task tracking and error recovery, and Gemini 2.5 pro archives the best performance.

Dari, the primary language of Afghanistan, is spoken by tens of millions of people yet remains largely absent from the misinformation detection literature. We address this gap with DariMis, the first manually annotated dataset of 9,224 Dari-language YouTube videos, labeled across two dimensions: Information Type (Misinformation, Partly True, True) and Harm Level (Low, Medium, High). A central empirical finding is that these dimensions are structurally coupled, not independent: 55.9 percent of Misinformation carries at least Medium harm potential, compared with only 1.0 percent of True content. This enables Information Type classifiers to function as implicit harm-triage filters in content moderation pipelines. We further propose a pair-input encoding strategy that represents the video title and description as separate BERT segment inputs, explicitly modeling the semantic relationship between headline claims and body content, a key signal of misleading information. An ablation study against single-field concatenation shows that pair-input encoding yields a 7.0 percentage point gain in Misinformation recall (60.1 percent to 67.1 percent), the safety-critical minority class, despite modest overall macro F1 differences (0.09 percentage points). We benchmark a Dari/Farsi-specialized model (ParsBERT) against XLM-RoBERTa-base; ParsBERT achieves the best test performance with accuracy of 76.60 percent and macro F1 of 72.77 percent. Bootstrap 95 percent confidence intervals are reported for all metrics, and we discuss both the practical significance and statistical limitations of the results.

Computational methods applied to legal scholarship hold the promise of analyzing law at scale. We start from a simple question: how often do courts implicitly apply statutory rules? This requires distinguishing legal reasoning from semantic similarity. We focus on implicit citation of the French Civil Code in first-instance court decisions and introduce a benchmark of 1,015 passage-article pairs annotated by three legal experts. We show that expert disagreement predicts model failures. Inter-annotator agreement is moderate ($κ$ = 0.33) with 43% of disagreements involving the boundary between factual description and legal reasoning. Our supervised ensemble achieves F1 = 0.70 (77% accuracy), but this figure conceals an asymmetry: 68% of false positives fall on the 33% of cases where the annotators disagreed. Despite these limits, reframing the task as top-k ranking and leveraging multi-model consensus yields 76% precision at k = 200 in an unsupervised setting. Moreover, the remaining false positives tend to surface legally ambiguous applications rather than obvious errors.

The growing use of large language models has increased interest in sharing textual data in a privacy-preserving manner. One prominent line of work addresses this challenge through text rewriting under Local Differential Privacy (LDP), where input texts are locally obfuscated before release with formal privacy guarantees. These guarantees are typically expressed by a parameter $\varepsilon$ that upper bounds the worst-case privacy loss. However, nominal $\varepsilon$ values are often difficult to interpret and compare across mechanisms. In this work, we investigate how to empirically calibrate across text rewriting mechanisms under LDP. We propose TeDA, which formulates calibration via a hypothesis-testing framework that instantiates text distinguishability audits in both surface and embedding spaces, enabling empirical assessment of indistinguishability from privatized texts. Applying this calibration to several representative mechanisms, we demonstrate that similar nominal $\varepsilon$ bounds can imply very different levels of distinguishability. Empirical calibration thus provides a more comparable footing for evaluating privacy-utility trade-offs, as well as a practical tool for mechanism comparison and analysis in real-world LDP text rewriting deployments.

Large language models (LLMs) inherently operate over a large generation space, yet conventional usage typically reports the most likely generation (MLG) as a point prediction, which underestimates the model's capability: although the top-ranked response can be incorrect, valid answers may still exist within the broader output space and can potentially be discovered through repeated sampling. This observation motivates moving from point prediction to set-valued prediction, where the model produces a set of candidate responses rather than a single MLG. In this paper, we propose a principled framework for set-valued prediction, which provides feasibility-aware coverage guarantees. We show that, given the finite-sampling nature of LLM generation, coverage is not always achievable: even with multiple samplings, LLMs may fail to yield an acceptable response for certain questions within the sampled candidate set. To address this, we establish a minimum achievable risk level (MRL), below which statistical coverage guarantees cannot be satisfied. Building on this insight, we then develop a data-driven calibration procedure that constructs prediction sets from sampled responses by estimating a rigorous threshold, ensuring that the resulting set contains a correct answer with a desired probability whenever the target risk level is feasible. Extensive experiments on six language generation tasks with five LLMs demonstrate both the statistical validity and the predictive efficiency of our framework.

Generalization is a central concept in machine learning theory, yet for quantum models, it is predominantly analyzed through uniform bounds that depend on a model's overall capacity rather than the specific function learned. These capacity-based uniform bounds are often too loose and entirely insensitive to the actual training and learning process. Previous theoretical guarantees have failed to provide non-uniform, data-dependent bounds that reflect the specific properties of the learned solution rather than the worst-case behavior of the entire hypothesis class. To address this limitation, we derive the first PAC-Bayesian generalization bounds for a broad class of quantum models by analyzing layered circuits composed of general quantum channels, which include dissipative operations such as mid-circuit measurements and feedforward. Through a channel perturbation analysis, we establish non-uniform bounds that depend on the norms of learned parameter matrices; we extend these results to symmetry-constrained equivariant quantum models; and we validate our theoretical framework with numerical experiments. This work provides actionable model design insights and establishes a foundational tool for a more nuanced understanding of generalization in quantum machine learning.

We study the theoretical behavior of denoising score matching--the learning task associated to diffusion models--when the data distribution is supported on a low-dimensional manifold and the score is parameterized using a random feature neural network. We derive asymptotically exact expressions for the test, train, and score errors in the high-dimensional limit. Our analysis reveals that, for linear manifolds the sample complexity required to learn the score function scales linearly with the intrinsic dimension of the manifold, rather than with the ambient dimension. Perhaps surprisingly, the benefits of low-dimensional structure starts to diminish once we have a non-linear manifold. These results indicate that diffusion models can benefit from structured data; however, the dependence on the specific type of structure is subtle and intricate.

We propose stepwise variational inference (VI) with vine copulas: a universal VI procedure that combines vine copulas with a novel stepwise estimation procedure of the variational parameters. Vine copulas consist of a nested sequence of trees built from copulas, where more complex latent dependence can be modeled with increasing number of trees. We propose to estimate the vine copula approximate posterior in a stepwise fashion, tree by tree along the vine structure. Further, we show that the usual backward Kullback-Leibler divergence cannot recover the correct parameters in the vine copula model, thus the evidence lower bound is defined based on the Rényi divergence. Finally, an intuitive stopping criterion for adding further trees to the vine eliminates the need to pre-define a complexity parameter of the variational distribution, as required for most other approaches. Thus, our method interpolates between mean-field VI (MFVI) and full latent dependence. In many applications, in particular sparse Gaussian processes, our method is parsimonious with parameters, while outperforming MFVI.

Electronic health records (EHRs) and other real-world clinical data are essential for clinical research, medical artificial intelligence, and life science, but their sharing is severely limited by privacy, governance, and interoperability constraints. These barriers create persistent data silos that hinder multi-center studies, large-scale model development, and broader biomedical discovery. Existing privacy-preserving approaches, including multi-party computation and related cryptographic techniques, provide strong protection but often introduce substantial computational overhead, reducing the efficiency of large-scale machine learning and foundation-model training. In addition, many such methods make data usable for restricted computation while leaving them effectively invisible to clinicians and researchers, limiting their value in workflows that still require direct inspection, exploratory analysis, and human interpretation. We propose a real-world-data transformation framework for privacy-preserving sharing of structured clinical records. Instead of converting data into opaque representations, our approach constructs transformed numeric views that preserve medical semantics and major statistical properties while, under a clearly specified threat model, provably breaking direct linkage between those views and protected patient-level attributes. Through collaboration between computer scientists and the AI agent \textbf{SciencePal}, acting as a constrained tool inventor under human guidance, we design three transformation operators that are non-reversible within this threat model, together with an additional mixing strategy for high-risk scenarios, supported by theoretical analysis and empirical evaluation under reconstruction, record linkage, membership inference, and attribute inference attacks.

Discovering governing Partial Differential Equations (PDEs) from sparse and noisy data is a challenging issue in data-driven scientific computing. Conventional sparse regression methods often suffer from two major limitations: (i) the instability of numerical differentiation under sparse and noisy data, and (ii) the restricted flexibility of a pre-defined candidate library. We propose Weak-PDE-Net, an end-to-end differentiable framework that can robustly identify open-form PDEs. Weak-PDE-Net consists of two interconnected modules: a forward response learner and a weak-form PDE generator. The learner embeds learnable Gaussian kernels within a lightweight MLP, serving as a surrogate model that adaptively captures system dynamics from sparse observations. Meanwhile, the generator integrates a symbolic network with an integral module to construct weak-form PDEs, avoiding explicit numerical differentiation and improving robustness to noise. To relax the constraints of the pre-defined library, we leverage Differentiable Neural Architecture Search strategy during training to explore the functional space, which enables the efficient discovery of open-form PDEs. The capability of Weak-PDE-Net in multivariable systems discovery is further enhanced by incorporating Galilean Invariance constraints and symmetry equivariance hypotheses to ensure physical consistency. Experiments on several challenging PDE benchmarks demonstrate that Weak-PDE-Net accurately recovers governing equations, even under highly sparse and noisy observations.

Personalized text-to-image generation lets users fine-tune diffusion models into repositories of concept-specific checkpoints, but serving these repositories efficiently is difficult for two reasons: natural-language requests are often ambiguous and can be misrouted to visually similar checkpoints, and standard post-training quantization can distort the fragile representations that encode personalized concepts. We present PersonalQ, a unified framework that connects checkpoint selection and quantization through a shared signal -- the checkpoint's trigger token. Check-in performs intent-aligned selection by combining intent-aware hybrid retrieval with LLM-based reranking over checkpoint context and asks a brief clarification question only when multiple intents remain plausible; it then rewrites the prompt by inserting the selected checkpoint's canonical trigger. Complementing this, Trigger-Aware Quantization (TAQ) applies trigger-aware mixed precision in cross-attention, preserving trigger-conditioned key/value rows (and their attention weights) while aggressively quantizing the remaining pathways for memory-efficient inference. Experiments show that PersonalQ improves intent alignment over retrieval and reranking baselines, while TAQ consistently offers a stronger compression-quality trade-off than prior diffusion PTQ methods, enabling scalable serving of personalized checkpoints without sacrificing fidelity.

Translating Natural Language to SQL (NL2SQL) remains a critical bottleneck for democratization of data in enterprises. Although Large Language Models (LLMs) like Gemini 2.5 and other LLMs have demonstrated impressive zero-shot capabilities, their high inference costs limit deployment at scale. This paper explores the efficacy of fine-tuning both large and small language models on NL2SQL tasks. Our research reveals a counter-intuitive scaling phenomenon. Fine-tuning large models (Gemini 2.5 Flash/Lite) on standard datasets yields negligible returns, often leading to overfitting on complex queries. Conversely, small models (Qwen) show significant gains. Fine-tuning improved the small model baseline from 36% to 45%, and further enriching the dataset with explicit Chain-of-Thought (CoT) reasoning surged accuracy to 54.5%(Fig 2). While this is still lower than the accuracy of large models like Gemini 2.5 , it does serve the business goal of significant cost reduction, latency in inference time and also meeting the business critical performance accuracy threshold.This paper demonstrates that transferring reasoning patterns enables compute-efficient smaller models to approach production-grade performance.

Expert eye movements provide a rich, passive source of domain knowledge in radiology, offering a powerful cue for integrating diagnostic reasoning into computer-aided analysis. However, direct integration into CNN-based systems, which historically have dominated the medical image analysis domain, is challenging: gaze recordings are sequential, temporally dense yet spatially sparse, noisy, and variable across experts. As a consequence, most existing image-based models utilize reduced representations such as heatmaps. In contrast, gaze naturally aligns with transformer architectures, as both are sequential in nature and rely on attention to highlight relevant input regions. In this work, we introduce FixationFormer, a transformer-based architecture that represents expert gaze trajectories as sequences of tokens, thereby preserving their temporal and spatial structure. By modeling gaze sequences jointly with image features, our approach addresses sparsity and variability in gaze data while enabling a more direct and fine-grained integration of expert diagnostic cues through explicit cross-attention between the image and gaze token sequences. We evaluate our method on three publicly available benchmark chest X-ray datasets and demonstrate that it achieves state-of-the-art classification performance, highlighting the value of representing gaze as a sequence in transformer-based medical image analysis.

Chest X-ray report generation and automated evaluation are limited by poor recognition of low-prevalence abnormalities and inadequate handling of clinically important language, including negation and ambiguity. We develop a clinician-guided framework combining human expertise and large language models for multi-label finding extraction from free-text chest X-ray reports and use it to define Ran Score, a finding-level metric for report evaluation. Using three non-overlapping MIMIC-CXR-EN cohorts from a public chest X-ray dataset and an independent ChestX-CN validation cohort, we optimize prompts, establish radiologist-derived reference labels and evaluate report generation models. The optimized framework improves the macro-averaged score from 0.753 to 0.956 on the MIMIC-CXR-EN development cohort, exceeds the CheXbert benchmark by 15.7 percentage points on directly comparable labels, and shows robust generalization on the ChestX-CN validation cohort. Here we show that clinician-guided prompt optimization improves agreement with a radiologist-derived reference standard and that Ran Score enables finding-level evaluation of report fidelity, particularly for low-prevalence abnormalities.

Retrieval-Augmented Generation (RAG) improves the reliability of large language model applications by grounding generation in retrieved evidence, but it also introduces a new attack surface: corpus poisoning. In this setting, an adversary injects or edits passages so that they are ranked into the Top-$K$ results for target queries and then affect downstream generation. Existing defences against corpus poisoning often rely on content filtering, auxiliary models, or generator-side reasoning, which can make deployment more difficult. We propose ProGRank, a post hoc, training-free retriever-side defence for dense-retriever RAG. ProGRank stress-tests each query--passage pair under mild randomized perturbations and extracts probe gradients from a small fixed parameter subset of the retriever. From these signals, it derives two instability signals, representational consistency and dispersion risk, and combines them with a score gate in a reranking step. ProGRank preserves the original passage content, requires no retraining, and also supports a surrogate-based variant when the deployed retriever is unavailable. Extensive experiments across three datasets, three dense retriever backbones, representative corpus poisoning attacks, and both retrieval-stage and end-to-end settings show that ProGRank provides stronger defence performance and a favorable robustness--utility trade-off. It also remains competitive under adaptive evasive attacks.

Existing dialogue systems rely on Query Suggestion (QS) to enhance user engagement. Recent efforts typically employ large language models with Click-Through Rate (CTR) model, yet fail in cold-start scenarios due to their heavy reliance on abundant online click data for effective CTR model training. To bridge this gap, we propose Cold-EQS, an iterative reinforcement learning framework for Cold-Start E-commerce Query Suggestion (EQS). Specifically, we leverage answerability, factuality, and information gain as reward to continuously optimize the quality of suggested queries. To continuously optimize our QS model, we estimate uncertainty for grouped candidate suggested queries to select hard and ambiguous samples from online user queries lacking click signals. In addition, we provide an EQS-Benchmark comprising 16,949 online user queries for offline training and evaluation. Extensive offline and online experiments consistently demonstrate a strong positive correlation between online and offline effectiveness. Both offline and online experimental results demonstrate the superiority of our Cold-EQS, achieving a significant +6.81% improvement in online chatUV.

The EU AI Act constitutes an important development in shaping the Union's digital regulatory architecture. The Act places fundamental rights at the heart of a risk-based governance framework. The article examines how the AI Act institutionalises a human-centric approach to AI and how the AI Act's provisions explicitly and implicitly embed the protection of rights enshrined in the EU Charter of Fundamental Rights. It argues that fundamental rights function not merely as aspirational goals, but as legal thresholds and procedural triggers across the lifecycle of an AI system. The analysis suggests that the AI Act has the potential to serve as a model for rights-preserving AI systems, while acknowledging that challenges will emerge at the level of implementation.

Video understanding with multimodal large language models (MLLMs) remains challenging due to the long token sequences of videos, which contain extensive temporal dependencies and redundant frames. Existing approaches typically treat MLLMs as passive recognizers, processing entire videos or uniformly sampled frames without adaptive reasoning. Recent agent-based methods introduce external tools, yet still depend on manually designed workflows and perception-first strategies, resulting in inefficiency on long videos. We present EVA, an Efficient Reinforcement Learning framework for End-to-End Video Agent, which enables planning-before-perception through iterative summary-plan-action-reflection reasoning. EVA autonomously decides what to watch, when to watch, and how to watch, achieving query-driven and efficient video understanding. To train such agents, we design a simple yet effective three-stage learning pipeline - comprising supervised fine-tuning (SFT), Kahneman-Tversky Optimization (KTO), and Generalized Reward Policy Optimization (GRPO) - that bridges supervised imitation and reinforcement learning. We further construct high-quality datasets for each stage, supporting stable and reproducible training. We evaluate EVA on six video understanding benchmarks, demonstrating its comprehensive capabilities. Compared with existing baselines, EVA achieves a substantial improvement of 6-12% over general MLLM baselines and a further 1-3% gain over prior adaptive agent methods. Our code and model are available at https://github.com/wangruohui/EfficientVideoAgent.

To address the critical scarcity of high-quality, publicly available counseling dialogue datasets, we created Multilingual KokoroChat by translating KokoroChat, a large-scale manually authored Japanese counseling corpus, into both English and Chinese. A key challenge in this process is that the optimal model for translation varies by input, making it impossible for any single model to consistently guarantee the highest quality. In a sensitive domain like counseling, where the highest possible translation fidelity is essential, relying on a single LLM is therefore insufficient. To overcome this challenge, we developed and employed a novel multi-LLM ensemble method. Our approach first generates diverse hypotheses from multiple distinct LLMs. A single LLM then produces a high-quality translation based on an analysis of the respective strengths and weaknesses of all presented hypotheses. The quality of ``Multilingual KokoroChat'' was rigorously validated through human preference studies. These evaluations confirmed that the translations produced by our ensemble method were preferred from any individual state-of-the-art LLM. This strong preference confirms the superior quality of our method's outputs. The Multilingual KokoroChat is available at https://github.com/UEC-InabaLab/MultilingualKokoroChat.

As artificial intelligence (AI) technologies continue to advance, effective risk assessment, regulation, and oversight are necessary to ensure that AI development and deployment align with ethical principles while preserving innovation and economic competitiveness. The adoption of the EU AI Act marks an important step in this direction, establishing a harmonised legal framework that includes detailed provisions on AI governance, as well as the creation of the European AI Office. This paper revisits the question of whether a more robust supranational agency dedicated to AI is still warranted and explores how such a body could enhance policy coherence, improve risk assessment capacities, and foster international cooperation. It also argues that a strengthened EU-level agency would also serve the Union's strategic aim of securing digital and technological sovereignty.

Due to the great saving of computation and memory overhead, token compression has become a research hot-spot for MLLMs and achieved remarkable progress in image-language tasks. However, for the video, existing methods still fall short of high-ratio token compression. We attribute this shortcoming to the insufficient modeling of temporal and continual video content, and propose a novel and training-free token pruning method for video MLLMs, termed ForestPrune, which achieves effective and high-ratio pruning via Spatial-temporal Forest Modeling. In practice, ForestPrune construct token forests across video frames based on the semantic, spatial and temporal constraints, making an overall comprehension of videos. Afterwards, ForestPrune evaluates the importance of token trees and nodes based on tree depth and node roles, thereby obtaining a globally optimal pruning decision. To validate ForestPrune, we apply it to two representative video MLLMs, namely LLaVA-Video and LLaVA-OneVision, and conduct extensive experiments on a bunch of video benchmarks. The experimental results not only show the great effectiveness for video MLLMs, e.g., retaining 95.8% average accuracy while reducing 90% tokens for LLaVA-OneVision, but also show its superior performance and efficiency than the compared token compression methods, e.g., +10.1% accuracy on MLVU and -81.4% pruning time than FrameFusion on LLaVA-Video.

The increasing memory demand of the Key-Value (KV) cache poses a significant bottleneck for Large Language Models (LLMs) in long-context applications. Existing low-rank compression methods often rely on irreversible parameter transformations, sacrificing the flexibility to switch back to full-precision inference when memory is abundant. In this paper, we propose EchoKV, a flexible KV cache compression scheme that enables on-demand transitions between standard and compressed inference. Unlike traditional compression-decompression paradigms, EchoKV utilizes a lightweight network to reconstruct the residual KV components from a partial subset, leveraging intrinsic inter-layer and intra-layer similarities among attention heads. We further introduce a two-stage fine-tuning strategy that allows for rapid, low-cost training (e.g., ~1 A100 GPU-hour for a 7B model). Experimental results on LongBench and RULER demonstrate that EchoKV consistently outperforms existing methods across various compression ratios while maintaining high throughput for short-context scenarios.

Assuming that neither source data nor the source model is accessible, black box domain adaptation represents a highly practical yet extremely challenging setting, as transferable information is restricted to the predictions of the black box source model, which can only be queried using target samples. Existing approaches attempt to extract transferable knowledge through pseudo label refinement or by leveraging external vision language models (ViLs), but they often suffer from noisy supervision or insufficient utilization of the semantic priors provided by ViLs, which ultimately hinder adaptation performance. To overcome these limitations, we propose a dual teacher distillation with subnetwork rectification (DDSR) model that jointly exploits the specific knowledge embedded in black box source models and the general semantic information of a ViL. DDSR adaptively integrates their complementary predictions to generate reliable pseudo labels for the target domain and introduces a subnetwork driven regularization strategy to mitigate overfitting caused by noisy supervision. Furthermore, the refined target predictions iteratively enhance both the pseudo labels and ViL prompts, enabling more accurate and semantically consistent adaptation. Finally, the target model is further optimized through self training with classwise prototypes. Extensive experiments on multiple benchmark datasets validate the effectiveness of our approach, demonstrating consistent improvements over state of the art methods, including those using source data or models.

Mitigating elderly loneliness requires policy interventions that achieve both adaptability and auditability. Existing methods struggle to reconcile these objectives: traditional agent-based models suffer from static rigidity, while direct large language model (LLM) controllers lack essential traceability. This work proposes a three-layer framework that separates diagnosis from control to achieve both properties simultaneously. LLMs operate strictly as diagnostic instruments that assess population state and generate structured risk evaluations, while deterministic formulas with explicit bounds translate these assessments into traceable parameter updates. This separation ensures that every policy decision can be attributed to inspectable rules while maintaining adaptive response to emergent needs. We validate the framework through systematic ablation across five experimental conditions in elderly care simulation. Results demonstrate that explicit control rules outperform end-to-end black-box LLM approaches by 11.7\% while preserving full auditability, confirming that transparency need not compromise adaptive performance.

Optimizing survival outcomes, such as patient survival or customer retention, is a critical objective in data-driven decision-making. Off-Policy Evaluation~(OPE) provides a powerful framework for assessing such decision-making policies using logged data alone, without the need for costly or risky online experiments in high-stakes applications. However, typical estimators are not designed to handle right-censored survival outcomes, as they ignore unobserved survival times beyond the censoring time, leading to systematic underestimation of the true policy performance. To address this issue, we propose a novel framework for OPE and Off-Policy Learning~(OPL) tailored for survival outcomes under censoring. Specifically, we introduce IPCW-IPS and IPCW-DR, which employ the Inverse Probability of Censoring Weighting technique to explicitly deal with censoring bias. We theoretically establish that our estimators are unbiased and that IPCW-DR achieves double robustness, ensuring consistency if either the propensity score or the outcome model is correct. Furthermore, we extend this framework to constrained OPL to optimize policy value under budget constraints. We demonstrate the effectiveness of our proposed methods through simulation studies and illustrate their practical impacts using public real-world data for both evaluation and learning tasks.

Combining Large Language Models (LLMs) with Reinforcement Learning (RL) enables agents to interpret language instructions more effectively for task execution. However, LLMs typically lack direct perception of the physical environment, which limits their understanding of environmental dynamics and their ability to generalize to unseen tasks. To address this limitation, we propose Visual-Language Knowledge-Guided Offline Reinforcement Learning (VLGOR), a framework that integrates visual and language knowledge to generate imaginary rollouts, thereby enriching the interaction data. The core premise of VLGOR is to fine-tune a vision-language model to predict future states and actions conditioned on an initial visual observation and high-level instructions, ensuring that the generated rollouts remain temporally coherent and spatially plausible. Furthermore, we employ counterfactual prompts to produce more diverse rollouts for offline RL training, enabling the agent to acquire knowledge that facilitates following language instructions while grounding in environments based on visual cues. Experiments on robotic manipulation benchmarks demonstrate that VLGOR significantly improves performance on unseen tasks requiring novel optimal policies, achieving a success rate over 24% higher than the baseline methods.

We propose the Identifiable Variational Dynamic Factor Model (iVDFM), which learns latent factors from multivariate time series with identifiability guarantees. By applying iVAE-style conditioning to the innovation process driving the dynamics rather than to the latent states, we show that factors are identifiable up to permutation and component-wise affine (or monotone invertible) transformations. Linear diagonal dynamics preserve this identifiability and admit scalable computation via companion-matrix and Krylov methods. We demonstrate improved factor recovery on synthetic data, stable intervention accuracy on synthetic SCMs, and competitive probabilistic forecasting on real-world benchmarks.

Whole-aircraft diagnosis for general aviation faces threefold challenges: data uncertainty, task heterogeneity, and computational inefficiency. Existing end-to-end approaches uniformly model health discrimination and fault characterization, overlooking intrinsic receptive field conflicts between global context modeling and local feature extraction, while incurring prohibitive training costs under severe class imbalance. To address these, this study proposes the Diagnosis Decomposition Framework (DDF), explicitly decoupling diagnosis into Anomaly Detection (AD) and Fault Classification (FC) subtasks via the Long-Micro Scale Diagnostician (LMSD). Employing a "long-range global screening and micro-scale local precise diagnosis" strategy, LMSD utilizes Convolutional Tokenizer with Multi-Head Self-Attention (ConvTokMHSA) for global operational pattern discrimination and Multi-Micro Kernel Network (MMK Net) for local fault feature extraction. Decoupled training separates "large-sample lightweight" and "small-sample complex" optimization pathways, significantly reducing computational overhead. Concurrently, Keyness Extraction Layer (KEL) via knowledge distillation furnishes physically traceable explanations for two-stage decisions, materializing interpretability-by-design. Experiments on the NGAFID real-world aviation dataset demonstrate approximately 4-8% improvement in Multi-Class Weighted Penalty Metric (MCWPM) over baselines with substantially reduced training time, validating comprehensive advantages in task adaptability, interpretability, and efficiency. This provides a deployable methodology for general aviation health management.

The rapid advancement of Vision-Language Models (VLMs) has brought their safety vulnerabilities into sharp focus. However, existing red teaming methods are fundamentally constrained by an inherent linear exploration paradigm, confining them to optimizing within a predefined strategy set and preventing the discovery of novel, diverse exploits. To transcend this limitation, we introduce TreeTeaming, an automated red teaming framework that reframes strategy exploration from static testing to a dynamic, evolutionary discovery process. At its core lies a strategic Orchestrator, powered by a Large Language Model (LLM), which autonomously decides whether to evolve promising attack paths or explore diverse strategic branches, thereby dynamically constructing and expanding a strategy tree. A multimodal actuator is then tasked with executing these complex strategies. In the experiments across 12 prominent VLMs, TreeTeaming achieves state-of-the-art attack success rates on 11 models, outperforming existing methods and reaching up to 87.60\% on GPT-4o. The framework also demonstrates superior strategic diversity over the union of previously public jailbreak strategies. Furthermore, the generated attacks exhibit an average toxicity reduction of 23.09\%, showcasing their stealth and subtlety. Our work introduces a new paradigm for automated vulnerability discovery, underscoring the necessity of proactive exploration beyond static heuristics to secure frontier AI models.

Confidence calibration assumes a unique ground-truth label per input, yet this assumption fails wherever annotators genuinely disagree. Post-hoc calibrators fitted on majority-voted labels, the standard single-label targets used in practice, can appear well-calibrated under conventional evaluation yet remain substantially miscalibrated against the underlying annotator distribution. We show that this failure is structural: under simplifying assumptions, Temperature Scaling is biased toward temperatures that underestimate annotator uncertainty, with true-label miscalibration increasing monotonically with annotation entropy. To address this, we develop a family of ambiguity-aware post-hoc calibrators that optimise proper scoring rules against the full label distribution and require no model retraining. Our methods span progressively weaker annotation requirements: Dirichlet-Soft leverages the full annotator distribution and achieves the best overall calibration quality across settings; Monte Carlo Temperature Scaling with a single annotation per example (MCTS S=1) matches full-distribution calibration across all benchmarks, demonstrating that pre-aggregated label distributions are unnecessary; and Label-Smooth Temperature Scaling (LS-TS) operates with voted labels alone by constructing data-driven pseudo-soft targets from the model's own confidence. Experiments on four benchmarks with real multi-annotator distributions (CIFAR-10H, ChaosNLI) and clinically-informed synthetic annotations (ISIC~2019, DermaMNIST) show that Dirichlet-Soft reduces true-label ECE by 55-87% relative to Temperature Scaling, while LS-TS reduces ECE by 9-77% without any annotator data.

Efficient solutions for satisfiability modulo theories (SMT) are integral in industrial applications such as hardware verification and design automation. Existing approaches are predominantly based on conflict-driven clause learning, which is structurally difficult to parallelize and therefore scales poorly. In this work, we introduce FourierSMT as a scalable and highly parallelizable continuous-variable optimization framework for SMT. We generalize the Walsh-Fourier expansion (WFE), called extended WFE (xWFE), from the Boolean domain to a mixed Boolean-real domain, which allows the use of gradient methods for SMT. This addresses the challenge of finding satisfying variable assignments to high-arity constraints by local updates of discrete variables. To reduce the evaluation complexity of xWFE, we present the extended binary decision diagram (xBDD) and map the constraints from xWFE to xBDDs. We then show that sampling the circuit-output probability (COP) of xBDDs under randomized rounding is equivalent to the expectation value of the xWFEs. This allows for efficient computation of the constraints. We show that the reduced problem is guaranteed to converge and preserves satisfiability, ensuring the soundness of the solutions. The framework is benchmarked for large-scale scheduling and placement problems with up to 10,000 variables and 700,000 constraints, achieving 8-fold speedups compared to state-of-the-art SMT solvers. These results pave the way for GPU-based optimization of SMTs with continuous systems.

Learning a generalist control policy for dexterous manipulation typically relies on large-scale datasets. Given the high cost of real-world data collection, a practical alternative is to generate synthetic data through simulation. However, the resulting synthetic data often exhibits a significant gap from real-world distributions. While many prior studies have proposed algorithms to bridge the Sim-to-Real discrepancy, there remains a lack of principled research that grounds these methods in real-world manipulation tasks, particularly their performance on generalist policies such as Vision-Language-Action (VLA) models. In this study, we empirically examine the primary determinants of Sim-to-Real generalization across four dimensions: multi-level domain randomization, photorealistic rendering, physics-realistic modeling, and reinforcement learning updates. To support this study, we design a comprehensive evaluation protocol to quantify the real-world performance of manipulation tasks. The protocol accounts for key variations in background, lighting, distractors, object types, and spatial features. Through experiments involving over 10k real-world trials, we derive critical insights into Sim-to-Real transfer. To inform and advance future studies, we release both the robotic platforms and the evaluation protocol for public access to facilitate independent verification, thereby establishing a realistic and standardized benchmark for dexterous manipulation policies.

Current large language models (LLMs) primarily rely on linear sequence generation and massive parameter counts, yet they severely struggle with complex algorithmic reasoning. While recent reasoning architectures, such as the Hierarchical Reasoning Model (HRM) and Tiny Recursive Model (TRM), demonstrate that compact recursive networks can tackle these tasks, their training dynamics often lack rigorous mathematical guarantees, leading to instability and representational collapse. We propose the Contraction Mapping Model (CMM), a novel architecture that reformulates discrete recursive reasoning into continuous Neural Ordinary and Stochastic Differential Equations (NODEs/NSDEs). By explicitly enforcing the convergence of the latent phase point to a stable equilibrium state and mitigating feature collapse with a hyperspherical repulsion loss, the CMM provides a mathematically grounded and highly stable reasoning engine. On the Sudoku-Extreme benchmark, a 5M-parameter CMM achieves a state-of-the-art accuracy of 93.7 %, outperforming the 27M-parameter HRM (55.0 %) and 5M-parameter TRM (87.4 %). Remarkably, even when aggressively compressed to an ultra-tiny footprint of just 0.26M parameters, the CMM retains robust predictive power, achieving 85.4 % on Sudoku-Extreme and 82.2 % on the Maze benchmark. These results establish a new frontier for extreme parameter efficiency, proving that mathematically rigorous latent dynamics can effectively replace brute-force scaling in artificial reasoning.

Large Language Models (LLMs) have become core cognitive components in modern artificial intelligence (AI) systems, combining internal knowledge with external context to perform complex tasks. However, LLMs typically treat all accessible data indiscriminately, lacking inherent awareness of knowledge ownership and access boundaries. This deficiency heightens risks of sensitive data leakage and adversarial manipulation, potentially enabling unauthorized system access and severe security crises. Existing protection strategies rely on rigid, uniform defense that prevent dynamic authorization. Structural isolation methods faces scalability bottlenecks, while prompt guidance methods struggle with fine-grained permissions distinctions. Here, we propose the Chain-of-Authorization (CoA) framework, a secure training and reasoning paradigm that internalizes authorization logic into LLMs' core capabilities. Unlike passive external defneses, CoA restructures the model's information flow: it embeds permission context at input and requires generating explicit authorization reasoning trajectory that includes resource review, identity resolution, and decision-making stages before final response. Through supervised fine-tuning on data covering various authorization status, CoA integrates policy execution with task responses, making authorization a causal prerequisite for substantive responses. Extensive evaluations show that CoA not only maintains comparable utility in authorized scenarios but also overcomes the cognitive confusion when permissions mismatches. It exhibits high rejection rates against various unauthorized and adversarial access. This mechanism leverages LLMs' reasoning capability to perform dynamic authorization, using natural language understanding as a proactive security mechanism for deploying reliable LLMs in modern AI systems.

Agentic computing systems, which autonomously spawn new functionalities based on natural language instructions, are becoming increasingly prevalent. While immensely capable, these systems raise serious security, privacy, and safety concerns. Fundamentally, the full set of functionalities offered by these systems, combined with their probabilistic execution flows, is not known beforehand. Given this lack of characterization, it is non-trivial to validate whether a system has successfully carried out the user's intended task or instead executed irrelevant actions, potentially as a consequence of compromise. In this paper, we propose Agent-Sentry, a framework that attempts to bound agentic systems to address this problem. Our key insight is that agentic systems are designed for specific use cases and therefore need not expose unbounded or unspecified functionalities. Once bounded, these systems become easier to scrutinize. Agent-Sentry operationalizes this insight by uncovering frequent functionalities offered by an agentic system, along with their execution traces, to construct behavioral bounds. It then learns a policy from these traces and blocks tool calls that deviate from learned behaviors or that misalign with user intent. Our evaluation shows that Agent-Sentry helps prevent over 90\% of attacks that attempt to trigger out-of-bounds executions, while preserving up to 98\% of system utility.

Multimodal stacks that mix ViTs, CNNs, GNNs, and transformer NLP strain embedded platforms because their compute/memory patterns diverge and hard real-time targets leave little slack. TRINE is a single-bitstream FPGA accelerator and compiler that executes end-to-end multimodal inference without reconfiguration. Layers are unified as DDMM/SDDMM/SpMM and mapped to a mode-switchable engine that toggles at runtime among weight/output-stationary systolic, 1xCS SIMD, and a routable adder tree (RADT) on a shared PE array. A width-matched, two-stage top-k unit enables in-stream token pruning, while dependency-aware layer offloading (DALO) overlaps independent kernels across reconfigurable processing units to sustain utilization. Evaluated on Alveo U50 and ZCU104, TRINE reduces latency by up to 22.57x vs. RTX 4090 and 6.86x vs. Jetson Orin Nano at 20-21 W; token pruning alone yields up to 7.8x on ViT-heavy pipelines, and DALO contributes up to 79% throughput improvement. With int8 quantization, accuracy drops remain <2.5% across representative tasks, delivering state-of-the-art latency and energy efficiency for unified vision, language, and graph workloads-in one bitstream.

Tool use enables large language models (LLMs) to access external information, invoke software systems, and act in digital environments beyond what can be solved from model parameters alone. Early research mainly studied whether a model could select and execute a correct single tool call. As agent systems evolve, however, the central problem has shifted from isolated invocation to multi-tool orchestration over long trajectories with intermediate state, execution feedback, changing environments, and practical constraints such as safety, cost, and verifiability. We comprehensively review recent progress in multi-tool LLM agents and analyzes the state of the art in this rapidly developing area. First, we unify task formulations and distinguish single-call tool use from long-horizon orchestration. Then, we organize the literature around six core dimensions: inference-time planning and execution, training and trajectory construction, safety and control, efficiency under resource constraints, capability completeness in open environments, and benchmark design and evaluation. We further summarize representative applications in software engineering, enterprise workflows, graphical user interfaces, and mobile systems. Finally, we discuss major challenges and outline future directions for building reliable, scalable, and verifiable multi-tool agents.

We introduce the Dual-View Pheromone Pathway Network (DPPN), an architecture that routes sparse attention through a persistent pheromone field over latent slot transitions, and use it to discover two independent requirements for persistent structural memory in neural networks. Through five progressively refined experiments using up to 10 seeds per condition across 5 model variants and 4 transfer targets, we identify a core principle: persistent memory requires a stable coordinate system, and any coordinate system learned jointly with the model is inherently unstable. We characterize three obstacles -- pheromone saturation, surface-structure entanglement, and coordinate incompatibility -- and show that neither contrastive updates, multi-source distillation, Hungarian alignment, nor semantic decomposition resolves the instability when embeddings are learned from scratch. Fixed random Fourier features provide extrinsic coordinates that are stable, structure-blind, and informative, but coordinate stability alone is insufficient: routing-bias pheromone does not transfer (10 seeds, p>0.05). DPPN outperforms transformer and random sparse baselines for within-task learning (AULC 0.700 vs 0.680 vs 0.670). Replacing routing bias with learning-rate modulation eliminates negative transfer: warm pheromone as a learning-rate prior achieves +0.003 on same-family tasks (17 seeds, p<0.05) while never reducing performance. A structure completion function over extrinsic coordinates produces +0.006 same-family bonus beyond regularization, showing the catch-22 between stability and informativeness is partially permeable to learned functions. The contribution is two independent requirements for persistent structural memory: (a) coordinate stability and (b) graceful transfer mechanism.

Task-oriented object detection (TOOD) atop CLIP offers open-vocabulary, prompt-driven semantics, yet dense per-window computation and heavy memory traffic hinder real-time, power-limited edge deployment. We present \emph{TorR}, a brain-inspired \textbf{algorithm--architecture co-design} that \textbf{replaces CLIP-style dense alignment with a hyperdimensional (HDC) associative reasoner} and turns temporal coherence into reuse. On the \emph{algorithm} side, TorR reformulates alignment as HDC similarity and graph composition, introducing \emph{partial-similarity reuse} via (i) query caching with per-class score accumulation, (ii) exact $δ$-updates when only a small set of hypervector bits change, and (iii) similarity/load-gated bypass under high system load. On the \emph{architecture} side, TorR instantiates a lane-scalable, bit-sliced item memory with bank/precision gating and a lightweight controller that schedules bypass/$δ$/full paths to meet RT-30/RT-60 targets as object counts vary. Synthesized in a TSMC 28\,nm process and exercised with a cycle-accurate simulator, TorR sustains real-time throughput with millijoule-scale energy per window ($\approx$50\,mJ at 60\,FPS; $\approx$113\,mJ at 30\,FPS) and low latency jitter, while delivering competitive AP@0.5 across five task prompts (mean 44.27\%) within a bounded margin to strong VLM baselines, but at orders-of-magnitude lower energy. The design exposes deployment-time configurability (effective dimension $D'$, thresholds, precision) to trade accuracy, latency, and energy for edge budgets.

Deep learning techniques for rumor detection typically utilize Graph Neural Networks (GNNs) to analyze post relations. These methods, however, falter due to over-smoothing issues when processing rumor propagation structures, leading to declining performance. Our investigation into this issue reveals that over-smoothing is intrinsically tied to the structural characteristics of rumor propagation trees, in which the majority of nodes are 1-level nodes. Furthermore, GNNs struggle to capture long-range dependencies within these trees. To circumvent these challenges, we propose a Pre-Trained Propagation Tree Transformer (P2T3) method based on pure Transformer architecture. It extracts all conversation chains from a tree structure following the propagation direction of replies, utilizes token-wise embedding to infuse connection information and introduces necessary inductive bias, and pre-trains on large-scale unlabeled datasets. Experiments indicate that P2T3 surpasses previous state-of-the-art methods in multiple benchmark datasets and performs well under few-shot conditions. P2T3 not only avoids the over-smoothing issue inherent in GNNs but also potentially offers a large model or unified multi-modal scheme for future social media research.

What should a developer inspect before deploying an LLM agent: the model, the tool code, the deployment configuration, or all three? In practice, many security failures in agent systems arise not from model weights alone, but from the surrounding software stack: tool functions that pass untrusted inputs to dangerous operations, exposed credentials in deployment artifacts, and over-privileged Model Context Protocol (MCP) configurations. We present Agent Audit, a security analysis system for LLM agent applications. Agent Audit analyzes Python agent code and deployment artifacts through an agent-aware pipeline that combines dataflow analysis, credential detection, structured configuration parsing, and privilege-risk checks. The system reports findings in terminal, JSON, and SARIF formats, enabling direct integration with local development workflows and CI/CD pipelines. On a benchmark of 22 samples with 42 annotated vulnerabilities, Agent Audit detects 40 vulnerabilities with 6 false positives, substantially improving recall over common SAST baselines while maintaining sub-second scan times. Agent Audit is open source and installable via pip, making security auditing accessible for agent systems. In the live demonstration, attendees scan vulnerable agent repositories and observe how Agent Audit identifies security risks in tool functions, prompts, and more. Findings are linked to source locations and configuration paths, and can be exported into VS Code and GitHub Code Scanning for interactive inspection.

We present UniQueR, a unified query-based feedforward framework for efficient and accurate 3D reconstruction from unposed images. Existing feedforward models such as DUSt3R, VGGT, and AnySplat typically predict per-pixel point maps or pixel-aligned Gaussians, which remain fundamentally 2.5D and limited to visible surfaces. In contrast, UniQueR formulates reconstruction as a sparse 3D query inference problem. Our model learns a compact set of 3D anchor points that act as explicit geometric queries, enabling the network to infer scene structure, including geometry in occluded regions--in a single forward pass. Each query encodes spatial and appearance priors directly in global 3D space (instead of per-frame camera space) and spawns a set of 3D Gaussians for differentiable rendering. By leveraging unified query interactions across multi-view features and a decoupled cross-attention design, UniQueR achieves strong geometric expressiveness while substantially reducing memory and computational cost. Experiments on Mip-NeRF 360 and VR-NeRF demonstrate that UniQueR surpasses state-of-the-art feedforward methods in both rendering quality and geometric accuracy, using an order of magnitude fewer primitives than dense alternatives.

Embodied Visual Tracking (EVT), a core dynamic task in embodied intelligence, requires an agent to precisely follow a language-specified target. Yet most existing methods rely on single-agent imitation learning, suffering from costly expert data and limited generalization due to static training environments. Inspired by competition-driven capability evolution, we propose CoMaTrack, a competitive game-theoretic multi-agent reinforcement learning framework that trains agents in a dynamic adversarial setting with competitive subtasks, yielding stronger adaptive planning and interference-resilient strategies. We further introduce CoMaTrack-Bench, the first benchmark for competitive EVT, featuring game scenarios between a tracker and adaptive opponents across diverse environments and instructions, enabling standardized robustness evaluation under active adversarial interactions. Experiments show that CoMaTrack achieves state-of-the-art results on both standard benchmarks and CoMaTrack-Bench. Notably, a 3B VLM trained with our framework surpasses previous single-agent imitation learning methods based on 7B models on the challenging EVT-Bench, achieving 92.1% in STT, 74.2% in DT, and 57.5% in AT. The benchmark code will be available at https://github.com/wlqcode/CoMaTrack-Bench

Surgical smoke severely degrades intraoperative video quality, obscuring anatomical structures and limiting surgical perception. Existing learning-based desmoking approaches rely on scarce paired supervision and deterministic restoration pipelines, making it difficult to perform exploration or reinforcement-driven refinement under real surgical conditions. We propose PhySe-RPO, a diffusion restoration framework optimized through Physics- and Semantics-Guided Relative Policy Optimization. The core idea is to transform deterministic restoration into a stochastic policy, enabling trajectory-level exploration and critic-free updates via group-relative optimization. A physics-guided reward imposes illumination and color consistency, while a visual-concept semantic reward learned from CLIP-based surgical concepts promotes smoke-free and anatomically coherent restoration. Together with a reference-free perceptual constraint, PhySe-RPO produces results that are physically consistent, semantically faithful, and clinically interpretable across synthetic and real robotic surgical datasets, providing a principled route to robust diffusion-based restoration under limited paired supervision.

Drone detection is pivotal in numerous security and counter-UAV applications. However, existing deep learning-based methods typically struggle to balance robust feature representation with computational efficiency. This challenge is particularly acute when detecting miniature drones against complex backgrounds under severe environmental interference. To address these issues, we introduce UAV-DETR, a novel framework that integrates a small-target-friendly architecture with real-time detection capabilities. Specifically, UAV-DETR features a WTConv-enhanced backbone and a Sliding Window Self-Attention (SWSA-IFI) encoder, capturing the high-frequency structural details of tiny targets while drastically reducing parameter overhead. Furthermore, we propose an Efficient Cross-Scale Feature Recalibration and Fusion Network (ECFRFN) to suppress background noise and aggregate multi-scale semantics. To further enhance accuracy, UAV-DETR incorporates a hybrid Inner-CIoU and NWD loss strategy, mitigating the extreme sensitivity of standard IoU metrics to minor positional deviations in small objects. Extensive experiments demonstrate that UAV-DETR significantly outperforms the baseline RT-DETR on our custom UAV dataset (+6.61% in mAP50:95, with a 39.8% reduction in parameters) and the public DUT-ANTI-UAV benchmark (+1.4% in Precision, +1.0% in F1-Score). These results establish UAV-DETR as a superior trade-off between efficiency and precision in counter-UAV object detection. The code is available at https://github.com/wd-sir/UAVDETR.

Unsupervised anomaly detection plays a pivotal role in industrial defect inspection and medical image analysis, with most methods relying on the reconstruction framework. However, these methods may suffer from over-generalization, enabling them to reconstruct anomalies well, which leads to poor detection performance. To address this issue, instead of focusing solely on normality reconstruction, we propose an innovative Uncertainty-Integrated Anomaly Perception and Restoration Attention Network (URA-Net), which explicitly restores abnormal patterns to their corresponding normality. First, unlike traditional image reconstruction methods, we utilize a pre-trained convolutional neural network to extract multi-level semantic features as the reconstruction target. To assist the URA-Net learning to restore anomalies, we introduce a novel feature-level artificial anomaly synthesis module to generate anomalous samples for training. Subsequently, a novel uncertainty-integrated anomaly perception module based on Bayesian neural networks is introduced to learn the distributions of anomalous and normal features. This facilitates the estimation of anomalous regions and ambiguous boundaries, laying the foundation for subsequent anomaly restoration. Then, we propose a novel restoration attention mechanism that leverages global normal semantic information to restore detected anomalous regions, thereby obtaining defect-free restored features. Finally, we employ residual maps between input features and restored features for anomaly detection and localization. The comprehensive experimental results on two industrial datasets, MVTec AD and BTAD, along with a medical image dataset, OCT-2017, unequivocally demonstrate the effectiveness and superiority of the proposed method.

Large language models (LLMs) are increasingly utilised for social simulation and persona generation, necessitating an understanding of how they represent geopolitical identities. In this paper, we analyse personas generated for Palestinian and Israeli identities by five popular LLMs across 640 experimental conditions, varying context (war vs non-war) and assigned roles. We observe significant distributional patterns in the generated attributes: Palestinian profiles in war contexts are frequently associated with lower socioeconomic status and survival-oriented roles, whereas Israeli profiles predominantly retain middle-class status and specialised professional attributes. When prompted with explicit instructions to avoid harmful assumptions, models exhibit diverse distributional changes, e.g., marked increases in non-binary gender inferences or a convergence toward generic occupational roles (e.g., "student"), while the underlying socioeconomic distinctions often remain. Furthermore, analysis of reasoning traces reveals an interesting dynamics between model reasoning and generation: while rationales consistently mention fairness-related concepts, the final generated personas follow the aforementioned diverse distributional changes. These findings illustrate a picture of how models interpret geopolitical contexts, while suggesting that they process fairness and adjust in varied ways; there is no consistent, direct translation of fairness concepts into representative outcomes.

With the rapid development and widespread application of Large Language Models (LLMs), their potential safety risks have attracted widespread attention. Reinforcement Learning from Human Feedback (RLHF) has been adopted to enhance the safety performance of LLMs. As a simple and effective alternative to RLHF, Direct Preference Optimization (DPO) is widely used for safety alignment. However, safety alignment still suffers from severe overfitting, which limits its actual performance. This paper revisits the overfitting phenomenon from the perspective of the model's comprehension of the training data. We find that the Imbalanced Preference Comprehension phenomenon exists between responses in preference pairs, which compromises the model's safety performance. To address this, we propose Balanced Direct Preference Optimization (B-DPO), which adaptively modulates optimization strength between preferred and dispreferred responses based on mutual information. A series of experimental results show that B-DPO can enhance the safety capability while maintaining the competitive general capabilities of LLMs on various mainstream benchmarks compared to state-of-the-art methods. \color{red}{Warning: This paper contains examples of harmful texts, and reader discretion is recommended.

Implicit bias induced by gradient-based algorithms is essential to the generalization of overparameterized models, yet its mechanisms can be subtle. This work leverages the Normalized Steepest Descent} (NSD) framework to investigate how optimization geometry shapes solutions on multiclass separable data. We introduce NucGD, a geometry-aware optimizer designed to enforce low rank structures through nuclear norm constraints. Beyond the algorithm itself, we connect NucGD with emerging low-rank projection methods, providing a unified perspective. To enable scalable training, we derive an efficient SVD-free update rule via asynchronous power iteration. Furthermore, we empirically dissect the impact of stochastic optimization dynamics, characterizing how varying levels of gradient noise induced by mini-batch sampling and momentum modulate the convergence toward the expected maximum margin solutions.Our code is accessible at: https://github.com/Tsokarsic/observing-the-implicit-bias-on-multiclass-seperable-data.

Context. Nowadays, artificial intelligence agent systems are transforming from single-tool interactions to complex multi-agent orchestrations. As a result, two competing communication protocols have emerged: a tool integration protocol that standardizes how agents invoke external tools, and an inter-agent delegation protocol that enables autonomous agents to discover and delegate tasks to one another. Despite widespread industry adoption by dozens of enterprise partners, no empirical comparison of these protocols exists in the literature. Objective. The goal of this work is to develop the first systematic benchmark comparing tool-integration-only, multi-agent delegation, and hybrid architectures across standardized queries at three complexity levels, and to quantify the trade-offs in response time, context window consumption, monetary cost, error recovery, and implementation complexity.

Tracking findings in longitudinal radiology reports is crucial for accurately identifying disease progression, and the time-consuming process would benefit from automatic summarization. This work introduces a structured summarization task, where we frame longitudinal report summarization as a timeline generation task, with dated findings organized in columns and temporally related findings grouped in rows. This structured summarization format enables straightforward comparison of findings across time and facilitates fact-checking against the associated reports. The timeline is generated using a 3-step LLM process of extracting findings, generating group names, and using the names to group the findings. To evaluate such systems, we create RadTimeline, a timeline dataset focused on tracking lung-related radiologic findings in chest-related imaging reports. Experiments on RadTimeline show tradeoffs of different-sized LLMs and prompting strategies. Our results highlight that group name generation as an intermediate step is critical for effective finding grouping. The best configuration has some irrelevant findings but very good recall, and grouping performance is comparable to human annotators.

Tables are pervasive in diverse documents, making table recognition (TR) a fundamental task in document analysis. Existing modular TR pipelines separately model table structure and content, leading to suboptimal integration and complex workflows. End-to-end approaches rely heavily on large-scale TR data and struggle in data-constrained scenarios. To address these issues, we propose TDATR (Table Detail-Aware Table Recognition) improves end-to-end TR through table detail-aware learning and cell-level visual alignment. TDATR adopts a ``perceive-then-fuse'' strategy. The model first performs table detail-aware learning to jointly perceive table structure and content through multiple structure understanding and content recognition tasks designed under a language modeling paradigm. These tasks can naturally leverage document data from diverse scenarios to enhance model robustness. The model then integrates implicit table details to generate structured HTML outputs, enabling more efficient TR modeling when trained with limited data. Furthermore, we design a structure-guided cell localization module integrated into the end-to-end TR framework, which efficiently locates cell and strengthens vision-language alignment. It enhances the interpretability and accuracy of TR. We achieve state-of-the-art or highly competitive performance on seven benchmarks without dataset-specific fine-tuning.

Language models increasingly "show their work" by writing step-by-step reasoning before answering. But are these reasoning steps genuinely used, or decorative narratives generated after the model has already decided? Consider: a medical AI writes "The patient's eosinophilia and livedo reticularis following catheterization suggest cholesterol embolization syndrome. Answer: B." If we remove the eosinophilia observation, does the diagnosis change? For most frontier models, the answer is no - the step was decorative. We introduce step-level evaluation: remove one reasoning sentence at a time and check whether the answer changes. This simple test requires only API access -- no model weights -- and costs approximately $1-2 per model per task. Testing 10 frontier models (GPT-5.4, Claude Opus, DeepSeek-V3.2, MiniMax-M2.5, Kimi-K2.5, and others) across sentiment, mathematics, topic classification, and medical QA (N=376-500 each), the majority produce decorative reasoning: removing any step changes the answer less than 17% of the time, while any single step alone recovers the answer. This holds even on math, where smaller models (0.8-8B) show genuine step dependence (55% necessity). Two models break the pattern: MiniMax-M2.5 on sentiment (37% necessity) and Kimi-K2.5 on topic classification (39%) - but both shortcut other tasks. Faithfulness is model-specific and task-specific. We also discover "output rigidity": on the same medical questions, Claude Opus writes 11 diagnostic steps while GPT-OSS-120B outputs a single token. Mechanistic analysis (attention patterns) confirms that CoT attention drops more in late layers for decorative tasks (33%) than faithful ones (20%). Implications: step-by-step explanations from frontier models are largely decorative, per-model per-domain evaluation is essential, and training objectives - not scale - determine whether reasoning is genuine.

Large Vision-Language Models (LVLMs) have shown strong performance across various multimodal tasks by leveraging the reasoning capabilities of Large Language Models (LLMs). However, processing visually complex and information-rich images, such as infographics or document layouts, requires these models to generate a large number of visual tokens, leading to significant computational overhead. To address this, we propose PinPoint, a novel two-stage framework that first identifies instruction-relevant image regions and then refines them to extract fine-grained visual features for improved reasoning and efficiency. Central to our approach is the Instruction-Region Alignment, which localizes relevant regions using both visual input and textual instructions. We further introduce new annotations that provide richer ground-truth supervision for instruction-relevant regions across challenging VQA benchmarks: InfographicVQA, MultiPageDocVQA, and SinglePageDocVQA. Experimental results show that PinPoint not only achieves superior accuracy compared to existing methods but also reduces computational overhead by minimizing irrelevant visual tokens.

Humans often juggle multiple, sometimes conflicting objectives and shift their priorities as circumstances change, rather than following a fixed objective function. In contrast, most computational decision-making and multi-objective RL methods assume static preference weights or a known scalar reward. In this work, we study sequential decision-making problem when these preference weights are unobserved latent variables that drift with context. Specifically, we propose Dynamic Preference Inference (DPI), a cognitively inspired framework in which an agent maintains a probabilistic belief over preference weights, updates this belief from recent interaction, and conditions its policy on inferred preferences. We instantiate DPI as a variational preference inference module trained jointly with a preference-conditioned actor-critic, using vector-valued returns as evidence about latent trade-offs. In queueing, maze, and multi-objective continuous-control environments with event-driven changes in objectives, DPI adapts its inferred preferences to new regimes and achieves higher post-shift performance than fixed-weight and heuristic envelope baselines.

Large language models (LLMs) have achieved remarkable success in various natural language processing tasks, yet they remain prone to generating factually incorrect outputs known as hallucinations. While recent approaches have shown promise for hallucination detection by repeatedly sampling from LLMs and quantifying the semantic inconsistency among the generated responses, they rely on fixed sampling budgets that fail to adapt to query complexity, resulting in computational inefficiency. We propose an Adaptive Bayesian Estimation framework for Semantic Entropy with Guided Semantic Exploration, which dynamically adjusts sampling requirements based on observed uncertainty. Our approach employs a hierarchical Bayesian framework to model the semantic distribution, enabling dynamic control of sampling iterations through variance-based thresholds that terminate generation once sufficient certainty is achieved. We also develop a perturbation-based importance sampling strategy to systematically explore the semantic space. Extensive experiments on four QA datasets demonstrate that our method achieves superior hallucination detection performance with significant efficiency gains. In low-budget scenarios, our approach requires about 50% fewer samples to achieve comparable detection performance to existing methods, while delivers an average AUROC improvement of 12.6% under the same sampling budget.

The core of molecular dynamics simulation fundamentally lies in the interatomic potential. Traditional empirical potentials lack accuracy, while first-principles methods are computationally prohibitive. Machine learning interatomic potentials (MLIPs) promise near-quantum accuracy at linear cost, but existing models still face challenges in efficiency and stability. We presents Machine Learning Advances Neural Network (MLANet), an efficient and robust graph neural network framework. MLANet introduces a dual-path dynamic attention mechanism for geometry-aware message passing and a multi-perspective pooling strategy to construct comprehensive system representations. This design enables highly accurate modeling of atomic environments while achieving exceptional computational efficiency, making high-fidelity simulations more accessible. Tested across a wide range of datasets spanning diverse systems, including organic molecules (e.g., QM7, MD17), periodic inorganic materials (e.g., Li-containing crystals), two-dimensional materials (e.g., bilayer graphene, black phosphorus), surface catalytic reactions (e.g., formate decomposition), and charged systems, MLANet maintains competitive prediction accuracy while its computational cost is markedly lower than mainstream equivariant models, and it enables stable long-time molecular dynamics simulations. MLANet provides an efficient and practical tool for large-scale, high-accuracy atomic simulations.

We introduce PolyVeil, a protocol for private Boolean summation across $k$ clients that encodes private bits as permutation matrices in the Birkhoff polytope. A two-layer architecture gives the server perfect simulation-based security (statistical distance zero) while a separate aggregator faces \#P-hard likelihood inference via the permanent and mixed discriminant. Two variants (full and compressed) differ in what the aggregator observes. We develop a finite-sample $(\varepsilon,δ)$-DP analysis with explicit constants. In the full variant, where the aggregator sees a doubly stochastic matrix per client, the log-Lipschitz constant grows as $n^4 K_t$ and a signal-to-noise analysis shows the DP guarantee is non-vacuous only when the private signal is undetectable. In the compressed variant, where the aggregator sees a single scalar, the univariate density ratio yields non-vacuous $\varepsilon$ at moderate SNR, with the optimal decoy count balancing CLT accuracy against noise concentration. This exposes a fundamental tension. \#P-hardness requires the full matrix view (Birkhoff structure visible), while non-vacuous DP requires the scalar view (low dimensionality). Whether both hold simultaneously in one variant remains open. The protocol needs no PKI, has $O(k)$ communication, and outputs exact aggregates.

Transformers have achieved great success across a wide range of applications, yet the theoretical foundations underlying their success remain largely unexplored. To demystify the strong capacities of transformers applied to versatile scenarios and tasks, we theoretically investigate utilizing transformers as students to learn from a class of teacher models. Specifically, the teacher models covered in our analysis include convolution layers with average pooling, graph convolution layers, and various classic statistical learning models, including a variant of sparse token selection models [Sanford et al., 2023, Wang et al., 2024] and group-sparse linear predictors [Zhang et al., 2025]. When learning from this class of teacher models, we prove that one-layer transformers with simplified "position-only'' attention can successfully recover all parameter blocks of the teacher models, thus achieving the optimal population loss. Building upon the efficient mimicry of trained transformers towards teacher models, we further demonstrate that they can generalize well to a broad class of out-of-distribution data under mild assumptions. The key in our analysis is to identify a fundamental bilinear structure shared by various learning tasks, which enables us to establish unified learning guarantees for these tasks when treating them as teachers for transformers.

The category of figurative language contains many varieties, some of which are non-compositional in nature. This type of phrase or multi-word expression (MWE) includes idioms, which represent a single meaning that does not consist of the sum of its words. For language models, this presents a unique problem due to tokenization and adjacent contextual embeddings. Many large language models have overcome this issue with large phrase vocabulary, though immediate recognition frequently fails without one- or few-shot prompting or instruction finetuning. The best results have been achieved with BERT-based or LSTM finetuning approaches. The model in this paper contains one such variety. We propose BERT- and RoBERTa-based models finetuned with a combination of slot loss and span contrastive loss (SCL) with hard negative reweighting to improve idiomaticity detection, attaining state of the art sequence accuracy performance on existing datasets. Comparative ablation studies show the effectiveness of SCL and its generalizability. The geometric mean of F1 and sequence accuracy (SA) is also proposed to assess a model's span awareness and general performance together.

Embodied agents for creative tasks like photography must bridge the semantic gap between high-level language commands and geometric control. We introduce PhotoAgent, an agent that achieves this by integrating Large Multimodal Models (LMMs) reasoning with a novel control paradigm. PhotoAgent first translates subjective aesthetic goals into solvable geometric constraints via LMM-driven, chain-of-thought (CoT) reasoning, allowing an analytical solver to compute a high-quality initial viewpoint. This initial pose is then iteratively refined through visual reflection within a photorealistic internal world model built with 3D Gaussian Splatting (3DGS). This ``mental simulation'' replaces costly and slow physical trial-and-error, enabling rapid convergence to aesthetically superior results. Evaluations confirm that PhotoAgent excels in spatial reasoning and achieves superior final image quality.

Classroom AI is rapidly expanding from low-level perception toward higher-level judgments about engagement, confusion, collaboration, and instructional quality. Yet classrooms are among the hardest real-world settings for multimodal vision: they are multi-party, noisy, privacy-sensitive, pedagogically diverse, and often multilingual. In this paper, we argue that classroom AI should be treated as a critical domain, where raw predictive accuracy is insufficient unless predictions are accompanied by verifiable evidence, calibrated uncertainty, and explicit deployment guardrails. We introduce NSCR, a neuro-symbolic framework that decomposes classroom analytics into four layers: perceptual grounding, symbolic abstraction, executable reasoning, and governance. NSCR adapts recent ideas from symbolic fact extraction and verifiable code generation to multimodal educational settings, enabling classroom observations from video, audio, ASR, and contextual metadata to be converted into typed facts and then composed by executable rules, programs, and policy constraints. Beyond the system design, we contribute a benchmark and evaluation protocol organized around five tasks: classroom state inference, discourse-grounded event linking, temporal early warning, collaboration analysis, and multilingual classroom reasoning. We further specify reliability metrics centered on abstention, calibration, robustness, construct alignment, and human usefulness. The paper does not report new empirical results; its contribution is a concrete framework and evaluation agenda intended to support more interpretable, privacy-aware, and pedagogically grounded multimodal AI for classrooms.

How should multi-agent systems be designed, and can that design knowledge be captured in a form that is inspectable, revisable, and transferable? We introduce ABSTRAL, a framework that treats MAS architecture as an evolving natural-language document, an artifact refined through contrastive trace analysis. Three findings emerge. First, we provide a precise measurement of the multi-agent coordination tax: under fixed turn budgets, ensembles achieve only 26% turn efficiency, with 66% of tasks exhausting the limit, yet still improve over single-agent baselines by discovering parallelizable task decompositions. Second, design knowledge encoded in documents transfers: topology reasoning and role templates learned on one domain provide a head start on new domains, with transferred seeds matching coldstart iteration 3 performance in a single iteration. Third, contrastive trace analysis discovers specialist roles absent from any initial design, a capability no prior system demonstrates. On SOPBench (134 bank tasks, deterministic oracle), ABSTRAL reaches 70% validation / 65.96% test pass rate with a GPT-4o backbone. We release the converged documents as inspectable design rationale.

The Random Forest model is one of the popular models of Machine learning. We present a quantum algorithm for testing (forecasting) process of the Random Forest machine learning model for the Regression problem. The presented algorithm is more efficient (in terms of query complexity or running time) than the classical counterpart.

Quantum simulators are a foundational component of the quantum software ecosystem. They are widely used to develop and debug quantum programs, validate compiler transformations, and support empirical claims about correctness and performance. In the absence of large-scale quantum hardware, simulator outputs are often treated as ground truth for algorithm development and system evaluation. However, quantum simulators also introduce unique implementation challenges. They must faithfully emulate quantum behavior while executing on classical hardware, requiring complex representations of quantum state evolution, operator composition, and noise modeling. Yet, we still lack a large-scale and in-depth study of failures in quantum simulators. To bridge this gap, this work presents a comprehensive empirical study of bugs in widely used open-source quantum simulators. We analyze 394 confirmed bugs from 12 simulators and manually categorize them based on root causes, failure manifestations, affected components, and discovery mechanisms. Our study reveals several key findings. First, bug discovery is largely user-driven, with most crashes, exceptions, and resource-related failures not detected by automated testing and identified after deployment. Second, logical correctness failures are widespread and often silent, producing plausible but incorrect outputs without triggering crashes or explicit error signals. Third, many critical failures originate in classical simulator infrastructure, such as memory management, indexing, configuration, and dependency compatibility, rather than in core quantum execution logic. These findings provide new insights into the reliability challenges of quantum simulators and highlight opportunities to improve testing and validation practices in the quantum software ecosystem.

This retrospective cohort study used continuous AI monitoring to estimate fall rates by exposure time rather than occupied bed-days. From August 2024 to December 2025, 3,980 eligible monitoring units contributed 292,914 hourly rows, yielding probability-weighted rates of 17.8 falls per 1,000 chair exposure-hours and 4.3 per 1,000 bed exposure-hours. Within the study window, 43 adjudicated falls matched the monitoring pipeline, and 40 linked to eligible exposure hours for the primary Poisson model, producing an adjusted chair-versus-bed rate ratio of 2.35 (95% confidence interval 0.87 to 6.33; p=0.0907). In a separate broader observation cohort (n=32 deduplicated events), 6 of 7 direct chair falls involved footrest-positioning failures. Because this was an observational study in a single health system, these findings remain hypothesis-generating and support testing safer chair setups rather than using chairs less.

Large language models (LLMs) can often produce substantially better outputs when allowed to use additional test-time computation, such as sampling, chain of thought, backtracking, or revising partial solutions. Despite the growing empirical success of such techniques, there is limited theoretical understanding of how inference time computation should be structured, or what constitutes an optimal use of a fixed computation budget. We model test-time computation as an algorithm interacting with a Markov chain: at any point, the algorithm may resume generation from any previously observed state. That is, unlike standard Markov chains where the states are drawn passively, we allow the algorithm to backtrack to any previously observed state of the Markov chain at any time. Many of the existing test-time algorithms, such as Chain-of-Thought (CoT) (Wei et al., 2023), Tree-of-Thoughts (ToT) (Yao et al., 2023), or Best-of-$k$ (Brown et al., 2024) could be seen as specific algorithms in this model. We prove that while backtracking can reduce the number of generations exponentially, a very limited form of backtracking is theoretically sufficient. Namely, we show that the optimal algorithm always generates a caterpillar tree. That is, if we remove the leaves of the state tree generated by the optimal algorithm, we obtain a path. Motivated by our characterization of the optimal algorithm, we present Caterpillar of Thoughts (CaT), a new test-time computation algorithm, reducing the number of token/state generations. Our empirical evaluation shows that CaT, compared to ToT, achieves a better success rate while also reducing the number of token generations.

Large Language Models (LLMs) are equipped with profound semantic knowledge, making them a natural choice for injecting semantic generalization into personalized search systems. However, in practice we find that directly fine-tuning LLMs on industrial personalized tasks (e.g. next item prediction) often yields suboptimal results. We attribute this bottleneck to a critical Knowledge--Action Gap: the inherent conflict between preserving pre-trained semantic knowledge and aligning with specific personalized actions by discriminative objectives. Empirically, action-only training objectives induce Semantic Collapse, such as attention ``sinks''. This degradation severely cripples the LLM's generalization, failing to bring improvements to personalized search systems. We propose KARMA (Knowledge--Action Regularized Multimodal Alignment), a unified framework that treats semantic reconstruction as a train-only regularizer. KARMA optimizes a next-interest embedding for retrieval (Action) while enforcing semantic decodability (Knowledge) through two complementary objectives: (i) history-conditioned semantic generation, which anchors optimization to the LLM's native next-token distribution, and (ii) embedding-conditioned semantic reconstruction, which constrains the interest embedding to remain semantically recoverable. On Taobao search system, KARMA mitigates semantic collapse (attention-sink analysis) and improves both action metrics and semantic fidelity. In ablations, semantic decodability yields up to +22.5 HR@200. With KARMA, we achieve +0.25 CTR AUC in ranking, +1.86 HR in pre-ranking and +2.51 HR in recalling. Deployed online with low inference overhead at ranking stage, KARMA drives +0.5% increase in Item Click.

Agricultural pest management increasingly relies on timely and accurate access to expert knowledge, yet high quality labeled data and continuous expert support remain limited, particularly for farmers operating in rural regions with unstable/no internet connectivity. At the same time, the rapid growth of AI and LLMs has created new opportunities to deliver practical decision support tools directly to end users in agriculture through compact and deployable systems. This work addresses (i) generating a structured insect information dataset, and (ii) adapting a lightweight LLM model ($\leq$ 7B) by fine tuning it for edge device uses in agricultural pest management. The textual data collection was done by reviewing and collecting information from available pest databases and published manuscripts on nine selected pest species. These structured reports were then reviewed and validated by a domain expert. From these reports, we constructed Q/A pairs to support model training and evaluation. A LoRA-based fine-tuning approach was applied to multiple lightweight LLMs and evaluated. Initial evaluation shows that Mistral 7B achieves an 88.9\% pass rate on the domain-specific Q/A task, substantially outperforming Qwen 2.5 7B (63.9\%), and LLaMA 3.1 8B (58.7\%). Notably, Mistral demonstrates higher semantic alignment (embedding similarity: 0.865) despite lower lexical overlap (BLEU: 0.097), indicating that semantic understanding and robust reasoning are more predictive of task success than surface-level conformity in specialized domains. By combining expert organized data, well-structured Q/A pairs, semantic quality control, and efficient model adaptation, this work contributes towards providing support for farmer facing agricultural decision support tools and demonstrates the feasibility of deploying compact, high-performing language models for practical field-level pest management guidance.

Large-scale machine learning workloads increasingly rely on multi-GPU systems, yet their performance is often limited by an overlooked component: the CPU. Through a detailed study of modern large language model (LLM) inference and serving workloads, we find that multi-GPU performance frequently degrades not because GPUs are saturated, but because CPUs fail to keep the GPUs busy. Under limited CPU allocations, systems exhibit symptoms such as delayed kernel launch, stalled communication, and increased tokenization latency, leading to severe GPU underutilization even when ample GPU resources are available. This work presents a systematic analysis of CPU-induced slowdowns in multi-GPU LLM inference. We show that these bottlenecks persist even in serving stacks that employ process-level separation and modern GPU-side optimizations such as CUDA Graphs. Since the marginal cost of additional CPU cores is small relative to GPU instance pricing, our evaluation indicates that increasing the number of CPU cores can substantially improve performance and stability at minimal additional cost. Under moderate serving load, we observe that CPU-starved configurations frequently time out, while providing adequate CPU resources restores responsiveness and reduces time-to-first-token (TTFT) latency by 1.36-5.40x across configurations, all without requiring additional GPUs. This work shows that CPU provisioning is a crucial factor in multi-GPU LLM inference configuration, helping prevent control-side bottlenecks.

As the Internet of Things (IoT) continues to expand across critical infrastructure, smart environments, and consumer devices, securing them against cyber threats has become increasingly vital. Traditional intrusion detection models often treat IoT threats as binary classification problems or rely on opaque models, thereby limiting trust. This work studies multiclass threat attribution in IoT environments using the CICIoT2023 dataset, grouping over 30 attack variants into 8 semantically meaningful classes. We utilize a combination of a gradient boosting model and SHAP (SHapley Additive exPlanations) to deliver both global and class-specific explanations, enabling detailed insight into the features driving each attack classification. The results show that the model distinguishes distinct behavioral signatures of the attacks using flow timing, packet size uniformity, TCP flag dynamics, and statistical variance. Additional analysis that exposes both feature attribution and the decision trajectory per class further validates these observed patterns. Our findings contribute to the development of more accurate and explainable intrusion detection systems, bridging the gap between high-performance machine learning and the need for trust and accountability in AI-driven cybersecurity for IoT environments.

The deployment of deep neural networks (DNNs) in safety-critical edge environments necessitates robustness against hardware-induced bit-flip errors. While empirical studies indicate that reducing numerical precision can improve fault tolerance, the theoretical basis of this phenomenon remains underexplored. In this work, we study resilience as a structural property of neural architectures rather than solely as a property of a dataset-specific trained solution. By deriving the expected squared error (MSE) under independent parameter bit flips across multiple numerical formats and layer primitives, we show that lower precision, higher sparsity, bounded activations, and shallow depth are consistently favored under this corruption model. We then argue that logic and lookup-based neural networks realize the joint limit of these design trends. Through ablation studies on the MLPerf Tiny benchmark suite, we show that the observed empirical trends are consistent with the theoretical predictions, and that LUT-based models remain highly stable in corruption regimes where standard floating-point models fail sharply. Furthermore, we identify a novel even-layer recovery effect unique to logic-based architectures and analyze the structural conditions under which it emerges. Overall, our results suggest that shifting from continuous arithmetic weights to discrete Boolean lookups can provide a favorable accuracy-resilience trade-off for hardware fault tolerance.

Observational studies can yield clinically actionable evidence at scale, but executing them on real-world databases is open-ended and requires coherent decisions across cohort construction, analysis, and reporting. Prior evaluations of LLM agents emphasize isolated steps or single answers, missing the integrity and internal structure of the resulting evidence bundle. To address this gap, we introduce RWE-bench, a benchmark grounded in MIMIC-IV and derived from peer-reviewed observational studies. Each task provides the corresponding study protocol as the reference standard, requiring agents to execute experiments in a real database and iteratively generate tree-structured evidence bundles. We evaluate six LLMs (three open-source, three closed-source) under three agent scaffolds using both question-level correctness and end-to-end task metrics. Across 162 tasks, task success is low: the best agent reaches 39.9%, and the best open-source model reaches 30.4%. Agent scaffolds also matter substantially, causing over 30% variation in performance metrics. Furthermore, we implement an automated cohort evaluation method to rapidly localize errors and identify agent failure modes. Overall, the results highlight persistent limitations in agents' ability to produce end-to-end evidence bundles, and efficient validation remains an important direction for future work. Code and data are available at https://github.com/somewordstoolate/RWE-bench.

As AI systems increasingly mediate negotiations, understanding how the number of negotiated issues impacts human performance is crucial for maintaining human agency. We designed a human-AI negotiation case study in a realistic property rental scenario, varying the number of negotiated issues; empirical findings show that without support, performance stays stable up to three issues but declines as additional issues increase cognitive load. To address this, we introduce a novel uncertainty-based visualization driven by Bayesian estimation of agreement probability. It shows how the space of mutually acceptable agreements narrows as negotiation progresses, helping users identify promising options. In a within-subjects experiment (N=32), it improved human outcomes and efficiency, preserved human control, and avoided redistributing value. Our findings surface practical limits on the complexity people can manage in human-AI negotiation, advance theory on human performance in complex negotiations, and offer validated design guidance for interactive systems.

Data scarcity remains a persistent challenge in low-resource domains. While existing data augmentation methods leverage the generative capabilities of large language models (LLMs) to produce large volumes of synthetic data, these approaches often prioritize quantity over quality and lack domain-specific strategies. In this work, we introduce DALDALL, a persona-based data augmentation framework tailored for legal information retrieval (IR). Our method employs domain-specific professional personas--such as attorneys, prosecutors, and judges--to generate synthetic queries that exhibit substantially greater lexical and semantic diversity than vanilla prompting approaches. Experiments on the CLERC and COLIEE benchmarks demonstrate that persona-based augmentation achieves improvement in lexical diversity as measured by Self-BLEU scores, while preserving semantic fidelity to the original queries. Furthermore, dense retrievers fine-tuned on persona-augmented data consistently achieve competitive or superior recall performance compared to those trained on original data or generic augmentations. These findings establish persona-based prompting as an effective strategy for generating high-quality training data in specialized, low-resource domains.

Video Object-Centric Learning seeks to decompose raw videos into a small set of object slots, but existing slot-attention models often suffer from severe over-fragmentation. This is because the model is implicitly encouraged to occupy all slots to minimize the reconstruction objective, thereby representing a single object with multiple redundant slots. We tackle this limitation with a reconstruction-guided slot curriculum (SlotCurri). Training starts with only a few coarse slots and progressively allocates new slots where reconstruction error remains high, thus expanding capacity only where it is needed and preventing fragmentation from the outset. Yet, during slot expansion, meaningful sub-parts can emerge only if coarse-level semantics are already well separated; however, with a small initial slot budget and an MSE objective, semantic boundaries remain blurry. Therefore, we augment MSE with a structure-aware loss that preserves local contrast and edge information to encourage each slot to sharpen its semantic boundaries. Lastly, we propose a cyclic inference that rolls slots forward and then backward through the frame sequence, producing temporally consistent object representations even in the earliest frames. All combined, SlotCurri addresses object over-fragmentation by allocating representational capacity where reconstruction fails, further enhanced by structural cues and cyclic inference. Notable FG-ARI gains of +6.8 on YouTube-VIS and +8.3 on MOVi-C validate the effectiveness of SlotCurri. Our code is available at github.com/wjun0830/SlotCurri.

Independently trained domain specialists can be fused post-hoc into a single model that outperforms any individual specialist, and the gain is predictable: gain = 0.82 x divergence - 2.72 (R^2 = 0.856, n=6, 3-26% divergence). This enables practitioners to estimate cooperative value before committing compute. Below ~3.3% divergence, gains approach zero.In the KALAVAI protocol, contributors fine-tune copies of a shared checkpoint independently, then submit for lightweight MoE routing (500 steps). Gains are consistent: +7.72% at 410M (+/-0.02%, 3 seeds), +7.49% at 1B (+/-0.01%, 3 seeds), +6.53% at 6.9B, each over the best specialist. The router matches domain-oracle routing within <10^{-5} nats. Cross-lingual fusion (Tamil/Yoruba/Welsh/Code) achieves +21.76%, with Yoruba perplexity falling 41.9 to 7.7. A 20-contributor federation achieves +16.71% (+/-0.07pp, 3 seeds).Three requirements bound the protocol. Shared initialisation is necessary: checkpoint mismatch degrades routing. Frozen layers are optional below ~10,000 steps and beneficial beyond. Learned routing is essential: uniform averaging degrades by -1.2% vs. best specialist, while any trained router achieves oracle-optimal assignment.

Large language models (LLMs) solve complex problems by generating multi-step reasoning traces. Yet these traces are typically analyzed from only one of two perspectives: the sequence of tokens across different reasoning steps in the generated text, or the hidden-state vectors across model layers within one step. We introduce PRISM (Probabilistic Reasoning Inspection through Semantic and Implicit Modeling), a framework and diagnostic tool for jointly analyzing both levels, providing a unified view of how reasoning evolves across steps and layers. Across multiple reasoning models and benchmarks, PRISM uncovers systematic patterns in the reasoning process, showing that failed trajectories are more likely to become trapped in unproductive verification loops and further diverge into distinct modes such as overthinking and premature commitment, which behave differently once a candidate answer is reached. It further reveals how prompting reshapes reasoning behavior beyond aggregate accuracy by altering both semantic transitions and internal computational patterns. By modeling reasoning trajectories as structured processes, PRISM makes these behaviors observable and analyzable rather than relying solely on final-task accuracy. Taken together, these insights position PRISM as a practical tool for analyzing and diagnosing reasoning processes in LLMs.

Automated classification of clinical transcriptions into medical specialties is essential for routing, coding, and clinical decision support, yet prior work on the widely used MTSamples benchmark suffers from severe data leakage caused by applying SMOTE oversampling before train test splitting. We first document this methodological flaw and establish a leakage free benchmark across 40 medical specialties (4966 records), revealing that the true task difficulty is substantially higher than previously reported. We then introduce CLiGNet (Clinical Label Interaction Graph Network), a neural architecture that combines a Bio ClinicalBERT text encoder with a two layer Graph Convolutional Network operating on a specialty label graph constructed from semantic similarity and ICD 10 chapter priors. Per label attention gates fuse document and label graph representations, trained with focal binary cross entropy loss to handle extreme class imbalance (181 to 1 ratio). Across seven baselines ranging from TF IDF classifiers to Clinical Longformer, CLiGNet without calibration achieves the highest macro F1 of 0.279, with an ablation study confirming that the GCN label graph provides the single largest component gain (increase of 0.066 macro F1). Adding per label Platt scaling calibration yields an expected calibration error of 0.007, demonstrating a principled trade off between ranking performance and probability reliability. We provide comprehensive failure analysis covering pairwise specialty confusions, rare class behaviour, document length effects, and token level Integrated Gradients attribution, offering actionable insights for clinical NLP system deployment.

Active learning reduces labeling costs by selecting samples that maximize information gain. A dominant framework, Query-by-Committee (QBC), typically relies on perturbation-based diversity by inducing model disagreement through random feature subsetting or data blinding. While this approximates one notion of epistemic uncertainty, it sacrifices direct characterization of the plausible hypothesis space. We propose the complementary approach: Rashomon Ensembled Active Learning (REAL) which constructs a committee by exhaustively enumerating the Rashomon Set of all near-optimal models. To address functional redundancy within this set, we adopt a PAC-Bayesian framework using a Gibbs posterior to weight committee members by their empirical risk. Leveraging recent algorithmic advances, we exactly enumerate this set for the class of sparse decision trees. Across synthetic and established active learning baselines, REAL outperforms randomized ensembles, particularly in moderately noisy environments where it strategically leverages expanded model multiplicity to achieve faster convergence.

Large language models excel on objectively verifiable tasks such as math and programming, where evaluation reduces to unit tests or a single correct answer. In contrast, real-world enterprise work is often subjective and context-dependent: success hinges on organizational goals, user intent, and the quality of intermediate artifacts produced across long, multi-tool workflows. We introduce LH-Bench, a three-pillar evaluation design that moves beyond binary correctness to score autonomous, long-horizon execution on subjective enterprise tasks. The pillars are: (i) expert-grounded rubrics that give LLM judges the domain context needed to score subjective work, (ii) curated ground-truth artifacts that enable stepwise reward signals (e.g., chapter-level annotation for content tasks), and (iii) pairwise human preference evaluation for convergent validation. We show that domain-authored rubrics provide substantially more reliable evaluation signals than LLM-authored rubrics (kappa = 0.60 vs. 0.46), and that human preference judgments confirm the same top-tier separation (p < 0.05), evidence that expert-grounded evaluation can scale without sacrificing reliability. We release public datasets and report results on two environments: Figma-to-code (33 real .fig tasks against the Figma API via MCP) and Programmatic content (41 courses comprising 183 individually-evaluated chapters on a course platform serving 30+ daily users).

Generating samples from a continuous probability density is a central algorithmic problem across statistics, engineering, and the sciences. For high-dimensional settings, Hamiltonian Monte Carlo (HMC) is the default algorithm across mainstream software packages. However, despite the extensive line of work on HMC and its widespread empirical success, it remains unclear how many iterations of HMC are required as a function of the dimension $d$. On one hand, a variety of results show that Metropolized HMC converges in $O(d^{1/4})$ iterations from a warm start close to stationarity. On the other hand, Metropolized HMC is significantly slower without a warm start, e.g., requiring $Ω(d^{1/2})$ iterations even for simple target distributions such as isotropic Gaussians. Finding a warm start is therefore the computational bottleneck for HMC. We resolve this issue for the well-studied setting of sampling from a probability distribution satisfying strong log-concavity (or isoperimetry) and third-order derivative bounds. We prove that \emph{non-Metropolized} HMC generates a warm start in $\tilde{O}(d^{1/4})$ iterations, after which we can exploit the warm start using Metropolized HMC. Our final complexity of $\tilde{O}(d^{1/4})$ is the fastest algorithm for high-accuracy sampling under these assumptions, improving over the prior best of $\tilde{O}(d^{1/2})$. This closes the long line of work on the dimensional complexity of MHMC for such settings, and also provides a simple warm-start prescription for practical implementations.

Accurate prediction of mechanical properties of steel during hot rolling processes, such as Thin Slab Direct Rolling (TSDR), remains challenging due to complex interactions among chemical compositions, processing parameters, and resultant microstructures. Traditional empirical and experimental methodologies, while effective, are often resource-intensive and lack adaptability to varied production conditions. Moreover, most existing approaches do not explicitly leverage the strong correlations among key mechanical properties, missing an opportunity to improve predictive accuracy through multitask learning. To address this, we present a multitask learning framework that injects multitask awareness into the prior of TabPFN--a transformer-based foundation model for in-context learning on tabular data--through novel fine-tuning strategies. Originally designed for single-target regression or classification, we augment TabPFN's prior with two complementary approaches: (i) target averaging, which provides a unified scalar signal compatible with TabPFN's single-target architecture, and (ii) task-specific adapters, which introduce task-specific supervision during fine-tuning. These strategies jointly guide the model toward a multitask-informed prior that captures cross-property relationships among key mechanical metrics. Extensive experiments on an industrial TSDR dataset demonstrate that our multitask adaptations outperform classical machine learning methods and recent state-of-the-art tabular learning models across multiple evaluation metrics. Notably, our approach enhances both predictive accuracy and computational efficiency compared to task-specific fine-tuning, demonstrating that multitask-aware prior adaptation enables foundation models for tabular data to deliver scalable, rapid, and reliable deployment for automated industrial quality control and process optimization in TSDR.

Existing NLP work commonly treats contradictions as errors to be resolved by choosing which statements to accept or discard. Yet a key aspect of human reasoning in social interactions and professional domains is the ability to hypothesize explanations that reconcile contradictions. For example, "Cassie hates coffee" and "She buys coffee everyday" may appear contradictory, yet both are compatible if Cassie has the unenviable daily chore of buying coffee for all her coworkers. Despite the growing reasoning capabilities of large language models (LLMs), their ability to hypothesize such reconciliatory explanations remains largely unexplored. To address this gap, we introduce the task of reconciliatory explanation generation, where models must generate explanations that effectively render contradictory statements compatible. We propose a novel method of repurposing existing natural language inference (NLI) datasets, and introduce quality metrics that enable scalable automatic evaluation. Experiments with 18 LLMs show that most models achieve limited success in this task, and that the benefit of extending test-time compute by "thinking" plateaus as model size increases. Our results highlight an under-explored dimension of LLM reasoning and the need to address this limitation in enhancing LLMs' downstream applications such as chatbots and scientific aids.

Pfeffer, Krügel, and Uhl (2025) report that OpenAI's reasoning model o1-mini produces more utilitarian responses to the trolley problem and footbridge dilemma than the non-reasoning model GPT-4o. I replicate their study with four current OpenAI models and extend it with prompt variant testing. The trolley finding does not survive: GPT-4o's low utilitarian rate doesn't reflect a deontological commitment but safety refusals triggered by the prompt's advisory framing. When framed as "Is it morally permissible...?" instead of "Should I...?", GPT-4o gives 99% utilitarian responses. All models converge on utilitarian answers when prompt confounds are removed. The footbridge finding survives with blemishes. Reasoning models tend to give more utilitarian responses than non-reasoning models across prompt variations. But often they refuse to answer the dilemma or, when they answer, give a non-utilitarian rather than a utilitarian answer. These results demonstrate that single-prompt evaluations of LLM moral reasoning are unreliable: multi-prompt robustness testing should be standard practice for any empirical claim about LLM behavior.

Driver heterogeneity is often reduced to labels or discrete regimes, compressing what is inherently dynamic into static categories. We introduce quantum-inspired representation that models each driver as an evolving latent state, presented as a density matrix with structured mathematical properties. Behavioral observations are embedded via non-linear Random Fourier Features, while state evolution blends temporal persistence of behavior with context-dependent profile activation. We evaluate our approach on empirical driving data, Third Generation Simulation Data (TGSIM), showing how driving profiles are extracted and analyzed.

This work proposes a novel immersive communication framework that leverages brain-computer interface (BCI) to acquire brain signals for inferring user-centric states (e.g., intention and perception-related discomfort), thereby enabling more personalized and robust immersive adaptation under strong individual variability. Specifically, we develop a personalized federated learning (PFL) model to analyze and process the collected brain signals, which not only accommodates neurodiverse brain-signal data but also prevents the leakage of sensitive brain-signal information. To address the energy bottleneck of continual on-device learning and inference on energy-limited immersive terminals (e.g., head-mounted display), we further embed spiking neural networks (SNNs) into the PFL. By exploiting sparse, event-driven spike computation, the SNN-enabled PFL reduces the computation and energy cost of training and inference while maintaining competitive personalization performance. Experiments on real brain-signal dataset demonstrate that our method achieves the best overall identification accuracy while reducing inference energy by 6.46$\times$ compared with conventional artificial neural network-based personalized baselines.

The quality of scientific code is a critical concern for the research community. Poorly written code can result in irreproducible results, incorrect findings, and slower scientific progress. In this study, we evaluate scientific code quality across three dimensions: reproducibility, readability, and reusability. We curated a corpus of 518 code repositories by analyzing Code Availability statements from all 1239 Nature publications in 2024. To assess code quality, we employed multiple methods, including manual attempts to reproduce Jupyter notebooks, documentation reviews, and analyses of code clones and mutation patterns. Our results reveal major challenges in scientific code quality. Of the 19 notebooks we attempted to execute, only two were reproducible, primarily due to missing data files and dependency issues. Code duplication was also common, with 326 clone classes of at least 10 lines and three instances found among 637 of the 1510 notebooks in our corpus. These duplications frequently involved tasks such as visualization, data processing, and statistical analysis. Moreover, our mutation analysis showed that scientific notebooks often exhibit tangled state changes, complicating comprehension and reasoning. The prevalence of these issues -- unreproducible code, widespread duplication, and tangled state management -- underscores the need for improved tools and abstractions to help science build reproducible, readable and reusable software.

Simulation and modeling are essential in product development, integrated into the design and manufacturing process to enhance efficiency and quality. They are typically represented as complex nonlinear differential algebraic equations. The growing diversity of product requirements demands multi-task optimization, a key challenge in simulation modeling research. A dual physics-informed neural network architecture has been proposed to decouple constraints and objective functions in parametric differential algebraic equation optimization problems. Theoretical analysis shows that introducing a relaxation variable with a global error bound ensures solution equivalence between the network and optimization problem. A genetic algorithm-enhanced training framework for physics-informed neural networks improves training precision and efficiency, avoiding redundant solving of differential algebraic equations. This approach enables generalization for multi-task objectives with a single, training maintaining real-time responsiveness to product requirements.

Recent progress in artificial intelligence has encouraged numerous attempts to understand and decode human visual system from brain signals. These prior works typically align neural activity independently with semantic and perceptual features extracted from images using pre-trained vision models. However, they fail to account for two key challenges: (1) the modality gap arising from the natural difference in the information level of representation between brain signals and images, and (2) the fact that semantic and perceptual features are highly entangled within neural activity. To address these issues, we utilize hyperbolic space, which is well-suited for considering differences in the amount of information and has the geometric property that geodesics between two points naturally bend toward the origin, where the representational capacity is lower. Leveraging these properties, we propose a novel framework, Hyperbolic Feature Interpolation (HyFI), which interpolates between semantic and perceptual visual features along hyperbolic geodesics. This enables both the fusion and compression of perceptual and semantic information, effectively reflecting the limited expressiveness of brain signals and the entangled nature of these features. As a result, it facilitates better alignment between brain and visual features. We demonstrate that HyFI achieves state-of-the-art performance in zero-shot brain-to-image retrieval, outperforming prior methods with Top-1 accuracy improvements of up to +17.3% on THINGS-EEG and +9.1% on THINGS-MEG.

Automatically generating source code from natural language using large language models (LLMs) is becoming common, yet security vulnerabilities persist despite advances in fine tuning and prompting. In this work, we systematically evaluate whether multi LLM ensembles and collaborative strategies can meaningfully improve secure code generation. We present MULTI-LLMSECCODEEVAL, a framework for assessing and enhancing security across the vulnerability management lifecycle by combining multiple LLMs with static analysis and structured collaboration. Using SecLLMEval and SecLLMHolmes, we benchmark ten pipelines spanning single model, ensemble, collaborative, and hybrid designs. Our results show that ensemble pipelines augmented with static analysis improve secure code generation over single LLM baselines by up to 47.3% on SecLLMEval and 19.3% on SecLLMHolmes, while purely LLM based collaborative pipelines yield smaller gains of 8.9% to 22.3%. Hybrid pipelines that integrate ensembling, detection, and patching achieve the strongest security performance, outperforming the best ensemble baseline by 1.78% to 4.72% and collaborative baselines by 19.81% to 26.78%. Ablation studies reveal that model scale alone does not ensure security. Smaller, structured multi model ensembles consistently outperform large monolithic LLMs. Overall, our findings demonstrate that secure code does not emerge from scale, but from carefully orchestrated multi model system design.

We present PopResume, a population-representative resume dataset for causal fairness auditing of LLM- and VLM-based resume screening systems. Unlike existing benchmarks that rely on manually injected demographic information and outcome-level disparities, PopResume is grounded in population statistics and preserves natural attribute relationships, enabling path-specific effect (PSE)-based fairness evaluation. We decompose the effect of a protected attribute on resume scores into two paths: the business necessity path, mediated by job-relevant qualifications, and the redlining path, mediated by demographic proxies. This distinction allows auditors to separate legally permissible from impermissible sources of disparity. Evaluating four LLMs and four VLMs on PopResume's 60.8K resumes across five occupations, we identify five representative discrimination patterns that aggregate metrics fail to capture. Our results demonstrate that PSE-based evaluation reveals fairness issues masked by outcome-level measures, underscoring the need for causally-grounded auditing frameworks in AI-assisted hiring.

Adversarial imitation learning (AIL) achieves high-quality imitation by mitigating compounding errors in behavioral cloning (BC), but often exhibits training instability due to adversarial optimization. To avoid this issue, a class of non-adversarial Q-based imitation learning (IL) methods, represented by IQ-Learn, has emerged and is widely believed to outperform BC by leveraging online environment interactions. However, this paper revisits IQ-Learn and demonstrates that it provably reduces to BC and suffers from an imitation gap lower bound with quadratic dependence on horizon, therefore still suffering from compounding errors. Theoretical analysis reveals that, despite using online interactions, IQ-Learn uniformly suppresses the Q-values for all actions on states uncovered by demonstrations, thereby failing to generalize. To address this limitation, we introduce a primal-dual framework for distribution matching, yielding a new Q-based IL method, Dual Q-DM. The key mechanism in Dual Q-DM is incorporating Bellman constraints to propagate high Q-values from visited states to unvisited ones, thereby achieving generalization beyond demonstrations. We prove that Dual Q-DM is equivalent to AIL and can recover expert actions beyond demonstrations, thereby mitigating compounding errors. To the best of our knowledge, Dual Q-DM is the first non-adversarial IL method that is theoretically guaranteed to eliminate compounding errors. Experimental results further corroborate our theoretical results.

Conversational automatic speech recognition remains challenging due to overlapping speech, far-field noise, and varying speaker counts. While recent LLM-based systems perform well on single-speaker benchmarks, their robustness in multi-speaker settings is unclear. We systematically compare LLM-based and modular pipeline approaches along four axes: overlap robustness, semantic fidelity, speaker count, and single- versus multi-channel input. To capture meaning-altering errors that conventional metrics miss, we introduce tcpSemER, which extends tcpWER by replacing Levenshtein distance with embedding-based semantic similarity. We further decompose tcpWER into overlapping and non-overlapping components for finer-grained analysis. Experiments across three datasets show that LLM-based systems are competitive in two-speaker settings but degrade as speaker count and overlap increase, whereas modular pipelines remain more robust.

As large language models (LLMs) are trained on increasingly vast and opaque text corpora, determining which data contributed to training has become essential for copyright enforcement, compliance auditing, and user trust. While prior work focuses on detecting whether a dataset was used in training (membership inference), the complementary problem -- verifying that a dataset was not used -- has received little attention. We address this gap by introducing PRISM, a test that detects dataset-level non-membership using only grey-box access to model logits. Our key insight is that two models that have not seen a dataset exhibit higher rank correlation in their normalized token log probabilities than when one model has been trained on that data. Using this observation, we construct a correlation-based test that detects non-membership. Empirically, PRISM reliably rules out membership in training data across all datasets tested while avoiding false positives, thus offering a framework for verifying that specific datasets were excluded from LLM training.

Video-based quality assurance (QA) for long-form gameplay video is labor-intensive and error-prone, yet valuable for assessing game stability and visual correctness over extended play sessions. Vision language models (VLMs) promise general-purpose visual reasoning capabilities and thus appear attractive for detecting visual bugs directly from video frames. Recent benchmarks suggest that VLMs can achieve promising results in detecting visual glitches on curated datasets. Building on these findings, we conduct a real-world study using industrial QA gameplay videos to evaluate how well VLMs perform in practical scenarios. Our study samples keyframes from long gameplay videos and asks a VLM whether each keyframe contains a bug. Starting from a single-prompt baseline, the model achieves a precision of 0.50 and an accuracy of 0.72. We then examine two common enhancement strategies used to improve VLM performance without fine-tuning: (1) a secondary judge model that re-evaluates VLM outputs, and (2) metadata-augmented prompting through the retrieval of prior bug reports. Across \textbf{100 videos} totaling \textbf{41 hours} and \textbf{19,738 keyframes}, these strategies provide only marginal improvements over the simple baseline, while introducing additional computational cost and output variance. Our findings indicate that off-the-shelf VLMs are already capable of detecting a certain range of visual bugs in QA gameplay videos, but further progress likely requires hybrid approaches that better separate textual and visual anomaly detection.

Patient simulation is essential for developing and evaluating mental health dialogue systems. As most existing approaches rely on snapshot-style prompts with limited profile information, homogeneous behaviors and incoherent disease progression in multi-turn interactions have become key chellenges. In this work, we propose DEPROFILE, a data-grounded patient simulation framework that constructs unified, multi-source patient profiles by integrating demographic attributes, standardized clinical symptoms, counseling dialogues, and longitudinal life-event histories from real-world data. We further introduce a Chain-of-Change agent to transform noisy longitudinal records into structured, temporally grounded memory representations for simulation. Experiments across multiple large language model (LLM) backbones show that with more comprehensive profile constructed by DEPROFILE, the dialogue realism, behavioral diversity, and event richness have consistently improved and exceed state-of-the-art baselines, highlighting the importance of grounding patient simulation in verifiable longitudinal evidence.

In solving partial differential equations (PDEs), machine learning utilizing physical laws has received considerable attention owing to advantages such as mesh-free solutions, unsupervised learning, and feasibility for solving high-dimensional problems. An effective approach is based on physics-informed neural networks (PINNs), which are based on deep neural networks known for their excellent performance in various academic and industrial applications. However, PINNs struggled with model training owing to significantly slow convergence because of a spectral bias problem. In this study, we propose a PINN-based method equipped with a coordinate-encoding layer on linear grid cells. The proposed method improves the training convergence speed by separating the local domains using grid cells. Moreover, it reduces the overall computational cost by using axis-independent linear grid cells. The method also achieves efficient and stable model training by adequately interpolating the encoded coordinates between grid points using natural cubic splines, which guarantees continuous derivative functions of the model computed for the loss functions. The results of numerical experiments demonstrate the effective performance and efficient training convergence speed of the proposed method.

Fully provisioned Message Passing Interface (MPI) parallelism achieves near-optimal wall-clock time for Computational Fluid Dynamics (CFD) solvers. This work addresses a complementary question for shared, cloud-managed clusters: can fine-grained CPU provisioning reduce resource reservation of low-load subdomains, improving cluster packing efficiency without unacceptably degrading performance? We propose rank-aware resource scheduling on Kubernetes, mapping each MPI rank to a pod whose CPU request is proportional to its subdomain cell count. We also demonstrate In-Place Pod Vertical Scaling (Kubernetes v1.35 GA) for mid-simulation CPU adjustment without pod restart. Three findings emerge. First, hard CPU limits via the Linux CFS bandwidth controller cause 78x slowdown through cascading stalls at MPI_Allreduce barriers; requests-only allocation eliminates throttling entirely. Second, on non-burstable c5.xlarge instances, concentric decomposition with equal CPU is 19% faster than the Scotch baseline, while adding proportional CPU yields a further 3% improvement. Third, at 16 MPI ranks on 101K-cell meshes, proportional allocation is 20% faster than equal allocation while reducing sparse-subdomain provisioned CPU by 82%, freeing 6.5 vCPU of scheduling headroom. Experiments are conducted on AWS EC2 c5.xlarge clusters (4-16 ranks) running k3s v1.35. All scripts and data are released as open source.

Privacy-preserving semantic understanding of human activities is important for indoor sensing, yet existing Wi-Fi CSI-based systems mainly focus on pose estimation or predefined action classification rather than fine-grained language generation. Mapping CSI to natural-language descriptions remains challenging because of the semantic gap between wireless signals and language and direction-sensitive ambiguities such as left/right limb confusion. We propose WiFi2Cap, a three-stage framework for generating action captions directly from Wi-Fi CSI. A vision-language teacher learns transferable supervision from synchronized video-text pairs, and a CSI student is aligned to the teacher's visual space and text embeddings. To improve direction-sensitive captioning, we introduce a Mirror-Consistency Loss that reduces mirrored-action and left-right ambiguities during cross-modal alignment. A prefix-tuned language model then generates action descriptions from CSI embeddings. We also introduce the WiFi2Cap Dataset, a synchronized CSI-RGB-sentence benchmark for semantic captioning from Wi-Fi signals. Experimental results show that WiFi2Cap consistently outperforms baseline methods on BLEU-4, METEOR, ROUGE-L, CIDEr, and SPICE, demonstrating effective privacy-friendly semantic sensing.

Distributional metrics such as Fréchet Audio Distance cannot score individual music clips and correlate poorly with human judgments, while the only per-sample learned metric achieving high human correlation is closed-source. We introduce MUQ-EVAL, an open-source per-sample quality metric for AIgenerated music built by training lightweight prediction heads on frozen MuQ-310M features using MusicEval, a dataset of generated clips from 31 text-to-music systems with expert quality ratings. Our simplest model, frozen features with attention pooling and a two-layer MLP, achieves system-level SRCC = 0.957 and utterance-level SRCC = 0.838 with human mean opinion scores. A systematic ablation over training objectives and adaptation strategies shows that no addition meaningfully improves the frozen baseline, indicating that frozen MuQ representations already capture quality-relevant information. Encoder choice is the dominant design factor, outweighing all architectural and training decisions. LoRA-adapted models trained on as few as 150 clips already achieve usable correlation, enabling personalized quality evaluators from individual listener annotations. A controlled degradation analysis reveals selective sensitivity to signal-level artifacts but insensitivity to musical-structural distortions. Our metric, MUQ-EVAL, is fully open-source, outperforms existing open per-sample metrics, and runs in real time on a single consumer GPU. Code, model weights, and evaluation scripts are available at https://github.com/dgtql/MuQ-Eval.

Tabular data are central to biomedical research, from liquid biopsy and bulk and single-cell transcriptomics to electronic health records and phenotypic profiling. Unlike images or sequences, however, tabular datasets lack intrinsic spatial organization: features are treated as unordered dimensions, and their relationships must be inferred implicitly by the model. This limits the ability of vision architectures to exploit local structure and higher-order feature interactions in non-spatial biomedical data. Here we introduce Dynamic Feature Mapping (Dynomap), an end-to-end deep learning framework that learns a task-optimized spatial topology of features directly from data. Dynomap jointly optimizes feature placement and prediction through a fully differentiable rendering mechanism, without relying on heuristics, predefined groupings, or external priors. By transforming high-dimensional tabular vectors into learned feature maps, Dynomap enables vision-based models to operate effectively on unordered biomedical inputs. Across multiple clinical and biological datasets, Dynomap consistently outperformed classical machine learning, modern deep tabular models, and existing vector-to-image approaches. In liquid biopsy data, Dynomap organized clinically relevant gene signatures into coherent spatial patterns and improved multiclass cancer subtype prediction accuracy by up to 18%. In a Parkinson disease voice dataset, it clustered disease-associated acoustic descriptors and improved accuracy by up to 8%. Similar gains and interpretable feature organization were observed in additional biomedical datasets. These results establish Dynomap as a general strategy for bridging tabular and vision-based deep learning and for uncovering structured, clinically relevant patterns in high-dimensional biomedical data.

The standard practice in Large Language Models (LLMs) is to base predictions on the final-layer token representations. Recent studies, however, show that intermediate layers encode substantial information, which may contain more task-relevant features than the final-layer representations alone. Importantly, it was shown that for different tasks, different layers may be optimal. In this work we introduce Inter-Layer Structural Encoders (ILSE), a powerful structural approach to learn one effective representation from the LLM's internal layer representations all together. Central to ILSE is Cayley-Encoder, a mathematically grounded geometric encoder that leverages expander Cayley graphs for efficient inter-layer information propagation. We evaluate ILSE across 13 classification and semantic similarity tasks with 9 pre-trained LLMs ranging from 14 million to 8 billion parameters. ILSE consistently outperforms baselines and existing approaches, achieving up to 44% improvement in accuracy and 25% in similarity metrics. We further show that ILSE is data-efficient in few-shot regimes and can make small LLMs competitive with substantially larger models.

Out-of-distribution (OOD) detection aims to identify inputs that differ from the training distribution in order to reduce unreliable predictions by deep neural networks. Among post-hoc feature-space approaches, OOD detection is commonly performed by approximating the in-distribution support in the representation space of a pretrained network. Existing methods often reflect a trade-off between compact parametric models, such as Mahalanobis-based scores, and more flexible but reference-based methods, such as k-nearest neighbors. Bounding-box abstraction provides an attractive intermediate perspective by representing in-distribution support through compact axis-aligned summaries of hidden activations. In this paper, we introduce Bounding Box Anomaly Scoring (BBAS), a post-hoc OOD detection method that leverages bounding-box abstraction. BBAS combines graded anomaly scores based on interval exceedances, monitoring variables adapted to convolutional layers, and decoupled clustering and box construction for richer and multi-layer representations. Experiments on image-classification benchmarks show that BBAS provides robust separation between in-distribution and out-of-distribution samples while preserving the simplicity, compactness, and updateability of the bounding-box approach.

Learning system dynamics from observations is a critical problem in many applications over various real-world complex systems, e.g., climate, ecology, and fluid systems. Recently, neural dynamics modeling method have become a prevalent solution that embeds the object's observations into a latent space before learning dynamics using neural methods such as neural Ordinary Differential Equations (ODE). Existing dynamics modeling methods induce a specific model for each observation of different complex systems, resulting in poor generalization across systems. Inspired by the great success of pre-trained models, we conduct a generalized Pre-trained Dynamics EncoDER (PDEDER) which can embed the original state observations into a latent space where the dynamics can be captured more easily. To conduct the generalized PDEDER, we pre-train any Pre-trained Language Model (PLM) by minimizing the Lyapunov exponent objective, which constrains the chaotic behavior of governing dynamics learned in the latent space. By penalizing the divergence of embedded observations, our PDEDER promotes locally stable and well-structured latent dynamics, thereby facilitating more effective dynamics modeling than in the original observation space. In addition, we incorporate reconstruction and forecasting objectives to mitigate the risk of obtaining an over-smoothed latent space. Specifically, we collect 152 sets of real-world and synthetic observations from 23 complex systems as pre-training corpora and employ them to pre-train PDEDER. Given any future dynamic observation, we can fine-tune PDEDER with any specific dynamics modeling method. We evaluate PDEDER on 12 dynamic systems by short/long-term forecasting under both in-domain and cross-domain settings, and the empirical results indicate the effectiveness and generalizability of PDEDER.

The adoption of large language models (LLMs) for structured information extraction from financial documents has accelerated rapidly, yet production deployments face fundamental architectural decisions with limited empirical guidance. We present a systematic benchmark comparing four multi-agent orchestration architectures: sequential pipeline, parallel fan-out with merge, hierarchical supervisor-worker and reflexive self-correcting loop. These are evaluated across five frontier and open-weight LLMs on a corpus of 10,000 SEC filings (10-K, 10-Q and 8-K forms). Our evaluation spans 25 extraction field types covering governance structures, executive compensation and financial metrics, measured along five axes: field-level F1, document-level accuracy, end-to-end latency, cost per document and token efficiency. We find that reflexive architectures achieve the highest field-level F1 (0.943) but at 2.3x the cost of sequential baselines, while hierarchical architectures occupy the most favorable position on the cost-accuracy Pareto frontier (F1 0.921 at 1.4x cost). We further present ablation studies on semantic caching, model routing and adaptive retry strategies, demonstrating that hybrid configurations can recover 89\% of the reflexive architecture's accuracy gains at only 1.15x baseline cost. Our scaling analysis from 1K to 100K documents per day reveals non-obvious throughput-accuracy degradation curves that inform capacity planning. These findings provide actionable guidance for practitioners deploying multi-agent LLM systems in regulated financial environments.

There are different goals for literature research, from understanding an unfamiliar topic to generate hypothesis for the next research project. The nature of literature research also varies according to user's familiarity level of the topic. For inexperienced researchers, identifying gaps in the existing literature and generating feasible hypothesis are crucial but challenging. While general ``deep research'' tools can be used, they are not designed for such use case, thus often not effective. In addition, the ``black box" nature and hallucination of Large Language Models (LLMs) often lead to distrust. In this paper, we introduce a human-agent collaborative visualization system AwesomeLit to address this need. It has several novel features: a transparent user-steerable agentic workflow; a dynamically generated query exploring tree, visualizing the exploration path and provenance; and a semantic similarity view, depicting the relationships between papers. It enables users to transition from general intentions to detailed research topics. Finally, a qualitative study involving several early researchers showed that AwesomeLit is effective in helping users explore unfamiliar topics, identify promising research directions, and improve confidence in research results.

We consider a PAC-Bayes type learning rule for binary classification, balancing the training error of a randomized ''posterior'' predictor with its KL divergence to a pre-specified ''prior''. This can be seen as an extension of a modified two-part-code Minimum Description Length (MDL) learning rule, to continuous priors and randomized predictions. With a balancing parameter of $λ=1$ this learning rule recovers an (empirical) Bayes posterior and a modified variant recovers the profile posterior, linking with standard Bayesian prediction (up to the treatment of the single-parameter noise level). However, from a risk-minimization prediction perspective, this Bayesian predictor overfits and can lead to non-vanishing excess loss in the agnostic case. Instead a choice of $λ\gg 1$, which can be seen as using a sample-size-dependent-prior, ensures uniformly vanishing excess loss even in the agnostic case. We precisely characterize the effect of under-regularizing (and over-regularizing) as a function of the balance parameter $λ$, understanding the regimes in which this under-regularization is tempered or catastrophic. This work extends previous work by Zhu and Srebro [2025] that considered only discrete priors to PAC Bayes type learning rules and, through their rigorous Bayesian interpretation, to Bayesian prediction more generally.

Language barriers affect 27.3 million U.S. residents with non-English language preference, yet professional medical translation remains costly and often unavailable. We evaluated four frontier large language models (GPT-5.1, Claude Opus 4.5, Gemini 3 Pro, Kimi K2) translating 22 medical documents into 8 languages spanning high-resource (Spanish, Chinese, Russian, Vietnamese), medium-resource (Korean, Arabic), and low-resource (Tagalog, Haitian Creole) categories using a five-layer validation framework. Across 704 translation pairs, all models achieved high semantic preservation (LaBSE greater than 0.92), with no significant difference between high- and low-resource languages (p = 0.066). Cross-model back-translation confirmed results were not driven by same-model circularity (delta = -0.0009). Inter-model concordance across four independently trained models was high (LaBSE: 0.946), and lexical borrowing analysis showed no correlation between English term retention and fidelity scores in low-resource languages (rho = +0.018, p = 0.82). These converging results suggest frontier LLMs preserve medical meaning across resource levels, with implications for language access in healthcare.

Productive human-AI collaboration requires appropriate reliance, yet contemporary AI systems are often miscalibrated, exhibiting systematic overconfidence or underconfidence. We investigate whether humans can learn to mentally recalibrate AI confidence signals through repeated experience. In a behavioral experiment (N = 200), participants predicted the AI's correctness across four AI calibration conditions: standard, overconfidence, underconfidence, and a counterintuitive "reverse confidence" mapping. Results demonstrate robust learning across all conditions, with participants significantly improving their accuracy, discrimination, and calibration alignment over 50 trials. We present a computational model utilizing a linear-in-log-odds (LLO) transformation and a Rescorla-Wagner learning rule to explain these dynamics. The model reveals that humans adapt by updating their baseline trust and confidence sensitivity, using asymmetric learning rates to prioritize the most informative errors. While humans can compensate for monotonic miscalibration, we identify a significant boundary in the reverse confidence scenario, where a substantial proportion of participants struggled to override initial inductive biases. These findings provide a mechanistic account of how humans adapt their trust in AI confidence signals through experience.

Retrieval-Augmented Generation (RAG) systems for biomedical literature are typically evaluated using ranking metrics like Mean Reciprocal Rank (MRR), which measure how well the system identifies the single most relevant chunk. We argue that for full-text scientific documents, this paradigm is incomplete: it rewards retrieval precision while ignoring retrieval breadth -- the ability to surface evidence from across a document's structural sections. We propose GraLC-RAG, a framework that unifies late chunking with graph-aware structural intelligence, introducing structure-aware chunk boundary detection, UMLS knowledge graph infusion, and graph-guided hybrid retrieval. We evaluate six strategies on 2,359 IMRaD-filtered PubMed Central articles using 2,033 cross-section questions and two metric families: standard ranking metrics (MRR, Recall@k) and structural coverage metrics (SecCov@k, CS Recall). Our results expose a sharp divergence: content-similarity methods achieve the highest MRR (0.517) but always retrieve from a single section, while structure-aware methods retrieve from up to 15.6x more sections. Generation experiments show that KG-infused retrieval narrows the answer-quality gap to delta-F1 = 0.009 while maintaining 4.6x section diversity. These findings demonstrate that standard metrics systematically undervalue structural retrieval and that closing the multi-section synthesis gap is a key open problem for biomedical RAG.

Adapting pretrained language models to low-resource, morphologically rich languages remains a significant challenge. Existing vocabulary expansion methods typically rely on arbitrarily segmented subword units, resulting in fragmented lexical representations and loss of critical morphological information. To address this limitation, we propose the Lexically Grounded Subword Embedding Initialization (LGSE) framework, which introduces morphologically informed segmentation for initializing embeddings of novel tokens. Instead of using random vectors or arbitrary subwords, LGSE decomposes words into their constituent morphemes and constructs semantically coherent embeddings by averaging pretrained subword or FastText-based morpheme representations. When a token cannot be segmented into meaningful morphemes, its embedding is constructed using character n-gram representations to capture structural information. During Language-Adaptive Pretraining, we apply a regularization term that penalizes large deviations of newly introduced embeddings from their initialized values, preserving alignment with the original pretrained embedding space while enabling adaptation to the target language. To isolate the effect of initialization, we retain the original pre-trained model vocabulary and tokenizer and update only the new embeddings during adaptation. We evaluate LGSE on three NLP tasks: Question Answering, Named Entity Recognition, and Text Classification, in two morphologically rich, low-resource languages: Amharic and Tigrinya, where morphological segmentation resources are available. Experimental results show that LGSE consistently outperforms baseline methods across all tasks, demonstrating the effectiveness of morphologically grounded embedding initialization for improving representation quality in underrepresented languages. Project resources are available in the GitHub link.

Attribution maps for semantic segmentation are almost always judged by visual plausibility. Yet looking convincing does not guarantee that the highlighted pixels actually drive the model's prediction, nor that attribution credit stays within the target region. These questions require a dedicated evaluation protocol. We introduce a reproducible benchmark that tests intervention-based faithfulness, off-target leakage, perturbation robustness, and runtime on Pascal VOC and SBD across three pretrained backbones. To further demonstrate the benchmark, we propose Dual-Evidence Attribution (DEA), a lightweight correction that fuses gradient evidence with region-level intervention signals through agreement-weighted fusion. DEA increases emphasis where both sources agree and retains causal support when gradient responses are unstable. Across all completed runs, DEA consistently improves deletion-based faithfulness over gradient-only baselines and preserves strong robustness, at the cost of additional compute from intervention passes. The benchmark exposes a faithfulness-stability tradeoff among attribution families that is entirely hidden under visual evaluation, providing a foundation for principled method selection in segmentation explainability. Code is available at https://github.com/anmspro/DEA.

Vision-language models (VLMs) adapted to the medical domain have shown strong performance on visual question answering benchmarks, yet their robustness against two critical failure modes, hallucination and sycophancy, remains poorly understood, particularly in combination. We evaluate six VLMs (three general-purpose, three medical-specialist) on three medical VQA datasets and uncover a grounding-sycophancy tradeoff: models with the lowest hallucination propensity are the most sycophantic, while the most pressure-resistant model hallucinates more than all medical-specialist models. To characterize this tradeoff, we propose three metrics: L-VASE, a logit-space reformulation of VASE that avoids its double-normalization; CCS, a confidence-calibrated sycophancy score that penalizes high-confidence capitulation; and Clinical Safety Index (CSI), a unified safety index that combines grounding, autonomy, and calibration via a geometric mean. Across 1,151 test cases, no model achieves a CSI above 0.35, indicating that none of the evaluated 7-8B parameter VLMs is simultaneously well-grounded and robust to social pressure. Our findings suggest that joint evaluation of both properties is necessary before these models can be considered for clinical use. Code is available at https://github.com/UTSA-VIRLab/AgreeOrRight

Despite recent advances in population-based structural health monitoring (PBSHM), knowledge transfer between highly-disparate structures (i.e., heterogeneous populations) remains a challenge. The current work proposes that heterogeneous transfer may be accomplished via intermediate structures that bridge the gap in information between the structures of interest. A key aspect of the technique is the idea that by varying parameters such as material properties and geometry, one structure can be continuously morphed into another. The approach is demonstrated via a case study involving the parameterisation of (and transfer between) simulated heterogeneous bridge designs (Case 1). Transfer between simplified physical representations of a 'bridge' and 'aeroplane' is then demonstrated in Case 2, via a chain of finite-element models. The facetious question 'When is a bridge not an aeroplane?' has been previously asked in the context of predicting positive transfer based on structural similarity. While the obvious answer to this question is 'Always,' the results presented in the current paper show that, in some cases, positive transfer can indeed be achieved between highly-disparate systems.

Causal discovery is challenging in general dynamical systems because, without strong structural assumptions, the underlying causal graph may not be identifiable even from interventional data. However, many real-world systems exhibit directional, cascade-like structure, in which components activate sequentially and upstream failures suppress downstream effects. We study causal discovery in such chain-reaction systems and show that the causal structure is uniquely identifiable from blocking interventions that prevent individual components from activating. We propose a minimal estimator with finite-sample guarantees, achieving exponential error decay and logarithmic sample complexity. Experiments on synthetic models and diverse chain-reaction environments demonstrate reliable recovery from a few interventions, while observational heuristics fail in regimes with delayed or overlapping causal effects.

LLMs often generate seemingly valid answers to flawed or ill-posed inputs. This is not due to missing knowledge: under discriminative prompting, the same models can mostly identify such issues, yet fail to reflect this in standard generative responses. This reveals a fundamental know-act gap between discriminative recognition and generative behavior. Prior work largely characterizes this issue in narrow settings, such as math word problems or question answering, with limited focus on how to integrate these two modes. In this work, we present a comprehensive analysis using FaultyScience, a newly constructed large-scale, cross-disciplinary benchmark of faulty scientific questions. We show that the gap is pervasive and stems from token-level autoregression, which entangles task selection (validate vs. answer) with content generation, preventing discriminative knowledge from being utilized. To address this, we propose DeIllusionLLM, a task-level autoregressive framework that explicitly models this decision. Through self-distillation, the model unifies discriminative judgment and generative reasoning within a single backbone. Empirically, DeIllusionLLM substantially reduces answer-despite-error failures under natural prompting while maintaining general reasoning performance, demonstrating that self-distillation is an effective and scalable solution for bridging the discriminative-generative know-act gap

Consumers are generally resistant to Artificial Intelligence (AI) involvement in moral decision-making, perceiving moral agency as requiring uniquely human traits. This research investigates whether consumers might instead accept AIs in the role of moral compliance, where AI upholds pre-existing moral norms without exercising subjective discretion. Across five studies this research shows that consumers evaluate AI more positively than human agents in moral compliance roles. The findings reveal that this preference arises from inferences of AI's lack of ulterior motives, which are often attributed to human agents. While previous studies have focused on AI as a decision-maker, this work demonstrates the critical role of upholding pre-existing rules, a role in which AI is perceived to excel. These findings contribute to understanding consumer acceptance of moral AI and provide actionable insights for organizations seeking to leverage AI in ethical oversight. By positioning AI as a moral compliance agent, companies can address consumer skepticism, enhance trust, and improve perceptions of corporate ethicality.

Users often rely on Large Language Models (LLMs) for processing multiple documents or performing analysis over a number of instances. For example, analysing the overall sentiment of a number of movie reviews requires an LLM to process the sentiment of each review individually in order to provide a final aggregated answer. While LLM performance on such individual tasks is generally high, there has been little research on how LLMs perform when dealing with multi-instance inputs. In this paper, we perform a comprehensive evaluation of the multi-instance processing (MIP) ability of LLMs for tasks in which they excel individually. The results show that all LLMs follow a pattern of slight performance degradation for small numbers of instances (approximately 20-100), followed by a performance collapse on larger instance counts. Crucially, our analysis shows that while context length is associated with this degradation, the number of instances has a stronger effect on the final results. This finding suggests that when optimising LLM performance for MIP, attention should be paid to both context length and, in particular, instance count.

Simulation remains a cornerstone of computer architecture research, yet full end-to-end application execution is prohibitively time-consuming. The industry-standard solution, SimPoint, mitigates this cost by selecting a small number of representative code regions that capture program phase behavior. In this work, we take a fresh look at phase behavior in the SPEC CPU 2017 Integer suite to assess how pronounced such behavior truly is and what accuracy can be expected from typical SimPoint usage. Based on previously published data, we argue that common SimPoint counts can induce substantial estimation errors. To explore this further, we recast SimPoint as a stratified sampling problem, which enables the derivation of a conservative confidence interval. The analysis indicates that significant errors are expected, and our empirical analysis confirms this: with 20 SimPoints, two applications exhibit 40-60% performance prediction error. We decompose SimPoint into its two fundamental components - stratification (clustering) and sample-unit selection (centroid choice) - and analyze their individual effects on accuracy. We then extend the approach into a two-phase (double) sampling scheme, in which a large preliminary random sample enables improved stratification and more representative region selection. Using this method, the maximum per-application error drops to 3%. Finally, we demonstrate that the proposed two-phase stratified framework achieves an order-of-magnitude reduction in required sample size compared to simple random sampling while maintaining a tight analytical confidence interval, suggesting a practical path toward statistically grounded and efficient architectural simulation.

Computer system simulation studies routinely rely on executing a limited number of short application regions, since full end-to-end simulation is prohibitively time-consuming. To preserve representativeness, existing methods employ either random sampling or phase-based characterization to identify representative regions. In this work, we revisit random sampling in the context of computer architecture simulation. To assess how the confidence level varies with different micro-architectural configurations, we examine how the sample standard deviation relates to the sample mean. We show that the ranked set sampling (RSS) technique - well established in the statistical literature - maps naturally to architectural simulation and yields significantly tighter confidence intervals than simple random sampling. Across our experiments, RSS reduces the confidence interval width by up to 50%. We further introduce a repeated subsampling scheme that identifies representative simulation regions for future studies. For a fixed sample size, this approach reduces the maximum observed error from 35% to 10%. Evaluating two selection criteria, we find that more informed subsample selection provides additional accuracy gains. Overall, our method achieves an average error below 2% and a maximum error of 3.5% across individual SPEC CPU 2017 Integer applications when simulating 30 regions of 1 million instructions each.

Sparse Autoencoders uncover thousands of features in vision models, yet explaining these features without requiring human intervention remains an open challenge. While previous work has proposed generating correlation-based explanations based on top activating input examples, we present a fundamentally different alternative based on causal interventions. We leverage the structure of Vision-Language Models and steer individual SAE features in the vision encoder after providing an empty image. Then, we prompt the language model to explain what it ``sees'', effectively eliciting the visual concept represented by each feature. Results show that Steering offers an scalable alternative that complements traditional approaches based on input examples, serving as a new axis for automated interpretability in vision models. Moreover, the quality of explanations improves consistently with the scale of the language model, highlighting our method as a promising direction for future research. Finally, we propose Steering-informed Top-k, a hybrid approach that combines the strengths of causal interventions and input-based approaches to achieve state-of-the-art explanation quality without additional computational cost.

With the increasing deployment of automated and agentic systems, ensuring the adversarial robustness of automatic speech recognition (ASR) models has become critical. We observe that changing the precision of an ASR model during inference reduces the likelihood of adversarial attacks succeeding. We take advantage of this fact to make the models more robust by simple random sampling of the precision during prediction. Moreover, the insight can be turned into an adversarial example detection strategy by comparing outputs resulting from different precisions and leveraging a simple Gaussian classifier. An experimental analysis demonstrates a significant increase in robustness and competitive detection performance for various ASR models and attack types.

Teachers face growing pressure to integrate AI tools into their classrooms, yet are rarely positioned as agentic decision-makers in this process. Understanding the criteria teachers use to evaluate AI tools, and the conditions that support such reasoning, is essential for responsible AI integration. We address this gap through a two-day national summit in which 61 U.S. K-12 mathematics educators developed personal rubrics for evaluating AI classroom tools. The summit was designed to support deliberative sensemaking, a process we conceptualize by integrating Technological Pedagogical Content Knowledge (TPACK) with deliberative agency. Teachers generated over 200 criteria - initial articulations spanning four higher-order themes (Practical, Equitable, Flexible, and Rigorous) - that addressed both AI outputs and the process of using AI. Criteria contained productive tensions (e.g., personalization versus fairness, adaptability versus efficiency), and the vast majority framed AI as an assistant rather than a coaching tool for professional learning. Analysis of surveys, interviews, and summit discussions revealed five mechanisms supporting deliberative sensemaking: time and space for deliberation, artifact-centered sensemaking, collaborative reflection through diverse viewpoints, knowledge-building, and psychological safety. Across these mechanisms, TPACK and agency operated in a mutually reinforcing cycle - knowledge-building enabled more grounded evaluative judgment, while the act of constructing criteria deepened teachers' understanding of tools. We discuss implications for edtech developers seeking practitioner input, school leaders making adoption decisions, educators and professional learning designers, and researchers working to elicit teachers' evaluative reasoning about rapidly evolving technologies.

As AI agents become the primary consumers of retrieval APIs, there is an opportunity to expose more of the retrieval pipeline to the caller. flexvec is a retrieval kernel that exposes the embedding matrix and score array as a programmable surface, allowing arithmetic operations on both before selection. We refer to composing operations on this surface at query time as Programmatic Embedding Modulation (PEM). This paper describes a set of such operations and integrates them into a SQL interface via a query materializer that facilitates composable query primitives. On a production corpus of 240,000 chunks, three composed modulations execute in 19 ms end-to-end on a desktop CPU without approximate indexing. At one million chunks, the same operations execute in 82 ms.

In-context learning (ICL) allows a model to adapt at inference time by conditioning on examples rather than updating parameters. Existing time-series foundation models use implicit positional context, retrieval, or task-specific objectives, but rarely explicit instruction-conditioned demonstrations. We present a foundation model for instruction-conditioned in-context time-series tasks based on a quantile-regression T5 encoder-decoder. Historical examples and queries are encoded with a structured tokenization scheme that marks target series, covariates, context, and task-specific future information. A hierarchical Transformer with per-example encoding, example-level fusion, and cross-example attention conditions decoding on demonstration pairs, enabling forecasting and related tasks without task-specific fine-tuning. We train on large-scale real and synthetic time series using supervised forecasting plus self-supervised tasks, including imputation, reconstruction, classification, anomaly detection, and source demixing. This multi-task training learns a distribution over task mappings and improves adaptation to local structure at inference time. Across diverse datasets, frequencies, and horizons, our method outperforms strong foundation baselines on point and probabilistic forecasting benchmarks, including fev-bench and GIFT-Eval, while remaining competitive on classification and anomaly detection.

Chain-of-thought (CoT) reasoning has been proposed as a transparency mechanism for large language models in safety-critical deployments, yet its effectiveness depends on faithfulness (whether models accurately verbalize the factors that actually influence their outputs), a property that prior evaluations have examined in only two proprietary models, finding acknowledgment rates as low as 25% for Claude 3.7 Sonnet and 39% for DeepSeek-R1. To extend this evaluation across the open-weight ecosystem, this study tests 12 open-weight reasoning models spanning 9 architectural families (7B-685B parameters) on 498 multiple-choice questions from MMLU and GPQA Diamond, injecting six categories of reasoning hints (sycophancy, consistency, visual pattern, metadata, grader hacking, and unethical information) and measuring the rate at which models acknowledge hint influence in their CoT when hints successfully alter answers. Across 41,832 inference runs, overall faithfulness rates range from 39.7% (Seed-1.6-Flash) to 89.9% (DeepSeek-V3.2-Speciale) across model families, with consistency hints (35.5%) and sycophancy hints (53.9%) exhibiting the lowest acknowledgment rates. Training methodology and model family predict faithfulness more strongly than parameter count, and keyword-based analysis reveals a striking gap between thinking-token acknowledgment (approximately 87.5%) and answer-text acknowledgment (approximately 28.6%), suggesting that models internally recognize hint influence but systematically suppress this acknowledgment in their outputs. These findings carry direct implications for the viability of CoT monitoring as a safety mechanism and suggest that faithfulness is not a fixed property of reasoning models but varies systematically with architecture, training method, and the nature of the influencing cue.

Large Language Models (LLMs) have demonstrated potential in code generation, yet they struggle with the multi-step, stateful reasoning required for offensive cybersecurity operations. Existing research often relies on static benchmarks that fail to capture the dynamic nature of real-world vulnerabilities. In this work, we introduce STRIATUM-CTF (A Search-based Test-time Reasoning Inference Agent for Tactical Utility Maximization in Cybersecurity), a modular agentic framework built upon the Model Context Protocol (MCP). By standardizing tool interfaces for system introspection, decompilation, and runtime debugging, STRIATUM-CTF enables the agent to maintain a coherent context window across extended exploit trajectories. We validate this approach not merely on synthetic datasets, but in a live competitive environment. Our system participated in a university-hosted Capture-the-Flag (CTF) competition in late 2025, where it operated autonomously to identify and exploit vulnerabilities in real-time. STRIATUM-CTF secured First Place, outperforming 21 human teams and demonstrating strong adaptability in a dynamic problem-solving setting. We analyze the agent's decision-making logs to show how MCP-based tool abstraction significantly reduces hallucination compared to naive prompting strategies. These results suggest that standardized context protocols are a critical path toward robust autonomous cyber-reasoning systems.

We introduce CAPITU, a benchmark for evaluating instruction-following capabilities of Large Language Models (LLMs) in Brazilian Portuguese. Unlike existing benchmarks that focus on English or use generic prompts, CAPITU contextualizes all tasks within eight canonical works of Brazilian literature, combining verifiable instruction constraints with culturally-grounded content. The benchmark comprises 59 instruction types organized into seven categories, all designed to be automatically verifiable without requiring LLM judges or human evaluation. Instruction types include Portuguese-specific linguistic constraints (word termination patterns like -ando/-endo/-indo, -inho/-inha, -mente) and structural requirements. We evaluate 18 state-of-the-art models across single-turn and multi-turn settings. Our results show that frontier reasoning models achieve strong performance (GPT-5.2 with reasoning: 98.5% strict accuracy), while Portuguese-specialized models offer competitive cost-efficiency (Sabiazinho-4: 87.0% at \$0.13 vs Claude-Haiku-4.5: 73.5% at \$1.12). Multi-turn evaluation reveals significant variation in constraint persistence, with conversation-level accuracy ranging from 60% to 96% across models. We identify specific challenges in morphological constraints, exact counting, and constraint persistence degradation across turns. We release the complete benchmark, evaluation code, and baseline results to facilitate research on instruction-following in Portuguese.

Large language models (LLMs) are increasingly deployed as autonomous agents in financial trading. However, they often exhibit a hazardous behavioral bias that we term uniform trust, whereby retrieved information is implicitly assumed to be factual and heterogeneous sources are treated as equally informative. This assumption stands in sharp contrast to human decision-making, which relies on selective filtering, cross-validation, and experience-driven weighting of information sources. As a result, LLM-based trading systems are particularly vulnerable to multi-source noise and misinformation, amplifying factual hallucinations and leading to unstable risk-return performance. To bridge this behavioral gap, we introduce TrustTrade (Trust-Rectified Unified Selective Trader), a multi-agent selective consensus framework inspired by human epistemic heuristics. TrustTrade replaces uniform trust with cross-agent consistency by aggregating information from multiple independent LLM agents and dynamically weighting signals based on their semantic and numerical agreement. Consistent signals are prioritized, while divergent, weakly grounded, or temporally inconsistent inputs are selectively discounted. To further stabilize decision-making, TrustTrade incorporates deterministic temporal signals as reproducible anchors and a reflective memory mechanism that adapts risk preferences at test time without additional training. Together, these components suppress noise amplification and hallucination-driven volatility, yielding more stable and risk-aware trading behavior. Across controlled backtesting in high-noise market environments (2024 Q1 and 2026 Q1), the proposed TrustTrade calibrates LLM trading behavior from extreme risk-return regimes toward a human-aligned, mid-risk and mid-return profile.

The 2022 U.S. Supreme Court decision in Dobbs v. Jackson Women's Health Organization reshaped the reproductive rights landscape, introducing new uncertainty and barriers to abortion access. We present a large-scale computational analysis of abortion discourse on Reddit, examining how barriers to access are articulated across information-seeking and information-sharing behaviors, different stages of abortion (before, during, after), and three phases of the Dobbs decision in 2022. Drawing on more than 17,000 posts from four abortion-related subreddits, we employed a multi-step pipeline to classify posts by information type, abortion stage, barrier category, and expressed emotions. Using a codebook of eight barrier types, including legal, financial, emotional, and social obstacles, we analyzed their associations with emotions and information behaviors. Topic modeling of model-generated barrier rationales further revealed how discourse evolved in response to shifting legal and cultural contexts. Our findings show that emotional and psychological barriers consistently dominate abortion narratives online, with emotions such as nervousness, confusion, fear, and sadness prevalent across discourse. By linking information behaviors, barriers, emotions, and temporal dynamics, this study provides a multi-dimensional account of how abortion is navigated in online communities.

Understanding cellular trajectories via time-resolved single-cell transcriptomics is vital for studying development, regeneration, and disease. A key challenge is inferring continuous trajectories from discrete snapshots. Biological complexity stems from stochastic cell fate decisions, temporal proliferation changes, and spatial environmental influences. Current methods often use deterministic interpolations treating cells in isolation, failing to capture the probabilistic branching, population shifts, and niche-dependent signaling driving real biological processes. We introduce Manifold Interpolating Optimal-Transport Flow (MIOFlow) 2.0. This framework learns biologically informed cellular trajectories by integrating manifold learning, optimal transport, and neural differential equations. It models three core processes: (1) stochasticity and branching via Neural Stochastic Differential Equations; (2) non-conservative population changes using a learned growth-rate model initialized with unbalanced optimal transport; and (3) environmental influence through a joint latent space unifying gene expression with spatial features like local cell type composition and signaling. By operating in a PHATE-distance matching autoencoder latent space, MIOFlow 2.0 ensures trajectories respect the data's intrinsic geometry. Empirical comparisons show expressive trajectory learning via neural differential equations outperforms existing generative models, including simulation-free flow matching. Validated on synthetic datasets, embryoid body differentiation, and spatially resolved axolotl brain regeneration, MIOFlow 2.0 improves trajectory accuracy and reveals hidden drivers of cellular transitions, like specific signaling niches. MIOFlow 2.0 thus bridges single-cell and spatial transcriptomics to uncover tissue-scale trajectories.

Preference-based fine-tuning has become an important component in training large language models, and the data used at this stage may contain sensitive user information. A central question is how to design a differentially private pipeline that is well suited to the distinct structure of reinforcement learning from human feedback. We propose a privacy-preserving framework that imposes differential privacy only on reward learning and derives the final policy from the resulting private reward model. Theoretically, we study the suboptimality gap and show that privacy contributes an additional additive term beyond the usual non-private statistical error. We also establish a minimax lower bound and show that the dominant term changes with sample size and privacy level, which in turn characterizes regimes in which the upper bound is rate-optimal up to logarithmic factors. Empirically, synthetic experiments confirm the scaling predicted by the theory, and experiments on the Anthropic HH-RLHF dataset using the Gemma-2B-IT model show stronger private alignment performance than existing differentially private baseline methods across privacy budgets.

This paper asks whether a bounded neural architecture can exhibit a meaningful division of labor between intuition and deliberation on a classic 64-item syllogistic reasoning benchmark. More broadly, the benchmark is relevant to ongoing debates about world models and multi-stage reasoning in AI. It provides a controlled setting for testing whether a learned system can develop structured internal computation rather than only one-shot associative prediction. Experiment 1 evaluates a direct neural baseline for predicting full 9-way human response distributions under 5-fold cross-validation. Experiment 2 introduces a bounded dual-path architecture with separate intuition and deliberation pathways, motivated by computational mental-model theory (Khemlani & Johnson-Laird, 2022). Under cross-validation, bounded intuition reaches an aggregate correlation of r = 0.7272, whereas bounded deliberation reaches r = 0.8152, and the deliberation advantage is significant across folds (p = 0.0101). The largest held-out gains occur for NVC, Eca, and Oca, suggesting improved handling of rejection responses and c-a conclusions. A canonical 80:20 interpretability run and a five-seed stability sweep further indicate that the deliberation pathway develops sparse, differentiated internal structure, including an Oac-leaning state, a dominant workhorse state, and several weakly used or unused states whose exact indices vary across runs. These findings are consistent with reasoning-like internal organization under bounded conditions, while stopping short of any claim that the model reproduces full sequential processes of model construction, counterexample search, and conclusion revision.

Generating synthetic populations from aggregate statistics is a core component of microsimulation, agent-based modeling, policy analysis, and privacy-preserving data release. Beyond classical census marginals, many applications require matching heterogeneous unary, binary, and ternary constraints derived from surveys, expert knowledge, or automatically extracted descriptions. Constructing populations that satisfy such multi-way constraints simultaneously poses a significant computational challenge. We consider populations where each individual is described by categorical attributes and the target is a collection of global frequency constraints over attribute combinations. Exact formulations scale poorly as the number and arity of constraints increase, especially when the constraints are numerous and overlapping. Grounded in methods from statistical physics, we propose a maximum-entropy relaxation of this problem. Multi-way cardinality constraints are matched in expectation rather than exactly, yielding an exponential-family distribution over complete population assignments and a convex optimization problem over Lagrange multipliers. We evaluate the approach on NPORS-derived scaling benchmarks with 4 to 40 attributes and compare it primarily against generalized raking. The results show that MaxEnt becomes increasingly advantageous as the number of attributes and ternary interactions grows, while raking remains competitive on smaller, lower-arity instances.

When we think of how we use smartphones, e-commerce, collaboration platforms, LLMs, etc., most of our interactions with computers are interactive and often urgent. Similar trends of interactivity and urgency are coming to HPC, with applications from simulations to data analysis and machine learning requiring more parallel computational capability and more interactivity. This chapter overviews the progress made so far along with some vectors of what the path forward will bring for greater integration of interactive and urgent HPC policies, techniques, and technologies into our HPC ecosystems.

Cycle-accurate simulators are widely used to study systolic accelerators, yet their accuracy and usability are often limited by weak validation against real hardware and poor integration with modern ML compiler stacks. This paper presents SCALE-Sim TPU, a validated and extended version of SCALE-Sim v3 for TPU-style accelerators. Specifically, we make three contributions: (1) We validate SCALE-Sim's systolic GEMM model against measurements on Google TPU v4 and show that simulated cycle counts exhibit a strong linear correlation with hardware latency, enabling a simple cycle-to-latency mapping. (2) We introduce lightweight learned latency models for non-systolic elementwise operations, achieving median relative errors below 3 percent using only tensor size and shape, substantially improving end-to-end latency estimation. (3) We integrate a StableHLO-based frontend that allows workloads from modern ML frameworks such as JAX and PyTorch to be simulated directly via a unified compiler IR. Together, these contributions improve the fidelity, coverage, and practicality of cycle-accurate simulation for whole-model performance analysis on TPUs.

Unstructured Electronic Health Record (EHR) data, such as clinical notes, contain clinical contextual observations that are not directly reflected in structured data fields. This additional information can substantially improve model learning. However, due to their unstructured nature, these data are often unavailable or impractical to use when deploying a model. We introduce a multimodal learning framework that leverages unstructured EHR data during training while producing a model that can be deployed using only structured EHR data. Using a cohort of 3,466 children evaluated for late talking, we generated note embeddings with BioClinicalBERT and encoded structured embeddings from demographics and medical codes. A note-based teacher model and a structured-only student model were jointly trained using contrastive learning and contrastive knowledge distillation loss, producing a strong classifier (AUROC = 0.985). Our proposed model reached an AUROC of 0.705, outperforming the structured-only baseline of 0.656. These results demonstrate that incorporating unstructured data during training enhances the model's capacity to identify task-relevant information within structured EHR data, enabling a deployable structured-only phenotype model.

Multimodal AI agents are increasingly automating complex real-world workflows that involve online web execution. However, current web-agent benchmarks suffer from a critical limitation: they focus entirely on web-based interaction and perception, lacking grounding in the user's real-world physical surroundings. This limitation prevents evaluation in crucial scenarios, such as when an agent must use egocentric visual perception (e.g., via AR glasses) to recognize an object in the user's surroundings and then complete a related task online. To address this gap, we introduce Ego2Web, the first benchmark designed to bridge egocentric video perception and web agent execution. Ego2Web pairs real-world first-person video recordings with web tasks that require visual understanding, web task planning, and interaction in an online environment for successful completion. We utilize an automatic data-generation pipeline combined with human verification and refinement to curate well-constructed, high-quality video-task pairs across diverse web task types, including e-commerce, media retrieval, knowledge lookup, etc. To facilitate accurate and scalable evaluation for our benchmark, we also develop a novel LLM-as-a-Judge automatic evaluation method, Ego2WebJudge, which achieves approximately 84% agreement with human judgment, substantially higher than existing evaluation methods. Experiments with diverse SoTA agents on our Ego2Web show that their performance is weak, with substantial headroom across all task categories. We also conduct a comprehensive ablation study on task design, highlighting the necessity of accurate video understanding in the proposed task and the limitations of current agents. We hope Ego2Web can be a critical new resource for developing truly capable AI assistants that can seamlessly see, understand, and act across the physical and digital worlds.

Large Language Models (LLMs) combined with Retrieval-Augmented Generation (RAG) and knowledge graphs offer new opportunities for interacting with engineering diagrams such as Piping and Instrumentation Diagrams (P&IDs). However, directly processing raw images or smart P&ID files with LLMs is often costly, inefficient, and prone to hallucinations. This work introduces ChatP&ID, an agentic framework that enables grounded and cost-effective natural-language interaction with P&IDs using Graph Retrieval-Augmented Generation (GraphRAG), a paradigm we refer to as GraphRAG for engineering diagrams. Smart P&IDs encoded in the DEXPI standard are transformed into structured knowledge graphs, which serve as the basis for graph-based retrieval and reasoning by LLM agents. This approach enables reliable querying of engineering diagrams while significantly reducing computational cost. Benchmarking across commercial LLM APIs (OpenAI, Anthropic) demonstrates that graph-based representations improve accuracy by 18% over raw image inputs and reduce token costs by 85% compared to directly ingesting smart P&ID files. While small open-source models still struggle to interpret knowledge graph formats and structured engineering data, integrating them with VectorRAG and PathRAG improves response accuracy by up to 40%. Notably, GPT-5-mini combined with ContextRAG achieves 91% accuracy at a cost of only $0.004 per task. The resulting ChatP&ID interface enables intuitive natural-language interaction with complex engineering diagrams and lays the groundwork for AI-assisted process engineering tasks such as Hazard and Operability Studies (HAZOP) and multi-agent analysis.

Operator learning models are rapidly emerging as the predictive core of digital twins for nuclear and energy systems, promising real-time field reconstruction from sparse sensor measurements. Yet their robustness to adversarial perturbations remains uncharacterized, a critical gap for deployment in safety-critical systems. Here we show that neural operators are acutely vulnerable to extremely sparse (fewer than 1% of inputs), physically plausible perturbations that exploit their sensitivity to boundary conditions. Using gradient-free differential evolution across four operator architectures, we demonstrate that minimal modifications trigger catastrophic prediction failures, increasing relative $L_2$ error from $\sim$1.5% (validated accuracy) to 37-63% while remaining completely undetectable by standard validation metrics. Notably, 100% of successful single-point attacks pass z-score anomaly detection. We introduce the effective perturbation dimension $d_{\text{eff}}$, a Jacobian-based diagnostic that, together with sensitivity magnitude, yields a two-factor vulnerability model explaining why architectures with extreme sensitivity concentration (POD-DeepONet, $d_{\text{eff}} \approx 1$) are not necessarily the most exploitable, since low-rank output projections cap maximum error, while moderate concentration with sufficient amplification (S-DeepONet, $d_{\text{eff}} \approx 4$) produces the highest attack success. Gradient-free search outperforms gradient-based alternatives (PGD) on architectures with gradient pathologies, while random perturbations of equal magnitude achieve near-zero success rates, confirming that the discovered vulnerabilities are structural. Our findings expose a previously overlooked attack surface in operator learning models and establish that these models require robustness guarantees beyond standard validation before deployment.

This paper investigates how software professionals perceive the economic implications of diversity in software engineering teams. Motivated by a gap in software engineering research, which has largely emphasized socio-technical and process-related outcomes, we adopted a qualitative interview approach to capture practitioners' reasoning about diversity in relation to economic and market-oriented considerations. Based on interviews with ten software professionals, our analysis indicates that diversity is perceived as economically relevant through its associations with cost reduction and containment, revenue generation, time to market, process efficiency, innovation, and market alignment. Participants typically grounded these perceptions in concrete project experiences rather than abstract economic reasoning, framing diversity as a practical resource that supports project delivery, competitiveness, and organizational viability. Our findings provide preliminary empirical insights into how economic aspects of diversity are understood in software engineering practice.

Textual Large Language Models (LLMs) provide a simple and familiar interface: a string of text is used for both input and output. However, the information conveyed to an LLM often has a richer structure and semantics, which is not conveyed in a string. For example, most prompts contain both instructions ("Summarize this paper into a paragraph") and data (the paper to summarize), but these are usually not distinguished when passed to the model. This can lead to model confusion and security risks, such as prompt injection attacks. This work addresses this shortcoming by introducing an LLM-native mark-up language, LLMON (LLM Object Notation, pronounced "Lemon"), that enables the structure and semantic metadata of the text to be communicated in a natural way to an LLM. This information can then be used during model training, model prompting, and inference implementation, leading to improvements in model accuracy, safety, and security. This is analogous to how programming language types can be used for many purposes, such as static checking, code generation, dynamic checking, and IDE highlighting. We discuss the general design requirements of an LLM-native markup language, introduce the LLMON markup language and show how it meets these design requirements, describe how the information contained in a LLMON artifact can benefit model training and inference implementation, and provide some preliminary empirical evidence of its value for both of these use cases. We also discuss broader issues and research opportunities that are enabled with an LLM-native approach.

Validation of flood models, used to support risk mitigation strategies, remains challenging due to limited observations during extreme events. High-frequency, high-resolution optical imagery (~3 m), such as PlanetScope, offers new opportunities for flood mapping, although applications remain limited by cloud cover and the lack of labeled training data during disasters. To address this, we develop a flood mapping framework that integrates PlanetScope optical imagery with topographic features using machine learning (ML) and deep learning (DL) algorithms. A Random Forest model was applied to expert-annotated flood masks to generate training labels for DL models, U-Net. Two U-Net models with ResNet18 backbone were trained using optical imagery only (4 bands) and optical imagery combined with Height Above Nearest Drainage (HAND) and topographic slope (6 bands). Hurricane Ida (September 2021), which caused catastrophic flooding across the eastern United States, including the New York City metropolitan area, was used as an example to evaluate the framework. Results demonstrate that the U-Net model with topographic features achieved very close performance to the optical-only configuration (F1=0.92 and IoU=0.85 by both modeling scenarios), indicating that HAND and slope provide only marginal value to inundation extent detection. The proposed framework offers a scalable and label-efficient approach for mapping inundation extent that enables modeling under data-scarce flood scenarios.

Large-scale distributed learning aims at minimizing a loss function $L$ that depends on a training dataset with respect to a $d$-length parameter vector. The distributed cluster typically consists of a parameter server (PS) and multiple workers. Gradient coding is a technique that makes the learning process resilient to straggling workers. It introduces redundancy within the assignment of data points to the workers and uses coding theoretic ideas so that the PS can recover $\nabla L$ exactly or approximately, even in the presence of stragglers. Communication-efficient gradient coding allows the workers to communicate vectors of length smaller than $d$ to the PS, thus reducing the communication time. While there have been schemes that address the exact recovery of $\nabla L$ within communication-efficient gradient coding, to the best of our knowledge the approximate variant has not been considered in a systematic manner. In this work we present constructions of communication-efficient approximate gradient coding schemes. Our schemes use structured matrices that arise from bipartite graphs, combinatorial designs and strongly regular graphs, along with randomization and algebraic constraints. We derive analytical upper bounds on the approximation error of our schemes that are tight in certain cases. Moreover, we derive a corresponding worst-case lower bound on the approximation error of any scheme. For a large class of our methods, under reasonable probabilistic worker failure models, we show that the expected value of the computed gradient equals the true gradient. This in turn allows us to prove that the learning algorithm converges to a stationary point over the iterations. Numerical experiments corroborate our theoretical findings.

Public procurement refers to the process by which public sector institutions, such as governments, municipalities, and publicly funded bodies, acquire goods and services. Swiss law requires the integration of ecological, social, and economic sustainability requirements into tender evaluations in the format of criteria that have to be fulfilled by a bidder. However, translating high-level sustainability regulations into concrete, verifiable, and sector-specific procurement criteria (such as selection criteria, award criteria, and technical specifications) remains a labor-intensive and error-prone manual task, requiring substantial domain expertise in several groups of goods and services and considerable manual effort. This paper presents a configurable, LLM-assisted pipeline that is presented as a software supporting the systematic generation and evaluation of sustainability-oriented procurement criteria catalogs for Switzerland. The system integrates in-context prompting, interchangeable LLM backends, and automated output validation to enable auditable criteria generation across different procurement sectors. As a proof of concept, we instantiate the pipeline using official sustainability guidelines published by the Swiss government and the European Commission, which are ingested as structured reference documents. We evaluate the system through a combination of automated quality checks, including an LLM-based evaluation component, and expert comparison against a manually curated gold standard. Our results demonstrate that the proposed pipeline can substantially reduce manual drafting effort while producing criteria catalogs that are consistent with official guidelines. We further discuss system limitations, failure modes, and design trade-offs observed during deployment, highlighting key considerations for integrating generative AI into public sector software workflows.

Biological neural networks continuously adapt and modify themselves in response to experiences throughout their lifetime - a capability largely absent in artificial neural networks. Hebbian plasticity offers a promising path toward rapid adaptation in changing environments. Here, we introduce Hebbian Attractor Networks (HAN), a class of plastic neural networks in which local weight update normalization induces emergent attractor dynamics. Unlike prior approaches, HANs employ dual-timescale plasticity and temporal averaging of pre- and postsynaptic activations to induce either co-dynamic limit cycles or fixed-point weight attractors. Using simulated locomotion benchmarks, we gain insight into how Hebbian update frequency and activation averaging influence weight dynamics and control performance. Our results show that slower updates, combined with averaged pre- and postsynaptic activations, promote convergence to stable weight configurations, while faster updates yield oscillatory co-dynamic systems. We further demonstrate that these findings generalize to high-dimensional quadrupedal locomotion with a simulated Unitree Go1 robot. These results highlight how the timing of plasticity shapes neural dynamics in embodied systems, providing a principled characterization of the attractor regimes that emerge in self-modifying networks.

Large language models such as ChatGPT have increased scholarly output, but whether this productivity boost produces genuine intellectual advancement remains untested. I address this gap by measuring the semantic novelty of 13,847 articles published between 2020 and 2025 in 44 Information Systems journals. Using SPECTER2 embeddings, I operationalize novelty as the cosine distance between each paper and its nearest prior neighbors. A difference-in-differences design with the November 2022 release of ChatGPT as the treatment break reveals a heterogeneous pattern: authors affiliated with institutions in non-English-dominant countries show a 0.18 standard deviation decline in relative novelty compared to authors in English-dominant countries (beta = -0.176, p < 0.001), equivalent to a 7-percentile-point drop in the novelty distribution. This finding is robust across alternative novelty specifications, treatment break dates, and sub-samples, and survives a placebo test at a pre-treatment break. I interpret these results through the lens of construal level theory, proposing that LLMs function as proximity tools that shift researchers from abstract, exploratory thinking toward concrete, convention-following execution. The paper contributes to the growing debate on whether LLM-driven productivity gains come at the cost of intellectual diversity.

We present an energy-aware collaborative exploration framework for a UAV-UGV team operating in unknown environments, where the UAV's energy constraint is modeled as a maximum flight-time limit. The UAV executes a sequence of energy-bounded exploration tours, while the UGV simultaneously explores on the ground and serves as a mobile charging station. Rendezvous is enforced under a shared time budget so that the vehicles meet at the end of each tour before the UAV reaches its flight-time limit. We construct a sparsely coupled air-ground roadmap using a density-aware layered probabilistic roadmap (PRM) and formulate tour selection over the roadmap as coupled orienteering problems (OPs) to maximize information gain subject to the rendezvous constraint. The resulting tours are constructed over collision-validated roadmap edges. We validate our method through simulation studies, benchmark comparisons, and real-world experiments.

Synthetic insider threat benchmarks face a consistency problem: corpora generated without an external factual constraint cannot rule out cross-artifact contradictions. The CERT dataset -- the field's canonical benchmark -- is also static, lacks cross-surface correlation scenarios, and predates the LLM era. We present OrgForge-IT, a verifiable synthetic benchmark in which a deterministic simulation engine maintains ground truth and language models generate only surface prose, making cross-artifact consistency an architectural guarantee. The corpus spans 51 simulated days, 2,904 telemetry records at a 96.4% noise rate, and four detection scenarios designed to defeat single-surface and single-day triage strategies across three threat classes and eight injectable behaviors. A ten-model leaderboard reveals several findings: (1) triage and verdict accuracy dissociate - eight models achieve identical triage F1=0.80 yet split between verdict F1=1.0 and 0.80; (2) baseline false-positive rate is a necessary companion to verdict F1, with models at identical verdict accuracy differing by two orders of magnitude on triage noise; (3) victim attribution in the vishing scenario separates tiers - Tier A models exonerate the compromised account holder while Tier B models detect the attack but misclassify the victim; (4) rigid multi-signal thresholds structurally exclude single-surface negligent insiders, demonstrating the necessity of parallel, threat-class-specific triage pipelines; and (5) agentic software-engineering training acts as a force multiplier for multi-day temporal correlation, but only when paired with frontier-level parameter scale. Finally, prompt sensitivity analysis reveals that unstructured prompts induce vocabulary hallucination, motivating a two-track scoring framework separating prompt adherence from reasoning capability. OrgForge-IT is open source under the MIT license.

Where there is growing interest in in-context language learning (ICLL) for unseen languages with large language models, such languages usually suffer from the lack of NLP tools, data resources, and researcher expertise. This means that progress is difficult to assess, the field does not allow for cheap large-scale experimentation, and findings on ICLL are often limited to very few languages and tasks. In light of such limitations, we introduce a framework (Rashid), for studying ICLL wherein we reversibly cipher high-resource languages (HRLs) to construct truly unseen languages with access to a wide range of resources available for HRLs, unlocking previously impossible exploration of ICLL phenomena. We use our framework to assess current methods in the field with SOTA evaluation tools and manual analysis, explore the utility of potentially expensive resources in improving ICLL, and test ICLL strategies on rich downstream tasks beyond machine translation. These lines of exploration showcase the possibilities enabled by our framework, as well as providing actionable insights regarding current performance and future directions in ICLL.

In the 1980s, high-performance computing (HPC) became another tool for research in the open (non-defense) science and engineering research communities. However, HPC came with a high price tag; the first Cray-2 machines, released in 1985, cost between \$12 million and \$17 million, according to the Computer History Museum, and were largely available only at government research labs or through national supercomputing centers. In the 1990s, with demand for HPC increasing due to vast datasets, more complex modeling, and the growing computational needs of scientific applications, researchers began experimenting with building HPC machines from clusters of servers running the Linux operating system. By the late 1990s, two approaches to Linux-based parallel computing had emerged: the personal computer cluster methodology that became known as Beowulf and the Roadrunner architecture aimed at a more cost-effective supercomputer. While Beowulf attracted attention because of its low cost and thereby greater accessibility, Roadrunner took a different approach. While still affordable compared to vector processors and other commercially available supercomputers, Roadrunner integrated its commodity components with specialized networking technology. Furthermore, these systems initially served different purposes. While Beowulf focused on providing affordable parallel workstations for individual researchers at NASA, Roadrunner set out to provide a multi-user system that could compete with the commercial supercomputers that dominated the market at the time. This paper analyzes the technical decisions, performance implications, and long-term influence of both approaches. Through this analysis, we can start to judge the impact of both Roadrunner and Beowulf on the development of Linux-based supercomputers.

Inference-time scaling has emerged as an effective way to improve generative models at test time by using a verifier to score and select candidate outputs. A common choice is to employ Multimodal Large Language Models (MLLMs) as verifiers, which can improve performance but introduce substantial inference-time cost. Indeed, diffusion pipelines operate in an autoencoder latent space to reduce computation, yet MLLM verifiers still require decoding candidates to pixel space and re-encoding them into the visual embedding space, leading to redundant and costly operations. In this work, we propose Verifier on Hidden States (VHS), a verifier that operates directly on intermediate hidden representations of Diffusion Transformer (DiT) single-step generators. VHS analyzes generator features without decoding to pixel space, thereby reducing the per-candidate verification cost while improving or matching the performance of MLLM-based competitors. We show that, under tiny inference budgets with only a small number of candidates per prompt, VHS enables more efficient inference-time scaling reducing joint generation-and-verification time by 63.3%, compute FLOPs by 51% and VRAM usage by 14.5% with respect to a standard MLLM verifier, achieving a +2.7% improvement on GenEval at the same inference-time budget.

The Model Context Protocol (MCP) has rapidly emerged as a universal standard for connecting AI assistants to external tools and data sources. While MCP simplifies integration between AI applications and various services, it introduces significant security vulnerabilities, particularly on the client side. In this work we conduct threat modelings of MCP implementations using STRIDE (Spoofing, Tampering, Repudiation, Information Disclosure, Denial of Service, Elevation of Privilege) and DREAD (Damage, Reproducibility, Exploitability, Affected Users, Discoverability) frameworks across five key components: (1) MCP Host and Client, (2) LLM, (3) MCP Server, (4) External Data Stores, and (5) Authorization Server. This comprehensive analysis reveals tool poisoning-where malicious instructions are embedded in tool metadata-as the most prevalent and impactful client-side vulnerability. We therefore focus our empirical evaluation on this critical attack vector, providing a systematic comparison of how seven major MCP clients validate and defend against tool poisoning attacks. Our analysis reveals significant security issues with most tested clients due to insufficient static validation and parameter visibility. We propose a multi-layered defense strategy encompassing static metadata analysis, model decision path tracking, behavioral anomaly detection, and user transparency mechanisms. This research addresses a critical gap in MCP security, which has primarily focused on server-side vulnerabilities, and provides actionable recommendations and mitigation strategies for securing AI agent ecosystems.

Current architecture proposals within standards development organizations such as ETSI and 3GPP enable sensing capabilities in mobile networks; however, they do not include a repository for storing sensing data. Such a repository can be used for AI model training and to complement ongoing sensing service provisioning by improving efficiency and accuracy. One way of realizing this is through the fusion of historical sensing data with live sensing data. In this paper, we study historical and live sensing data fusion for Integrated Sensing and Communication in future 6G systems and introduce a Sensing Data Storage Function to store historical sensing data and sensing results. We show how the Sensing Data Storage Function can be used with other network functions in a 6G architecture proposition for Integrated Sensing and Communication. We validate our proposal with a measurement model and show performance improvements in terms of detection probability and false-alarm rate. The network functionality to fuse and process sensing data combines live sensing measurements with previously sensed historical sensing data using a map-aware hard filter that rejects detections consistent with known static structures. Our simulation illustrates that, for a traffic junction scenario, map-aware hard filtering substantially reduces false alarms without degrading detection probability.

Serverless computing is a cloud execution model where developers run code, and the server management is handled by the cloud provider. Serverless computing is increasingly gaining popularity as more systems adopt it to enhance scalability and reduce operational costs. While it has numerous benefits, it also embodies unique challenges inherent to serverless computing. One such challenge is Technical Debt (TD), which is exacerbated by the complexities of the serverless paradigm. While prior work has investigated the activities and bad practices that lead to TD in serverless computing, there remains a gap in understanding how TD manifests, the challenges it poses, and the solutions proposed to address TD issues in serverless systems. This study aims to investigate TD in the serverless context using Stack Overflow (SO) as a knowledge base. We collected 78,867 serverless questions on SO and labeled them as TD or non-TD using deep learning. Moreover, we conducted an in-depth analysis to identify types of TD in serverless settings, associated issues, and proposed solutions. We found that 37% of the serverless questions on SO are TD-related. We also identified six serverless-specific issues. Our research highlights the need for tools that can effectively detect TD in serverless applications.

Defining a constructive process to build general capabilities for language models in an automatic manner is considered an open problem in artificial intelligence. Towards this, we consider the problem of building a curriculum of tasks that grows a model via relevant skill discovery. We provide a concrete framework for this task, using a family of tasks called cross-entropy games, which we postulate is universal in a suitable sense. We show that if it is possible to grow the curriculum for relevant skill discovery by iterating a greedy optimization algorithm, then, under natural assumptions, there is essentially only one meta-objective possible (up to a few hyperparameters). We call the resulting process cognitive training. We postulate that, given sufficiently capable language models as players and meta-samplers and sufficient training time, cognitive training provides a principled way to relevant skill discovery; and hence to the extent general capabilities are achievable via greedy curriculum learning, cognitive training would be a solution.

Software analytics often builds from labeled data. Labeling can be slow, error prone, and expensive. When human expertise is scarce, SE researchers sometimes ask large language models (LLMs) for the missing labels. While this has been successful in some domains, recent results show that LLM-based labeling has blind spots. Specifically, their labeling is not effective for higher dimensional multi-objective problems. To address this task, we propose a novel LLM prompting strategy called SynthCore. When one opinion fails, SynthCore's combines multiple separated opinions generated by LLMs (with no knowledge of each others' answers) into an ensemble of few-shot learners. Simpler than other strategies (e.g. chain-of-thought, multi-agent-debate, etc) SynthCore aggregates results from multiple single prompt sessions (with no crossover between them). SynthCore has been tested on 49 SE multi-objective optimization tasks, handling tasks as diverse as software project management, Makefile configuration, and hyperparameter optimization. SynthCore's ensemble found optimizations that are better than state-of-the-art alternative approaches (Gaussian Process Models, Tree of Parzen Estimators, active learners in both exploration and exploitation mode). Importantly, these optimizations were made using data labeled by LLMs, without any human opinions. From these experiments, we conclude that ensembles of few shot learners can successfully annotate high dimensional multi-objective tasks. Further, we speculate that other successful few-shot prompting results could be quickly and easily enhanced using SynthCore's ensemble approach. To support open science, all our data and scripts are available at https://github.com/lohithsowmiyan/lazy-llm/tree/clusters.

Hybrid language models combining attention with state space models (SSMs) or linear attention offer improved efficiency, but whether both components are genuinely utilized remains unclear. We present a functional component ablation framework applied to two sub-1B hybrid models -- Qwen3.5-0.8B (sequential: Gated DeltaNet + softmax attention) and Falcon-H1-0.5B (parallel: Mamba-2 + attention) -- with a pure Transformer control (Qwen2.5-0.5B). Through group ablations, layer-wise sweeps, positional ablations, matched random controls, and perplexity analysis across five benchmarks, we establish four findings: (1) both component types are essential and neither is bypassed; (2) the alternative component (linear attention or SSM) is the primary language modeling backbone, causing >35,000x perplexity degradation when removed versus ~82x for attention; (3) component importance follows a positional gradient, with early layers being disproportionately critical; and (4) hybrid architectures exhibit 20-119x greater resilience to random layer removal than pure Transformers, revealing built-in functional redundancy between component types. These results provide actionable guidance for hybrid model compression, architecture design, and fault-tolerant deployment.

Autonomous aerial and aquatic robots that attain mobility by perturbing their medium, such as multicopters and torpedoes, produce wake effects that act as disturbances for adjacent robots. Wake effects are hard to model and predict due to the chaotic spatio-temporal dynamics of the fluid, entangled with the physical geometry of the robots and their complex motion patterns. Data-driven approaches using neural networks typically learn a memory-less function that maps the current states of the two robots to a force observed by the "sufferer" robot. Such models often perform poorly in agile scenarios: since the wake effect has a finite propagation time, the disturbance observed by a sufferer robot is some function of relative states in the past. In this work, we present an empirical study of the properties a wake-effect predictor must satisfy to accurately model the interactions between two robots mediated by a fluid. We explore seven data-driven models designed to capture the spatio-temporal evolution of fluid wake effects in four different media. This allows us to introspect the models and analyze the reasons why certain features enable improved accuracy in prediction across predictors and fluids. As experimental validation, we develop a planar rectilinear gantry for two spinning monocopters to test in real-world data with feedback control. The conclusion is that support of history of previous states as input and transport delay prediction substantially helps to learn an accurate wake-effect predictor.

The growing complexity of modern control tasks calls for controllers that can react online as objectives and disturbances change, while preserving closed-loop stability. Recent approaches for improving the performance of nonlinear systems while preserving closed-loop stability rely on time-invariant recurrent neural-network controllers, but offer no principled way to update the controller during operation. Most importantly, switching from one stabilizing policy to another can itself destabilize the closed-loop. We address this problem by introducing a stability-preserving update mechanism for nonlinear, neural-network-based controllers. Each controller is modeled as a causal operator with bounded $\ell_p$-gain, and we derive gain-based conditions under which the controller may be updated online. These conditions yield two practical update schemes, time-scheduled and state-triggered, that guarantee the closed-loop remains $\ell_p$-stable after any number of updates. Our analysis further shows that stability is decoupled from controller optimality, allowing approximate or early-stopped controller synthesis. We demonstrate the approach on nonlinear systems with time-varying objectives and disturbances, and show consistent performance improvements over static and naive online baselines while guaranteeing stability.

We derive posterior contraction rates (PCRs) and finite-sample Bernstein von Mises (BvM) results for non-parametric Bayesian models by extending the diffusion-based framework of Mou et al. (2024) to the infinite-dimensional setting. The posterior is represented as the invariant measure of a Langevin stochastic partial differential equation (SPDE) on a separable Hilbert space, which allows us to control posterior moments and obtain non-asymptotic concentration rates in Hilbert norms under various likelihood curvature and regularity conditions. We also establish a quantitative Laplace approximation for the posterior. The theory is illustrated in a nonparametric linear Gaussian inverse problem.

Always-on sensing is essential for next-generation edge/wearable AI systems, yet continuous high-fidelity RGB video capture remains prohibitively expensive for resource-constrained mobile and edge platforms. We present a new paradigm for efficient streaming video understanding: grayscale-always, color-on-demand. Through preliminary studies, we discover that color is not always necessary. Sparse RGB frames suffice for comparable performance when temporal structure is preserved via continuous grayscale streams. Building on this insight, we propose ColorTrigger, an online training-free trigger that selectively activates color capture based on windowed grayscale affinity analysis. Designed for real-time edge deployment, ColorTrigger uses lightweight quadratic programming to detect chromatic redundancy causally, coupled with credit-budgeted control and dynamic token routing to jointly reduce sensing and inference costs. On streaming video understanding benchmarks, ColorTrigger achieves 91.6% of full-color baseline performance while using only 8.1% RGB frames, demonstrating substantial color redundancy in natural videos and enabling practical always-on video sensing on resource-constrained devices.

Federated Learning (FL) is constrained by the communication and energy limitations of decentralized edge devices. While gradient sparsification via Top-K magnitude pruning effectively reduces the communication payload, it remains inherently energy-agnostic. It assumes all parameter updates incur identical downstream transmission and memory-update costs, ignoring hardware realities. We formalize the pruning process as an energy-constrained projection problem that accounts for the hardware-level disparities between memory-intensive and compute-efficient operations during the post-backpropagation phase. We propose Cost-Weighted Magnitude Pruning (CWMP), a selection rule that prioritizes parameter updates based on their magnitude relative to their physical cost. We demonstrate that CWMP is the optimal greedy solution to this constrained projection and provide a probabilistic analysis of its global energy efficiency. Numerical results on a non-IID CIFAR-10 benchmark show that CWMP consistently establishes a superior performance-energy Pareto frontier compared to the Top-K baseline.

Enhancing reader engagement while preserving informational fidelity is a central challenge in controllable text generation for news media. Optimizing news headlines for reader engagement is often conflated with clickbait, resulting in exaggerated or misleading phrasing that undermines editorial trust. We frame clickbait not as a separate stylistic category, but as an extreme outcome of disproportionate amplification of otherwise legitimate engagement cues. Based on this view, we formulate headline rewriting as a controllable generation problem, where specific engagement-oriented linguistic attributes are selectively strengthened under explicit constraints on semantic faithfulness and proportional emphasis. We present a guided headline rewriting framework built on a large language model (LLM) that uses the Future Discriminators for Generation (FUDGE) paradigm for inference-time control. The LLM is steered by two auxiliary guide models: (1) a clickbait scoring model that provides negative guidance to suppress excessive stylistic amplification, and (2) an engagement-attribute model that provides positive guidance aligned with target clickability objectives. Both guides are trained on neutral headlines drawn from a curated real-world news corpus. At the same time, clickbait variants are generated synthetically by rewriting these original headlines using an LLM under controlled activation of predefined engagement tactics. By adjusting guidance weights at inference time, the system generates headlines along a continuum from neutral paraphrases to more engaging yet editorially acceptable formulations. The proposed framework provides a principled approach for studying the trade-off between attractiveness, semantic preservation, and clickbait avoidance, and supports responsible LLM-based headline optimization in journalistic settings.

As LLM agent ecosystems grow, the number of available skills (tools, plugins) has reached tens of thousands, making it infeasible to inject all skills into an agent's context. This creates a need for skill routing -- retrieving the most relevant skills from a large pool given a user task. The problem is compounded by pervasive functional overlap in community skill repositories, where many skills share similar names and purposes yet differ in implementation details. Despite its practical importance, skill routing remains under-explored. Current agent architectures adopt a progressive disclosure design -- exposing only skill names and descriptions to the agent while keeping the full implementation body hidden -- implicitly treating metadata as sufficient for selection. We challenge this assumption through a systematic empirical study on a benchmark of ~$80K skills and 75 expert-verified queries. Our key finding is that the skill body (full implementation text) is the decisive signal: removing it causes 29--44 percentage point degradation across all retrieval methods, and cross-encoder attention analysis reveals 91.7% of attention concentrating on the body field. Motivated by this finding, we propose SkillRouter, a two-stage retrieve-and-rerank pipeline totaling only 1.2B parameters (0.6B encoder + 0.6B reranker). SkillRouter achieves 74.0% top-1 routing accuracy and delivers the strongest average result among the compact and zero-shot baselines we evaluate, while remaining deployable on consumer hardware.

Community Notes have emerged as an effective crowd-sourced mechanism for combating online deception on social media platforms. However, its reliance on human contributors limits both the timeliness and scalability. In this work, we study the automated Community Notes generation method for image-based contextual deception, where an authentic image is paired with misleading context (e.g., time, entity, and event). Unlike prior work that primarily focuses on deception detection (i.e., judging whether a post is true or false in a binary manner), Community Notes-style systems need to generate concise and grounded notes that help users recover the missing or corrected context. This problem remains underexplored due to three reasons: (i) datasets that support the research are scarce; (ii) methods must handle the dynamic nature of contextual deception; (iii) evaluation is difficult because standard metrics do not capture whether notes actually improve user understanding. To address these gaps, we curate a real-world dataset, XCheck, comprising X posts with associated Community Notes and external contexts. We further propose the Automated Context-Corrective Note generation method, named ACCNote, which is a retrieval-augmented, multi-agent collaboration framework built on large vision-language models. Finally, we introduce a new evaluation metric, Context Helpfulness Score (CHS), that aligns with user study outcomes rather than relying on lexical overlap. Experiments on our XCheck dataset show that the proposed ACCNote improves both deception detection and note generation performance over baselines, and exceeds a commercial tool GPT5-mini. Together, our dataset, method, and metric advance practical automated generation of context-corrective notes toward more responsible online social networks.

Agent skills are modular instruction packages that combine YAML metadata, natural language instructions, and embedded code, and they have reached 196K publicly available instances, yet no mechanism exists to detect clone relationships among them. This gap creates systemic risks: a vulnerability in a widely copied skill silently persists across derivatives with no alert to maintainers. Existing clone detectors, designed for single-modality source code, cannot handle the multi-modal structure of skills, where clone evidence is distributed across three interleaved content channels. We present SkillClone, the first multi-modal clone detection approach for agent skills. SkillClone fuses flat TF-IDF similarity with per-channel decomposition (YAML, NL, code) through logistic regression, combining strong detection with interpretable type classification. We construct SkillClone-Bench, a balanced benchmark of 300 ground-truth pairs with stratified difficulty. On SkillClone-Bench, SkillClone achieves F1 of 0.939 with precision 0.952, outperforming flat TF-IDF (F1 = 0.881) and achieving 4.2x higher Type-4 (semantic) recall than MinHash. Applying SkillClone to 20K skills reveals 258K clone pairs involving 75% of all skills, with 40% crossing author boundaries. A deduplication analysis shows the ecosystem is inflated 3.5x: only 5,642 unique skill concepts underlie the 20K listed skills, and 41% of skills in clone families are superseded by a strictly better variant.

Reinforcement learning with verifiable rewards (RLVR) has significantly improved reasoning in large language models (LLMs), yet the token-level mechanisms underlying these improvements remain unclear. We present a systematic empirical study of RLVR's distributional effects organized around three main analyses: (1) token-level characterization of distributional shifts between base and RL models, (2) the impact of token-level distributional shifts on sequence-level reasoning performance through cross-sampling interventions, and (3) fine-grained mechanics of these shifts at the token level. We find that RL fine-tuning induces highly sparse and targeted changes, with only a small fraction of token distributions exhibiting meaningful divergence between the base and RL policies. We further characterize the structure and evolution of these shifts through analyses of token entropy, positional concentration, and reallocation of probability mass. To assess the functional importance of these sparse changes, we conduct cross-sampling experiments that selectively swap token choices between the base and RL models with varying intervention budgets. We show that inserting only a small fraction of RL-sampled tokens into base generations progressively recovers RL performance gains, while injecting a similarly small number of base token choices into otherwise RL-generated sequences collapses performance to base levels, isolating a small set of token-level decisions directly responsible for RLVR's performance gains. Finally, we explore divergence-weighted variants of the advantage signal as a diagnostic intervention, finding that they can yield improvements over baselines. Together, our results shed light on the distributional changes induced by RLVR and provide a fine-grained, token-level lens for understanding RLVR fine-tuning as a targeted refinement process.

Cryptographic migration driven by algorithm deprecation, regulatory change, and post-quantum readiness requires more than an inventory of cryptographic assets. Existing Cryptographic Bills of Materials (CBOMs) are typically tool- or inventory-derived. They lack architectural intent, rationale, and security context, limiting their usefulness for migration planning. This paper introduces Security-Aware Architecture Tradeoff Analysis Method (SATAM), a security-aware adaptation of scenario-based architecture evaluation that derives an architecture-grounded, context-sensitive CBOM. SATAM integrates established approaches: ATAM, arc42, STRIDE, ADR, and CARAF. These are included to identify and analyze security-relevant cryptographic decision points and document them as explicit architectural decisions. These artifacts are used to annotate CBOM entries with architectural context, security intent, and migration-critical metadata using CycloneDX-compatible extensions. Following a Design Science Research approach, the paper presents the method design, a conceptual traceability model, and an illustrative application. The results demonstrate that architecture-derived CBOMs capture migration-relevant context that is typically absent from inventory-based approaches. Thereby, SATAM improves availability of information required for informed cryptographic migration planning and long-term cryptographic agility.

We present mmFHE, the first system that enables fully homomorphic encryption (FHE) for end-to-end mmWave radar sensing. mmFHE encrypts raw range profiles on a lightweight edge device and executes the entire mmWave signal-processing and ML inference pipeline homomorphically on an untrusted cloud that operates exclusively on ciphertexts. At the core of mmFHE is a library of seven composable, data-oblivious FHE kernels that replace standard DSP routines with fixed arithmetic circuits. These kernels can be flexibly composed into different application-specific pipelines. We demonstrate this approach on two representative tasks: vital-sign monitoring and gesture recognition. We formally prove two cryptographic guarantees for any pipeline assembled from this library: input privacy, the cloud learns nothing about the sensor data; and data obliviousness, the execution trace is identical on the cloud regardless of the data being processed. These guarantees effectively neutralize various supervised and unsupervised privacy attacks on raw data, including re-identification and data-dependent privacy leakage. Evaluation on three public radar datasets (270 vital-sign recordings, 600 gesture trials) shows that encryption introduces negligible error: HR/RR MAE <10^-3 bpm versus plaintext, and 84.5% gesture accuracy (vs. 84.7% plaintext) with end-to-end cloud GPU latency of 103s for a 10s vital-sign window and 37s for a 3s gesture window. These results show that privacy-preserving end-to-end mmWave sensing is feasible on commodity hardware today.

"Code-as-Policy" considers how executable code can complement data-intensive Vision-Language-Action (VLA) methods, yet their effectiveness as autonomous controllers for embodied manipulation remains underexplored. We present CaP-X, an open-access framework for systematically studying Code-as-Policy agents in robot manipulation. At its core is CaP-Gym, an interactive environment in which agents control robots by synthesizing and executing programs that compose perception and control primitives. Building on this foundation, CaP-Bench evaluates frontier language and vision-language models across varying levels of abstraction, interaction, and perceptual grounding. Across 12 models, CaP-Bench reveals a consistent trend: performance improves with human-crafted abstractions but degrades as these priors are removed, exposing a dependence on designer scaffolding. At the same time, we observe that this gap can be mitigated through scaling agentic test-time computation--through multi-turn interaction, structured execution feedback, visual differencing, automatic skill synthesis, and ensembled reasoning--substantially improves robustness even when agents operate over low-level primitives. These findings allow us to derive CaP-Agent0, a training-free framework that recovers human-level reliability on several manipulation tasks in simulation and on real embodiments. We further introduce CaP-RL, showing reinforcement learning with verifiable rewards improves success rates and transfers from sim2real with minimal gap. Together, CaP-X provides a principled, open-access platform for advancing embodied coding agents.

Offline Reinforcement Learning (RL) aims to learn optimal policies from fixed offline datasets, without further interactions with the environment. Such methods train an offline policy (or value function), and apply it at inference time without further refinement. We introduce an inference time adaptation framework inspired by model predictive control (MPC) that utilizes a pretrained policy along with a learned world model of state transitions and rewards. While existing world model and diffusion-planning methods use learned dynamics to generate imagined trajectories during training, or to sample candidate plans at inference time, they do not use inference-time information to optimize the policy parameters on the fly. In contrast, our design is a Differentiable World Model (DWM) pipeline that enables endto-end gradient computation through imagined rollouts for policy optimization at inference time based on MPC. We evaluate our algorithm on D4RL continuous-control benchmarks (MuJoCo locomotion tasks and AntMaze), and show that exploiting inference-time information to optimize the policy parameters yields consistent gains over strong offline RL baselines.

Symbolic regression aims to discover human-interpretable equations that explain observational data. However, existing approaches rely heavily on discrete structure search (e.g., genetic programming), which often leads to high computational cost, unstable performance, and limited scalability to large equation spaces. To address these challenges, we propose SRCO, a unified embedding-driven framework for symbolic regression that transforms symbolic structures into a continuous, optimizable representation space. The framework consists of three key components: (1) structure embedding: we first generate a large pool of exploratory equations using traditional symbolic regression algorithms and train a Transformer model to compress symbolic structures into a continuous embedding space; (2) continuous structure search: the embedding space enables efficient exploration using gradient-based or sampling-based optimization, significantly reducing the cost of navigating the combinatorial structure space; and (3) coefficient optimization: for each discovered structure, we treat symbolic coefficients as learnable parameters and apply gradient optimization to obtain accurate numerical values. Experiments on synthetic and real-world datasets show that our approach consistently outperforms state-of-the-art methods in equation accuracy, robustness, and search efficiency. This work introduces a new paradigm for symbolic regression by bridging symbolic equation discovery with continuous embedding learning and optimization.

Consider a market of competing model providers selling query access to models with varying costs and capabilities. Customers submit problem instances and are willing to pay up to a budget for a verifiable solution. An arbitrageur efficiently allocates inference budget across providers to undercut the market, thus creating a competitive offering with no model-development risk. In this work, we initiate the study of arbitrage in AI model markets, empirically demonstrating the viability of arbitrage and illustrating its economic consequences. We conduct an in-depth case study of SWE-bench GitHub issue resolution using two representative models, GPT-5 mini and DeepSeek v3.2. In this verifiable domain, simple arbitrage strategies generate net profit margins of up to 40%. Robust arbitrage strategies that generalize across different domains remain profitable. Distillation further creates strong arbitrage opportunities, potentially at the expense of the teacher model's revenue. Multiple competing arbitrageurs drive down consumer prices, reducing the marginal revenue of model providers. At the same time, arbitrage reduces market segmentation and facilitates market entry for smaller model providers by enabling earlier revenue capture. Our results suggest that arbitrage can be a powerful force in AI model markets with implications for model development, distillation, and deployment.

Quantum computers provide a super-exponential speedup for performing a Fourier transform over the symmetric group, an ability for which practical use cases have remained elusive so far. In this work, we leverage this ability to unlock spectral methods for machine learning over permutation-structured data, which appear in applications such as multi-object tracking and recommendation systems. It has been shown previously that a powerful way of building probabilistic models over permutations is to use the framework of non-Abelian harmonic analysis, as the model's group Fourier spectrum captures the interaction complexity: "low frequencies" correspond to low order correlations, and "high frequencies" to more complex ones. This can be used to construct a Markov chain model driven by alternating steps of diffusion (a group-equivariant convolution) and conditioning (a Bayesian update). However, this approach is computationally challenging and hence limited to simple approximations. Here we construct a quantum algorithm that encodes the exact probabilistic model -- a classically intractable object -- into the amplitudes of a quantum state by making use of the Quantum Fourier Transform (QFT) over the symmetric group. We discuss the scaling, limitations, and practical use of such an approach, which we envision to be a first step towards useful applications of non-Abelian QFTs.

The development of new drugs is a tedious, time-consuming, and expensive process, for which the average costs are estimated to be up to around $2.5 billion. The first step in this long process is the design of the new drug, for which de novo drug design, assisted by artificial intelligence, has blossomed in recent years and revolutionized the field. In particular, generative artificial intelligence has delivered promising results in drug discovery and development, reducing costs and the time to solution. However, classical generative models, such as generative adversarial networks (GANs), are difficult to train due to barren plateaus and prone to mode collapse. Quantum computing may be an avenue to overcome these issues and provide models with fewer parameters, thereby enhancing the generalizability of GANs. We propose a new style-based quantum GAN (QGAN) architecture for drug design that implements noise encoding at every rotational gate of the circuit and a gradient penalty in the loss function to mitigate mode collapse. Our pipeline employs a variational autoencoder to represent the molecular structure in a latent space, which is then used as input to our QGAN. Our baseline model runs on up to 15 qubits to validate our architecture on quantum simulators, and a 156-qubit IBM Heron quantum computer in the five-qubit setup is used for inference to investigate the effects of using real quantum hardware on the analysis. We benchmark our results against classical models as provided by the MOSES benchmark suite.

Diffusion Transformers (DiTs) power high-fidelity video world models but remain computationally expensive due to sequential denoising and costly spatio-temporal attention. Training-free feature caching accelerates inference by reusing intermediate activations across denoising steps; however, existing methods largely rely on a Zero-Order Hold assumption i.e., reusing cached features as static snapshots when global drift is small. This often leads to ghosting artifacts, blur, and motion inconsistencies in dynamic scenes. We propose \textbf{WorldCache}, a Perception-Constrained Dynamical Caching framework that improves both when and how to reuse features. WorldCache introduces motion-adaptive thresholds, saliency-weighted drift estimation, optimal approximation via blending and warping, and phase-aware threshold scheduling across diffusion steps. Our cohesive approach enables adaptive, motion-consistent feature reuse without retraining. On Cosmos-Predict2.5-2B evaluated on PAI-Bench, WorldCache achieves \textbf{2.3$\times$} inference speedup while preserving \textbf{99.4\%} of baseline quality, substantially outperforming prior training-free caching approaches. Our code can be accessed on \href{https://umair1221.github.io/World-Cache/}{World-Cache}.

Latent diffusion models (LDMs) enable high-fidelity synthesis by operating in learned latent spaces. However, training state-of-the-art LDMs requires complex staging: a tokenizer must be trained first, before the diffusion model can be trained in the frozen latent space. We propose UNITE - an autoencoder architecture for unified tokenization and latent diffusion. UNITE consists of a Generative Encoder that serves as both image tokenizer and latent generator via weight sharing. Our key insight is that tokenization and generation can be viewed as the same latent inference problem under different conditioning regimes: tokenization infers latents from fully observed images, whereas generation infers them from noise together with text or class conditioning. Motivated by this, we introduce a single-stage training procedure that jointly optimizes both tasks via two forward passes through the same Generative Encoder. The shared parameters enable gradients to jointly shape the latent space, encouraging a "common latent language". Across image and molecule modalities, UNITE achieves near state of the art performance without adversarial losses or pretrained encoders (e.g., DINO), reaching FID 2.12 and 1.73 for Base and Large models on ImageNet 256 x 256. We further analyze the Generative Encoder through the lenses of representation alignment and compression. These results show that single stage joint training of tokenization & generation from scratch is feasible.

We present UniMotion, to our knowledge the first unified framework for simultaneous understanding and generation of human motion, natural language, and RGB images within a single architecture. Existing unified models handle only restricted modality subsets (e.g., Motion-Text or static Pose-Image) and predominantly rely on discrete tokenization, which introduces quantization errors and disrupts temporal continuity. UniMotion overcomes both limitations through a core principle: treating motion as a first-class continuous modality on equal footing with RGB. A novel Cross-Modal Aligned Motion VAE (CMA-VAE) and symmetric dual-path embedders construct parallel continuous pathways for Motion and RGB within a shared LLM backbone. To inject visual-semantic priors into motion representations without requiring images at inference, we propose Dual-Posterior KL Alignment (DPA), which distills a vision-fused encoder's richer posterior into the motion-only encoder. To address the cold-start problem -- where text supervision alone is too sparse to calibrate the newly introduced motion pathway -- we further propose Latent Reconstruction Alignment (LRA), a self-supervised pre-training strategy that uses dense motion latents as unambiguous conditions to co-calibrate the embedder, backbone, and flow head, establishing a stable motion-aware foundation for all downstream tasks. UniMotion achieves state-of-the-art performance across seven tasks spanning any-to-any understanding, generation, and editing among the three modalities, with especially strong advantages on cross-modal compositional tasks.

Recent progress in latent world models (e.g., V-JEPA2) has shown promising capability in forecasting future world states from video observations. Nevertheless, dense prediction from a short observation window limits temporal context and can bias predictors toward local, low-level extrapolation, making it difficult to capture long-horizon semantics and reducing downstream utility. Vision--language models (VLMs), in contrast, provide strong semantic grounding and general knowledge by reasoning over uniformly sampled frames, but they are not ideal as standalone dense predictors due to compute-driven sparse sampling, a language-output bottleneck that compresses fine-grained interaction states into text-oriented representations, and a data-regime mismatch when adapting to small action-conditioned datasets. We propose a VLM-guided JEPA-style latent world modeling framework that combines dense-frame dynamics modeling with long-horizon semantic guidance via a dual-temporal pathway: a dense JEPA branch for fine-grained motion and interaction cues, and a uniformly sampled VLM \emph{thinker} branch with a larger temporal stride for knowledge-rich guidance. To transfer the VLM's progressive reasoning signals effectively, we introduce a hierarchical pyramid representation extraction module that aggregates multi-layer VLM representations into guidance features compatible with latent prediction. Experiments on hand-manipulation trajectory prediction show that our method outperforms both a strong VLM-only baseline and a JEPA-predictor baseline, and yields more robust long-horizon rollout behavior.

Large Language Models (LLMs) and Vision Language Models (VLMs) have shown impressive reasoning abilities, yet they struggle with spatial understanding and layout consistency when performing fine-grained visual editing. We introduce a Structured Reasoning framework that performs text-conditioned spatial layout editing via scene-graph reasoning. Given an input scene graph and a natural-language instruction, the model reasons over the graph to generate an updated scene graph that satisfies the text condition while maintaining spatial coherence. By explicitly guiding the reasoning process through structured relational representations, our approach improves both interpretability and control over spatial relationships. We evaluate our method on a new text-guided layout editing benchmark encompassing sorting, spatial alignment, and room-editing tasks. Our training paradigm yields an average 15% improvement in IoU and 25% reduction in center-distance error compared to Chain of Thought Fine-tuning (CoT-SFT) and vanilla GRPO baselines. Compared to SOTA zero-shot LLMs, our best models achieve up to 20% higher mIoU, demonstrating markedly improved spatial precision.

Many multimodal tasks, such as image captioning and visual question answering, require vision-language models (VLMs) to associate objects with their properties and spatial relations. Yet it remains unclear where and how such associations are computed within VLMs. In this work, we show that VLMs rely on two concurrent mechanisms to represent such associations. In the language model backbone, intermediate layers represent content-independent spatial relations on top of visual tokens corresponding to objects. However, this mechanism plays only a secondary role in shaping model predictions. Instead, the dominant source of spatial information originates in the vision encoder, whose representations encode the layout of objects and are directly exploited by the language model backbone. Notably, this spatial signal is distributed globally across visual tokens, extending beyond object regions into surrounding background areas. We show that enhancing these vision-derived spatial representations globally across all image tokens improves spatial reasoning performance on naturalistic images. Together, our results clarify how spatial association is computed within VLMs and highlight the central role of vision encoders in enabling spatial reasoning.

Weight-Decomposed Low-Rank Adaptation (DoRA) extends LoRA by decoupling weight magnitude from direction, but its forward pass requires the row-wise norm of W + sBA, a computation that every major framework we surveyed implements by materializing the dense [d_out, d_in] product BA. At d_in = 8192 and rank r = 384, a single module's norm requires about 512 MB of transient working memory in bf16, making high-rank DoRA costly and often infeasible on common single-GPU setups once hundreds of adapted modules and checkpointing are involved. We present two systems contributions. A factored norm decomposes the squared norm into base, cross, and Gram terms computable through O(d_out r + r^2) intermediates, eliminating the dense product. Fused Triton kernels collapse the four-kernel DoRA composition into a single pass, reducing memory traffic by about 4x and using a numerically stable form that avoids catastrophic cancellation in the near-unity rescaling regime where magnitude scales concentrate in practice. Across six 8-32B vision-language models (VLMs) on three NVIDIA GPUs (RTX 6000 PRO, H200, B200) at r = 384 in bf16, the fused implementation is 1.5-2.0x faster than Hugging Face PEFT's DoRA implementation for inference and 1.5-1.9x faster for gradient computation (optimizer step excluded), with up to 7 GB lower peak VRAM. Microbenchmarks on six GPUs spanning four architecture generations (L40S, A100, RTX 6000 PRO, H200, B200, B300) confirm 1.5-2.7x compose-kernel speedup. Final-logit cosine similarity exceeds 0.9999 across all model/GPU pairs, and multi-seed training curves match within 7.1 x 10^-4 mean per-step loss delta over 2000 steps.

The process of discovery requires active exploration -- the act of collecting new and informative data. However, efficient autonomous exploration remains a major unsolved problem. The dominant paradigm addresses this challenge by using Reinforcement Learning (RL) to train agents with intrinsic motivation, maximizing a composite objective of extrinsic and intrinsic rewards. We suggest that this approach incurs unnecessary overhead: while policy optimization is necessary for precise task execution, employing such machinery solely to expand state coverage may be inefficient. In this paper, we propose a new paradigm that explicitly separates exploration from exploitation and bypasses RL during the exploration phase. Our method uses a tree-search strategy inspired by the Go-With-The-Winner algorithm, paired with a measure of epistemic uncertainty to systematically drive exploration. By removing the overhead of policy optimization, our approach explores an order of magnitude more efficiently than standard intrinsic motivation baselines on hard Atari benchmarks. Further, we demonstrate that the discovered trajectories can be distilled into deployable policies using existing supervised backward learning algorithms, achieving state-of-the-art scores by a wide margin on Montezuma's Revenge, Pitfall!, and Venture without relying on domain-specific knowledge. Finally, we demonstrate the generality of our framework in high-dimensional continuous action spaces by solving the MuJoCo Adroit dexterous manipulation and AntMaze tasks in a sparse-reward setting, directly from image observations and without expert demonstrations or offline datasets. To the best of our knowledge, this has not been achieved before.

We propose TiCo, a simple post-training method for enabling spoken dialogue models (SDMs) to follow time-constrained instructions and generate responses with controllable duration. This capability is valuable for real-world spoken language systems such as voice assistants and interactive agents, where controlling response duration can improve interaction quality. However, despite their strong ability to generate natural spoken responses, existing models lack time awareness and struggle to follow duration-related instructions (e.g., "Please generate a response lasting about 15 seconds"). Through an empirical evaluation of both open-source and commercial SDMs, we show that they frequently fail to satisfy such time-control requirements. TiCo addresses this limitation by enabling models to estimate elapsed speaking time during generation through Spoken Time Markers (STM) (e.g., <10.6 seconds>). These markers help the model maintain awareness of time and adjust the remaining content to meet the target duration. TiCo is simple and efficient: it requires only a small amount of data and no additional question-answer pairs, relying instead on self-generation and reinforcement learning. Experimental results show that TiCo significantly improves adherence to duration constraints while preserving response quality.

Hundreds of millions of people rely on large language models (LLMs) for education, work, and even healthcare. Yet these models are known to reproduce and amplify social biases present in their training data. Moreover, text-based interfaces remain a barrier for many, for example, users with limited literacy, motor impairments, or mobile-only devices. Voice interaction promises to expand accessibility, but unlike text, speech carries identity cues that users cannot easily mask, raising concerns about whether accessibility gains may come at the cost of equitable treatment. Here we show that audio-enabled LLMs exhibit systematic gender discrimination, shifting responses toward gender-stereotyped adjectives and occupations solely on the basis of speaker voice, and amplifying bias beyond that observed in text-based interaction. Thus, voice interfaces do not merely extend text models to a new modality but introduce distinct bias mechanisms tied to paralinguistic cues. Complementary survey evidence ($n=1,000$) shows that infrequent chatbot users are most hesitant to undisclosed attribute inference and most likely to disengage when such practices are revealed. To demonstrate a potential mitigation strategy, we show that pitch manipulation can systematically regulate gender-discriminatory outputs. Overall, our findings reveal a critical tension in AI development: efforts to expand accessibility through voice interfaces simultaneously create new pathways for discrimination, demanding that fairness and accessibility be addressed in tandem.

Sodium-ion batteries require anodes that combine high capacity, low operating voltage, fast Na-ion transport, and mechanical stability, which conventional anodes struggle to deliver. Here, we use the SpookyNet machine-learning force field (MLFF) together with all-electron density-functional theory calculations to characterize Na storage in aminobenzene-functionalized Janus graphene (Na$_x$AB) at room-temperature. Simulations across state of charge reveal a three-stage storage mechanism-site-specific adsorption at aminobenzene groups and Na$_n$@AB$_m$ structure formation, followed by interlayer gallery filling-contrasting the multi-stage pore-, graphite-interlayer-, and defect-controlled behavior in hard carbon. This leads to an OCV profile with an extended low-voltage plateau of 0.15 V vs. Na/Na$^{+}$, an estimated gravimetric capacity of $\sim$400 mAh g$^{-1}$, negligible volume change, and Na diffusivities of $\sim10^{-6}$ cm$^{2}$ s$^{-1}$, two to three orders of magnitude higher than in hard carbon. Our results establish Janus aminobenzene-graphene as a promising, structurally defined high-capacity Na-ion anode and illustrate the power of MLFF-based simulations for characterizing electrode materials.

The increasing heterogeneity of high-performance computing (HPC) systems and the transition to exascale architectures require systematic and reproducible performance evaluation across diverse workloads. While continuous integration (CI) ensures functional correctness in software engineering, performance and energy efficiency in HPC are typically evaluated outside CI workflows, motivating continuous benchmarking (CB) as a complementary approach. Integrating benchmarking into CI workflows enables reproducible evaluation, early detection of regressions, and continuous validation throughout the software development lifecycle. We present exaCB, a framework for continuous benchmarking developed in the context of the JUPITER exascale system. exaCB enables application teams to integrate benchmarking into their workflows while supporting large-scale, system-wide studies through reusable CI/CD components, established harnesses, and a shared reporting protocol. The framework supports incremental adoption, allowing benchmarks to be onboarded easily and to evolve from basic runnability to more advanced instrumentation and reproducibility. The approach is demonstrated in JUREAP, the early-access program for JUPITER, where exaCB enabled continuous benchmarking of over 70 applications at varying maturity levels, supporting cross-application analysis, performance tracking, and energy-aware studies. These results illustrate the practicality using exaCB for continuous benchmarking for exascale HPC systems across large, diverse collections of scientific applications.

Diffusion language models (DLMs) have emerged as a promising alternative to autoregressive (AR) models for language modeling, allowing flexible generation order and parallel generation of multiple tokens. However, this flexibility introduces a challenge absent in AR models: the \emph{decoding strategy} -- which determines the order and number of tokens generated at each iteration -- critically affects sampling efficiency. Among decoding strategies explored in practice, confidence-based methods, which adaptively select which and how many tokens to unmask based on prediction confidence, have shown strong empirical performance. Despite this success, our theoretical understanding of confidence-based decoding remains limited. In this work, we develop the first theoretical analysis framework for confidence-based decoding in DLMs. We focus on an entropy sum-based strategy that continues unmasking tokens within each iteration until the cumulative entropy exceeds a threshold, and show that it achieves $\varepsilon$-accurate sampling in KL divergence with an expected number of iterations $\widetilde O(H(X_0)/\varepsilon)$, where $H(X_0)$ denotes the entropy of the target data distribution. Notably, this strategy yields substantial sampling acceleration when the data distribution has low entropy relative to the sequence length, while automatically adapting to the intrinsic complexity of data without requiring prior knowledge or hyperparameter tuning. Overall, our results provide a theoretical foundation for confidence-based decoding and may inform the design of more efficient decoding strategies for DLMs.

Large language model (LLM)-based systems are becoming increasingly popular for solving tasks by constructing executable workflows that interleave LLM calls, information retrieval, tool use, code execution, memory updates, and verification. This survey reviews recent methods for designing and optimizing such workflows, which we treat as agentic computation graphs (ACGs). We organize the literature based on when workflow structure is determined, where structure refers to which components or agents are present, how they depend on each other, and how information flows between them. This lens distinguishes static methods, which fix a reusable workflow scaffold before deployment, from dynamic methods, which select, generate, or revise the workflow for a particular run before or during execution. We further organize prior work along three dimensions: when structure is determined, what part of the workflow is optimized, and which evaluation signals guide optimization (e.g., task metrics, verifier signals, preferences, or trace-derived feedback). We also distinguish reusable workflow templates, run-specific realized graphs, and execution traces, separating reusable design choices from the structures actually deployed in a given run and from realized runtime behavior. Finally, we outline a structure-aware evaluation perspective that complements downstream task metrics with graph-level properties, execution cost, robustness, and structural variation across inputs. Our goal is to provide a clear vocabulary, a unified framework for positioning new methods, a more comparable view of existing body of literature, and a more reproducible evaluation standard for future work in workflow optimizations for LLM agents.

Despite their output being ultimately consumed by human viewers, 3D Gaussian Splatting (3DGS) methods often rely on ad-hoc combinations of pixel-level losses, resulting in blurry renderings. To address this, we systematically explore perceptual optimization strategies for 3DGS by searching over a diverse set of distortion losses. We conduct the first-of-its-kind large-scale human subjective study on 3DGS, involving 39,320 pairwise ratings across several datasets and 3DGS frameworks. A regularized version of Wasserstein Distortion, which we call WD-R, emerges as the clear winner, excelling at recovering fine textures without incurring a higher splat count. WD-R is preferred by raters more than $2.3\times$ over the original 3DGS loss, and $1.5\times$ over current best method Perceptual-GS. WD-R also consistently achieves state-of-the-art LPIPS, DISTS, and FID scores across various datasets, and generalizes across recent frameworks, such as Mip-Splatting and Scaffold-GS, where replacing the original loss with WD-R consistently enhances perceptual quality within a similar resource budget (number of splats for Mip-Splatting, model size for Scaffold-GS), and leads to reconstructions being preferred by human raters $1.8\times$ and $3.6\times$, respectively. We also find that this carries over to the task of 3DGS scene compression, with $\approx 50\%$ bitrate savings for comparable perceptual metric performance.

Diffusion Language Models (DLMs) offer attractive advantages over Auto-Regressive (AR) models, such as full-attention parallel decoding and flexible generation. However, they suffer from a notable train-inference mismatch: DLMs are trained with a static, single-step masked prediction objective, but deployed through a multi-step progressive denoising trajectory. We propose MemDLM (Memory-Enhanced DLM), which narrows this gap by embedding a simulated denoising process into training via Bi-level Optimization. An inner loop updates a set of fast weights, forming a Parametric Memory that captures the local trajectory experience of each sample, while an outer loop updates the base model conditioned on this memory. By offloading memorization pressure from token representations to parameters, MemDLM yields faster convergence and lower training loss. Moreover, the inner loop can be re-enabled at inference time as an adaptation step, yielding additional gains on long-context understanding. We find that, when activated at inference time, this Parametric Memory acts as an emergent in-weight retrieval mechanism, helping MemDLM further reduce token-level attention bottlenecks on challenging Needle-in-a-Haystack retrieval tasks. Code: https://github.com/JarvisPei/MemDLM.

Decision Boundary Maps (DBMs) are an effective tool for visualising machine learning classification boundaries. Yet, DBM quality strongly depends on the dimensionality reduction (DR) technique and high dimensional space used for the data points. For complex ML datasets, DR can create many mixed classes which, in turn, yield DBMs that are hard to use. We propose a new technique to compute DBMs by transforming data space into Shapley space and computing DR on it. Compared to standard DBMs computed directly from data, our maps have similar or higher quality metric values and visibly more compact, easier to explore, decision zones.

Generative Recommender Systems using semantic ids, such as TIGER (Rajput et al., 2023), have emerged as a widely adopted competitive paradigm in sequential recommendation. However, existing architectures are designed solely for semantic retrieval and do not address concerns such as monetization via ad revenue and incorporation of bids for commercial retrieval. We propose GEM-Rec, a unified framework that integrates commercial relevance and monetization objectives directly into the generative sequence. We introduce control tokens to decouple the decision of whether to show an ad from which item to show. This allows the model to learn valid placement patterns directly from interaction logs, which inherently reflect past successful ad placements. Complementing this, we devise a Bid-Aware Decoding mechanism that handles real-time pricing, injecting bids directly into the inference process to steer the generation toward high-value items. We prove that this approach guarantees allocation monotonicity, ensuring that higher bids weakly increase an ad's likelihood of being shown without requiring model retraining. Experiments demonstrate that GEM-Rec allows platforms to dynamically optimize for semantic relevance and platform revenue.

Recent advances in text-to-image (T2I) generation via reinforcement learning (RL) have benefited from reward models that assess semantic alignment and visual quality. However, most existing reward models pay limited attention to fine-grained spatial relationships, often producing images that appear plausible overall yet contain inaccuracies in object positioning. In this work, we present \textbf{SpatialReward}, a verifiable reward model explicitly designed to evaluate spatial layouts in generated images. SpatialReward adopts a multi-stage pipeline: a \emph{Prompt Decomposer} extracts entities, attributes, and spatial metadata from free-form prompts; expert detectors provide accurate visual grounding of object positions and attributes; and a vision-language model applies chain-of-thought reasoning over grounded observations to assess complex spatial relations that are challenging for rule-based methods. To more comprehensively evaluate spatial relationships in generated images, we introduce \textbf{SpatRelBench}, a benchmark covering object attributes, orientation, inter-object relations, and rendered text placement. Experiments on Stable Diffusion and FLUX show that incorporating SpatialReward into RL training consistently improves spatial consistency and overall generation quality, with results aligned more closely to human judgments. These findings indicate that verifiable reward models hold considerable potential for enabling more accurate and controllable optimization in text-to-image generation models.

Conversation is ubiquitous in social life, but the empirical study of this interactive process has been thwarted by tools that are insufficiently modular and unadaptive to researcher needs. To relieve many constraints in conversation research, the current tutorial presents an overview and introduction to a new tool, Dyadic (https://www.chatdyadic.com/), a web-based platform for studying human-human and human-AI conversations using text-based or voice-based chats. Dyadic is distinct from other platforms by offering studies with multiple modalities, AI suggestions (e.g., in human-human studies, AI can suggest responses to a participant), live monitoring (e.g., researchers can evaluate, in real time, chats between communicators), and survey deployment (e.g., Likert-type scales, feeling thermometers, and open-ended text boxes can be sent to humans for in situ evaluations of the interaction), among other consequential features. No coding is required to operate Dyadic directly, and integrations with existing survey platforms are offered.

The limited availability of dysarthric speech data makes cross-lingual detection an important but challenging problem. A key difficulty is that speech representations often encode language-dependent structure that can confound dysarthria detection. We propose a representation-level language shift (LS) that aligns source-language self-supervised speech representations with the target-language distribution using centroid-based vector adaptation estimated from healthy-control speech. We evaluate the approach on oral DDK recordings from Parkinson's disease speech datasets in Czech, German, and Spanish under both cross-lingual and multilingual settings. LS substantially improves sensitivity and F1 in cross-lingual settings, while yielding smaller but consistent gains in multilingual settings. Representation analysis further shows that LS reduces language identity in the embedding space, supporting the interpretation that LS removes language-dependent structure.

Modern time series forecasting is evaluated almost entirely through passive observation of single historical trajectories, rendering claims about a model's robustness to non-stationarity fundamentally unfalsifiable. We propose a paradigm shift toward interventionist, exact-statistical benchmarking. By systematically titrating calibrated Gaussian observation noise into known chaotic and stochastic dynamical systems, we transform forecasting from a black-box sequence matching game into an exact distributional inference task. Because the underlying data-generating process and noise variance are mathematically explicit, evaluation can rely on exact negative log-likelihoods and calibrated distributional tests rather than heuristic approximations. To fully leverage this framework, we extend the Fern architecture into a probabilistic generative model that natively parameterizes the Symmetric Positive Definite (SPD) cone, outputting calibrated joint covariance structures without the computational bottleneck of generic Jacobian modeling. Under this rigorous evaluation, we find that state-of-the-art zero-shot foundation models behave consistently with the context-parroting mechanism, failing systematically under non-stationary regime shifts and elevated noise. In contrast, Fern explicitly captures the invariant measure and multivariate geometry of the underlying dynamics, maintaining structural fidelity and statistically sharp calibration precisely where massive sequence-matching models collapse.

The slow, sequential nature of autoregressive (AR) language models has driven the adoption of parallel decoding methods. However, these non-AR models often sacrifice generation quality as they struggle to model the complex joint distribution of token sequences. To narrow this performance gap, we introduce Gumbel Distillation, a novel distillation technique that enables parallel decoders to learn this distribution effectively. Our method leverages the Gumbel-Max trick to create a deterministic mapping from a latent Gumbel noise space to the output tokens of a high-performing AR teacher. As a model-agnostic technique, Gumbel Distillation seamlessly integrates with diverse parallel decoding architectures, including MDLM and BD3-LM. Experiments on LM1B and OpenWebText show that Gumbel Distillation substantially improves the generation quality of parallel language models, achieving a 30.0% improvement in MAUVE score and 10.5% in generative perplexity over MDLM trained on OpenWebText dataset. Code available at https://github.com/hxixixh/gumbel-distill.

A Large Language Model (LLM) as judge evaluates the quality of victim Machine Learning (ML) models, specifically LLMs, by analyzing their outputs. An LLM as judge is the combination of one model and one specifically engineered judge prompt that contains the criteria for the analysis. The resulting automation of the analysis scales up the complex evaluation of the victim models' free-form text outputs by faster and more consistent judgments compared to human reviewers. Thus, quality and security assessments of LLMs can cover a wide range of the victim models' use cases. Being a comparably new technique, LLMs as judges lack a thorough investigation for their reliability and agreement to human judgment. Our work evaluates the applicability of LLMs as automated quality assessors of victim LLMs. We test the efficacy of 37 differently sized conversational LLMs in combination with 5 different judge prompts, the concept of a second-level judge, and 5 models fine-tuned for the task as assessors. As assessment objective, we curate datasets for eight different categories of judgment tasks and the corresponding ground-truth labels based on human assessments. Our empirical results show a high correlation of LLMs as judges with human assessments, when combined with a suitable prompt, in particular for GPT-4o, several open-source models with $\geqslant$ 32B parameters, and a few smaller models like Qwen2.5 14B.

While large language models (LLMs) are pretrained on massive amounts of data, their knowledge coverage remains incomplete in specialized, data-scarce domains, motivating extensive efforts to study synthetic data generation for knowledge injection. We propose SPA (Scaling Prompt-engineered Augmentation), a simple but tough-to-beat baseline that uses a small set of carefully designed prompts to generate large-scale synthetic data for knowledge injection. Through systematic comparisons, we find that SPA outperforms several strong baselines. Furthermore, we identify two key limitations of prior approaches: (1) while RL-based methods may improve the token efficiency of LLM-based data augmentation at small scale, they suffer from diversity collapse as data scales, leading to diminishing returns; and (2) while multi-stage prompting may outperform simple augmentation methods, their advantages can disappear after careful prompt tuning. Our results suggest that, for knowledge injection, careful prompt design combined with straightforward large-scale augmentation can be surprisingly effective, and we hope SPA can serve as a strong baseline for future studies in this area. Our code is available at https://github.com/Tangkexian/SPA.

Multi-agent applications often execute complex tasks as multi-stage workflows, where each stage is an LLM call whose output becomes part of context for subsequent steps. Existing LLM serving systems largely assume homogeneous clusters with identical model replicas. This design overlooks the potential of heterogeneous deployments, where models of different sizes and capabilities enable finer trade-offs between latency and performance. However, heterogeneity introduces new challenges in scheduling across models with diverse throughput and performance. We present Chimera, a predictive scheduling system for multi-agent workflow serving on heterogeneous LLM clusters that jointly improves end-to-end latency and task performance. Chimera applies semantic routing to estimate per-model confidence scores for each request, predicts the total remaining output length of the workflow, and estimates per-model congestion using in-flight predicted token volumes for load balancing. We evaluate Chimera on representative agentic workflows for code generation and math reasoning using multiple heterogeneous LLM configurations. Across comparable settings, Chimera traces the best latency-performance frontier, reducing end-to-end latency by 1.2--2.4$\times$ and improving task performance by 8.0-9.5 percentage points on average over competitive baselines including vLLM.

This is the fourth paper in the CayleyPy project, which applies AI methods to the exploration of large graphs. In this work, we suggest the existence of a new discrete version of holographic string dualities for this setup, and discuss their relevance to AI systems and mathematics. Many modern AI tasks -- such as those addressed by GPT-style language models or RL systems -- can be viewed as direct analogues of predicting particle trajectories on graphs. We investigate this problem for a large family of Cayley graphs, for which we show that surprisingly it admits a dual description in terms of discrete strings. We hypothesize that such dualities may extend to a range of AI systems where they can lead to more efficient computational approaches. In particular, string holographic images of states are proposed as natural candidates for data embeddings, motivated by the "complexity = volume" principle in AdS/CFT. For Cayley graphs of the symmetric group S_n, our results indicate that the corresponding dual objects are flat, planar polygons. The diameter of the graph is equal to the number of integer points inside the polygon scaled by n. Vertices of the graph can be mapped holographically to paths inside the polygon, and the usual graph distances correspond to the area under the paths, thus directly realising the "complexity = volume" paradigm. We also find evidence for continuous CFTs and dual strings in the large n limit. We confirm this picture and other aspects of the duality in a large initial set of examples. We also present new datasets (obtained by a combination of ML and conventional tools) which should be instrumental in establishing the duality for more general cases.

Recent advances in Multimodal Large Language Models (MLLMs) have enabled automated generation of structured layouts from natural language descriptions. Existing methods typically follow a code-only paradigm that generates code to represent layouts, which are then rendered by graphic engines to produce final images. However, they are blind to the rendered visual outcome, making it difficult to guarantee readability and aesthetics. In this paper, we identify visual feedback as a critical factor in layout generation and propose Visual Feedback Layout Model (VFLM), a self-improving framework that leverages visual feedback iterative refinement. VFLM is capable of performing adaptive reflective generation, which leverages visual information to reflect on previous issues and iteratively generates outputs until satisfactory quality is achieved. It is achieved through reinforcement learning with a visually grounded reward model that incorporates OCR accuracy. By rewarding only the final generated outcome, we can effectively stimulate the model's iterative and reflective generative capabilities. Experiments across multiple benchmarks show that VFLM consistently outperforms advanced MLLMs, existing layout models, and code-only baselines, establishing visual feedback as critical for design-oriented MLLMs. Our code and data are available at https://github.com/FolSpark/VFLM.

In Machine Translation, Large Language Models (LLMs) have generally underperformed compared to conventional encoder-decoder systems and thus see limited adoption. However, LLMs excel at modeling contextual information, making them a natural fit for document-level translation tasks where coherence across sentences is crucial. Despite this potential, document-level MT with LLMs faces two key challenges: (1) the scarcity of large-scale, high-quality document-level parallel data; and (2) the propensity of LLMs to introduce hallucinations and omissions during generation. To address these challenges, we propose a two-stage fine-tuning strategy leveraging LLM-augmented document-level data. First, we augment data by converting summarization data into document-level parallel data using a LLM, and then filter it using multiple metrics, leveraging sacreBLEU, COMET, and LaBSE-based cosine similarity-to improve data quality. Finally, we employ a two-stage fine-tuning strategy: first fine-tuning on the abundant sentence-level MT resources, and then on the filtered document-level corpus.

Recent advances in large language models (LLMs) have enabled the automation of an increasing number of programming tasks, including code generation for scientific and engineering domains. In rapidly evolving software ecosystems such as quantum software development, where frameworks expose complex abstractions, a central question is how best to incorporate domain knowledge into LLM-based assistants while preserving maintainability as libraries evolve. In this work, we study specialization strategies for Qiskit code generation using the Qiskit-HumanEval benchmark. We compare a parameter-specialized fine-tuned baseline introduced in prior work against a range of recent general-purpose LLMs enhanced with retrieval-augmented generation (RAG) and agent-based inference with execution feedback. Our results show that modern general-purpose LLMs consistently outperform the parameter-specialized baseline. While the fine-tuned model achieves approximately 47% pass@1 on Qiskit-HumanEval, recent general-purpose models reach 60-65% under zero-shot and retrieval-augmented settings, and up to 85% for the strongest evaluated model when combined with iterative execution-feedback agents -representing an improvement of more than 20% over zero-shot general-purpose performance and more than 35% over the parameter-specialized baseline. Agentic execution feedback yields the most consistent improvements, albeit at increased runtime cost, while RAG provides modest and model-dependent gains. These findings indicate that performance gains can be achieved without domain-specific fine-tuning, instead relying on inference-time augmentation, thereby enabling a more flexible and maintainable approach to LLM-assisted quantum software development.

Autonomous agents operating in continuous environments must decide not only what to do, but when to act. We introduce a lightweight adaptive temporal control system that learns the optimal interval between cognitive ticks from experience, replacing ad hoc biologically inspired timers with a principled learned policy. The policy state is augmented with a predictive hyperbolic spread signal (a "curvature signal" shorthand) derived from hyperbolic geometry: the mean pairwise Poincare distance among n sampled futures embedded in the Poincare ball. High spread indicates a branching, uncertain future and drives the agent to act sooner; low spread signals predictability and permits longer rest intervals. We further propose an interval-aware reward that explicitly penalises inefficiency relative to the chosen wait time, correcting a systematic credit-assignment failure of naive outcome-based rewards in timing problems. We additionally introduce a joint spatio-temporal embedding (ATCPG-ST) that concatenates independently normalised state and position projections in the Poincare ball; spatial trajectory divergence provides an independent timing signal unavailable to the state-only variant (ATCPG-SO). This extension raises mean hyperbolic spread (kappa) from 1.88 to 3.37 and yields a further 5.8 percent efficiency gain over the state-only baseline. Ablation experiments across five random seeds demonstrate that (i) learning is the dominant efficiency factor (54.8 percent over no-learning), (ii) hyperbolic spread provides significant complementary gain (26.2 percent over geometry-free control), (iii) the combined system achieves 22.8 percent efficiency over the fixed-interval baseline, and (iv) adding spatial position information to the spread embedding yields an additional 5.8 percent.

Cardiovascular disease remains the leading cause of global mortality, with progress hindered by human interpretation of complex cardiac tests. Current AI vision-language models are limited to single-modality inputs and are non-interactive. We present MARCUS (Multimodal Autonomous Reasoning and Chat for Ultrasound and Signals), an agentic vision-language system for end-to-end interpretation of electrocardiograms (ECGs), echocardiograms, and cardiac magnetic resonance imaging (CMR) independently and as multimodal input. MARCUS employs a hierarchical agentic architecture comprising modality-specific vision-language expert models, each integrating domain-trained visual encoders with multi-stage language model optimization, coordinated by a multimodal orchestrator. Trained on 13.5 million images (0.25M ECGs, 1.3M echocardiogram images, 12M CMR images) and our novel expert-curated dataset spanning 1.6 million questions, MARCUS achieves state-of-the-art performance surpassing frontier models (GPT-5 Thinking, Gemini 2.5 Pro Deep Think). Across internal (Stanford) and external (UCSF) test cohorts, MARCUS achieves accuracies of 87-91% for ECG, 67-86% for echocardiography, and 85-88% for CMR, outperforming frontier models by 34-45% (P<0.001). On multimodal cases, MARCUS achieved 70% accuracy, nearly triple that of frontier models (22-28%), with 1.7-3.0x higher free-text quality scores. Our agentic architecture also confers resistance to mirage reasoning, whereby vision-language models derive reasoning from unintended textual signals or hallucinated visual content. MARCUS demonstrates that domain-specific visual encoders with an agentic orchestrator enable multimodal cardiac interpretation. We release our models, code, and benchmark open-source.

Autonomous agents operating in uncertain environments must balance fast responses with goal-directed planning. Classical MF RL often converges slowly and may induce unsafe exploration, whereas MB methods are computationally expensive and sensitive to model mismatch. This paper presents a human-inspired hybrid RL architecture integrating Pavlovian, Instrumental MF, and Instrumental MB components. Inspired by Pavlovian and Instrumental learning from neuroscience, the framework considers contextual radio cues, here intended as georeferenced environmental features acting as CS, to shape intrinsic value signals and bias decision-making. Learning is further modulated by internal motivational drives through a dedicated motivational signal. A Bayesian arbitration mechanism adaptively blends MF and MB estimates based on predicted reliability. Simulation results show that the hybrid approach accelerates learning, improves operational safety, and reduces navigation in high-uncertainty regions compared to standard RL baselines. Pavlovian conditioning promotes safer exploration and faster convergence, while arbitration enables a smooth transition from exploration to efficient, plan-driven exploitation. Overall, the results highlight the benefits of biologically inspired modularity for robust and adaptive autonomous systems under uncertainty.

We study calibeating, the problem of post-processing external forecasts online to minimize cumulative losses and match an informativeness-based benchmark. Unlike prior work, which analyzed calibeating for specific losses with specific arguments, we reduce calibeating to existing online learning techniques and obtain results for general proper losses. More concretely, we first show that calibeating is minimax-equivalent to regret minimization. This recovers the $O(\log T)$ calibeating rate of Foster and Hart [FH23] for the Brier and log losses and its optimality, and yields new optimal calibeating rates for mixable losses and general bounded losses. Second, we prove that multi-calibeating is minimax-equivalent to the combination of calibeating and the classical expert problem. This yields new optimal multi-calibeating rates for mixable losses, including Brier and log losses, and general bounded losses. Finally, we obtain new bounds for achieving calibeating and calibration simultaneously for the Brier loss. For binary predictions, our result gives the first calibrated algorithm that at the same time also achieves the optimal $O(\log T)$ calibeating rate.

Metacognition -- the ability to assess one's own cognitive performance -- is documented across species, with internal confidence estimates serving as a key signal for adaptive behavior. While confidence can be extracted from Large Language Model (LLM) outputs, whether models actively use these signals to regulate behavior remains a fundamental question. We investigate this through a four-phase abstention paradigm.Phase 1 established internal confidence estimates in the absence of an abstention option. Phase 2 revealed that LLMs apply implicit thresholds to these estimates when deciding to answer or abstain. Confidence emerged as the dominant predictor of behavior, with effect sizes an order of magnitude larger than knowledge retrieval accessibility (RAG scores) or surface-level semantic features. Phase 3 provided causal evidence through activation steering: manipulating internal confidence signals correspondingly shifted abstention rates. Finally, Phase 4 demonstrated that models can systematically vary abstention policies based on instructed thresholds.Our findings indicate that abstention arises from the joint operation of internal confidence representations and threshold-based policies, mirroring the two-stage metacognitive control found in biological systems. This capacity is essential as LLMs transition into autonomous agents that must recognize their own uncertainty to decide when to act or seek help.

The development of machine learning interatomic potentials faces a critical computational bottleneck with the generation and labeling of useful training datasets. We present a novel application of determinantal point processes (DPPs) to the task of selecting informative subsets of atomic configurations to label with reference energies and forces from costly quantum mechanical methods. Through experiments with hafnium oxide data, we show that DPPs are competitive with existing approaches to constructing compact but diverse training sets by utilizing kernels of molecular descriptors, leading to improved accuracy and robustness in machine learning representations of molecular systems. Our work identifies promising directions to employ DPPs for unsupervised training data curation with heterogeneous or multimodal data, or in online active learning schemes for iterative data augmentation during molecular dynamics simulation.

We study multimodal survival analysis integrating clinical text, tabular covariates, and genomic profiles using locally deployable large language models (LLMs). As many institutions face tight computational and privacy constraints, this setting motivates the use of lightweight, on-premises models. Our approach jointly estimates calibrated survival probabilities and generates concise, evidence-grounded prognosis text via teacher-student distillation and principled multimodal fusion. On a TCGA cohort, it outperforms standard baselines, avoids reliance on cloud services and associated privacy concerns, and reduces the risk of hallucinated or miscalibrated estimates that can be observed in base LLMs.

A celebrated method for Variational Inequalities (VIs) is Extragradient (EG), which can be viewed as a standard discrete-time integration scheme. With this view in mind, in this paper we show that EG may suffer from discretization bias when applied to non-linear vector fields, conservative or otherwise. To resolve this discretization shortcoming, we introduce RAndomized Mid-Point for debiAsed Gradient Extrapolation (RAMPAGE) and its variance-reduced counterpart, RAMPAGE+ which leverages antithetic sampling. In contrast with EG, both methods are unbiased. Furthermore, leveraging negative correlation, RAMPAGE+ acts as an unbiased, geometric path-integrator that completely removes internal first-order terms from the variance, provably improving upon RAMPAGE. We further demonstrate that both methods enjoy provable $\mathcal{O}(1/k)$ convergence guarantees for a range of problems including root finding under co-coercive, co-hypomonotone, and generalized Lipschitzness regimes. Furthermore, we introduce symmetrically scaled variants to extend our results to constrained VIs. Finally, we provide convergence guarantees of both methods for stochastic and deterministic smooth convex-concave games. Somewhat interestingly, despite being a randomized method, RAMPAGE+ attains purely deterministic bounds for a number of the studied settings.

This paper proposes $\mathrm{dynActivation}$, a per-layer trainable activation defined as $f_i(x) = \mathrm{BaseAct}(x)(α_i - β_i) + β_i x$, where $α_i$ and $β_i$ are lightweight learned scalars that interpolate between the base nonlinearity and a linear path and $\mathrm{BaseAct}(x)$ resembles any ReLU-like function. The static and dynamic ReLU-like variants are then compared across multiple vision tasks, language modeling tasks, and ablation studies. The results suggest that dynActivation variants tend to linearize deep layers while maintaining high performance, which can improve training efficiency by up to $+54\%$ over ReLU. On CIFAR-10, dynActivation(Mish) improves over static Mish by up to $+14.02\%$ on AttentionCNN with an average improvment by $+6.00\%$, with a $24\%$ convergence-AUC reduction relative to Mish (2120 vs. 2785). In a 1-to-75-layer MNIST depth-scaling study, dynActivation never drops below $95\%$ test accuracy ($95.3$--$99.3\%$), while ReLU collapses below $80\%$ at 25 layers. Under FGSM at $\varepsilon{=}0.08$, dynActivation(Mish) incurs a $55.39\%$ accuracy drop versus $62.79\%$ for ReLU ($7.40\%$ advantage). Transferred to language modeling, a new proposed dynActGLU-variant achieves a $10.3\%$ relative perplexity reduction over SwiGLU at 5620 steps (4.047 vs. 4.514), though the gap vanishes at 34300 steps.

Recent advances in cross-view geo-localization (CVGL) methods have shown strong potential for supporting unmanned aerial vehicle (UAV) navigation in GNSS-denied environments. However, existing work predominantly focuses on matching UAV views to onboard map tiles, which introduces an inherent trade-off between accuracy and storage overhead, and overlooks the importance of the UAV's heading during navigation. Moreover, the substantial discrepancies and varying overlaps in cross-view scenarios have been insufficiently considered, limiting their generalization to real-world scenarios. In this paper, we present Bearing-UAV, a purely vision-driven cross-view navigation method that jointly predicts UAV absolute location and heading from neighboring features, enabling accurate, lightweight, and robust navigation in the wild. Our method leverages global and local structural features and explicitly encodes relative spatial relationships, making it robust to cross-view variations, misalignment, and feature-sparse conditions. We also present Bearing-UAV-90k, a multi-city benchmark for evaluating cross-view localization and navigation. Extensive experiments show encouraging results that Bearing-UAV yields lower localization error than previous matching/retrieval paradigm across diverse terrains. Our code and dataset will be made publicly available.

Just as people improve decision-making by consulting diverse human advisors, they can now also consult with multiple AI systems. Prior work on group decision-making shows that advice aggregation creates pressure to conform, leading to overreliance. However, the conditions under which multi-AI consultation improves or undermines human decision-making remain unclear. We conducted experiments with three tasks in which participants received advice from panels of AIs. We varied panel size, within-panel consensus, and the human-likeness of presentation. Accuracy improved for small panels relative to a single AI; larger panels yielded no gains. The level of within-panel consensus affected participants' reliance on AI advice: High consensus fostered overreliance; a single dissent reduced pressure to conform; wide disagreement created confusion and undermined appropriate reliance. Human-like presentations increased perceived usefulness and agency in certain tasks, without raising conformity pressure. These findings yield design implications for presenting multi-AI advice that preserve accuracy while mitigating conformity.

Quantum computers have the potential to solve certain complex problems in a much more efficient way than classical computers. Nevertheless, current quantum computer implementations are limited by high physical error rates. This issue is addressed by Quantum Error Correction (QEC) codes, which use multiple physical qubits to form a logical qubit to achieve a lower logical error rate, with the surface code being one of the most commonly used. The most time-critical step in this process is interpreting the measurements of the physical qubits to determine which errors have most likely occurred - a task called decoding. Consequently, the main challenge for QEC is to achieve error correction with high accuracy within the tight $1μs$ decoding time budget imposed by superconducting qubits. State-of-the-art QEC approaches trade accuracy for latency. In this work, we propose an FPGA accelerator for a Neural Network based decoder as a way to achieve a lower logical error rate than current methods within the tight time constraint, for code distance up to d=7. We achieved this goal by applying different hardware-aware optimizations to a high-accuracy GNN-based decoder. In addition, we propose several accelerator optimizations leading to the FPGA-based decoder achieving a latency smaller than $1μs$, with a lower error rate compared to the state-of-the-art.

Semantic data and knowledge infrastructures must reconcile two fundamentally different forms of representation: natural language, in which most knowledge is created and communicated, and formal semantic models, which enable machine-actionable integration, interoperability, and reasoning. Bridging this gap remains a central challenge, particularly when full semantic formalization is required at the point of data entry. Here, we introduce the Semantic Ladder, an architectural framework that enables the progressive formalization of data and knowledge. Building on the concept of modular semantic units as identifiable carriers of meaning, the framework organizes representations across levels of increasing semantic explicitness, ranging from natural language text snippets to ontology-based and higher-order logical models. Transformations between levels support semantic enrichment, statement structuring, and logical modelling while preserving semantic continuity and traceability. This approach enables the incremental construction of semantic knowledge spaces, reduces the semantic parsing burden, and supports the integration of heterogeneous representations, including natural language, structured semantic models, and vector-based embeddings. The Semantic Ladder thereby provides a foundation for scalable, interoperable, and AI-ready data and knowledge infrastructures.

While both classical and neural network classifiers can achieve high accuracy, they fall short on offering uncertainty bounds on their predictions, making them unfit for safety-critical applications. Existing kernel-based classifiers that provide such bounds scale with $\mathcal O (n^{\sim3})$ in time, making them computationally intractable for large datasets. To address this, we propose a novel, computationally efficient classification algorithm based on the Nadaraya-Watson estimator, for whose estimates we derive frequentist uncertainty intervals. We evaluate our classifier on synthetically generated data and on electrocardiographic heartbeat signals from the MIT-BIH Arrhythmia database. We show that the method achieves competitive accuracy $>$\SI{96}{\percent} at $\mathcal O(n)$ and $\mathcal O(\log n)$ operations, while providing actionable uncertainty bounds. These bounds can, e.g., aid in flagging low-confidence predictions, making them suitable for real-time settings with resource constraints, such as diagnostic monitoring or implantable devices.

Text-driven video moment retrieval (VMR) remains challenging due to limited capture of hidden temporal dynamics in untrimmed videos, leading to imprecise grounding in long sequences. Traditional methods rely on natural language queries (NLQs) or static image augmentations, overlooking motion sequences and suffering from high computational costs in Transformer-based architectures. Existing approaches fail to integrate subtitle contexts and generated temporal priors effectively, we therefore propose a novel two-stage framework for enhanced temporal grounding. In the first stage, LLM-guided subtitle matching identifies relevant textual cues from video subtitles, fused with the query to generate auxiliary short videos via text-to-video models, capturing implicit motion information as temporal priors. In the second stage, augmented queries are processed through a multi-modal controlled Mamba network, extending text-controlled selection with video-guided gating for efficient fusion of generated priors and long sequences while filtering noise. Our framework is agnostic to base retrieval models and widely applicable for multimodal VMR. Experimental evaluations on the TVR benchmark demonstrate significant improvements over state-of-the-art methods, including reduced computational overhead and higher recall in long-sequence grounding.

Reinforcement learning with verifiable rewards (RLVR) has substantially improved the reasoning capabilities of large language models. While existing analyses identify that RLVR-induced changes are sparse, they primarily focus on the \textbf{magnitude} of these updates, largely overlooking their \textbf{direction}. In this work, we argue that the direction of updates is a more critical lens for understanding RLVR's effects, which can be captured by the signed, token-level log probability difference $Δ\log p$ between the base and final RLVR models. Through statistical analysis and token-replacement interventions, we demonstrate that $Δ\log p$ more effectively identifies sparse, yet reasoning-critical updates than magnitude-based metrics (\eg divergence or entropy). Building on this insight, we propose two practical applications: (1) a \textit{test-time extrapolation} method that amplifies the policy along the learned $Δ\log p$ direction to improve reasoning accuracy without further training; (2) a \textit{training-time reweighting} method that focuses learning on low-probability (corresponding to higher $Δ\log p$) tokens, which improves reasoning performance across models and benchmarks. Our work establishes the direction of change as a key principle for analyzing and improving RLVR.

Deductive verification provides strong correctness guarantees for code by extracting verification conditions (VCs) and writing formal proofs for them. The expertise-intensive task of VC proving is the main bottleneck in this process, and has been partly automated owing to recent advances in Large Language Model (LLM) agents. However, existing proof agents are not able to discover helper lemmas - auxiliary lemmas that aid in proving - and thus fall short as programs grow in size and complexity. In this paper, we argue that VC proving for program verification is more than a purely mathematical task, and benefits considerably from program comprehension. Our key insight is that human-proof engineers often discover and apply helper lemmas based on their understanding of the program semantics, which are not directly reflected in the VCs produced by VC generators. Inspired by this insight, we propose an LLM agent, LemmaNet, that discovers helper lemmas in two ways. Specifically, the agent first synthesizes lemmas offline by directly analyzing the source code and specifications, and then relating this semantic understanding to the mechanical, verbose encoding produced by VC generators. As the proof unfolds, LemmaNet then adapts existing helper lemmas online to accommodate evolving proof states, enabling the agent to effectively discharge complex VCs on-the-fly. We evaluate LemmaNet on SV-COMP and established real-world subjects, including modules of the Linux kernel, Contiki OS, standard C++ library, and X.509 parser. Our experimental results demonstrate that LemmaNet significantly outperforms state-of-the-art approaches, highlighting the importance of program comprehension-aided lemma discovery in agentic program verification.

Generative AI is accelerating software development, but may quietly shift where the real risks lie. As AI generates code faster than teams can understand it, two under appreciated forms of debt accumulate: cognitive debt, the erosion of shared understanding across a team, and intent debt, the absence of externalized rationale that both developers and AI agents need to work safely with code. This article proposes a Triple Debt Model for reasoning about software health built around three interacting debt types: technical debt in code, cognitive debt in people, and intent debt in externalized knowledge. Cognitive debt concerns what people understand; intent debt concerns what is explicitly captured for humans and machines to use. We discuss how generative AI changes the relative importance of these debt types, how each can be diagnosed and mitigated, and surfaced points of debate for practitioners.

Predicting narrative similarity can be understood as an inherently interpretive task: different, equally valid readings of the same text can produce divergent interpretations and thus different similarity judgments, posing a fundamental challenge for semantic evaluation benchmarks that encode a single ground truth. Rather than treating this multiperspectivity as a challenge to overcome, we propose to incorporate it in the decision making process of predictive systems. To explore this strategy, we created an ensemble of 31 LLM personas. These range from practitioners following interpretive frameworks to more intuitive, lay-style characters. Our experiments were conducted on the SemEval-2026 Task 4 dataset, where the system achieved an accuracy score of 0.705. Accuracy improves with ensemble size, consistent with Condorcet Jury Theorem-like dynamics under weakened independence. Practitioner personas perform worse individually but produce less correlated errors, yielding larger ensemble gains under majority voting. Our error analysis reveals a consistent negative association between gender-focused interpretive vocabulary and accuracy across all persona categories, suggesting either attention to dimensions not relevant for the benchmark or valid interpretations absent from the ground truth. This finding underscores the need for evaluation frameworks that account for interpretive plurality.

Foundation models are now increasingly being developed for Earth observation (EO), yet they often rely on stochastic masking that do not explicitly enforce physics constraints; a critical trustworthiness limitation, in particular for predictive models that guide public health decisions. In this work, we propose SpecTM (Spectral Targeted Masking), a physics-informed masking design that encourages the reconstruction of targeted bands from cross-spectral context during pretraining. To achieve this, we developed an adaptable multi-task (band reconstruction, bio-optical index inference, and 8-day-ahead temporal prediction) self-supervised learning (SSL) framework that encodes spectrally intrinsic representations via joint optimization, and evaluated it on a downstream microcystin concentration regression model using NASA PACE hyperspectral imagery over Lake Erie. SpecTM achieves R^2 = 0.695 (current week) and R^2 = 0.620 (8-day-ahead) predictions surpassing all baseline models by (+34% (0.51 Ridge) and +99% (SVR 0.31)) respectively. Our ablation experiments show targeted masking improves predictions by +0.037 R^2 over random masking. Furthermore, it outperforms strong baselines with 2.2x superior label efficiency under extreme scarcity. SpecTM enables physics-informed representation learning across EO domains and improves the interpretability of foundation models.

Clinical decision-making agents can benefit from reusing prior decision experience. However, many memory-augmented methods store experiences as independent records without explicit relational structure, which may introduce noisy retrieval, unreliable reuse, and in some cases even hurt performance compared to direct LLM inference. We propose GSEM (Graph-based Self-Evolving Memory), a clinical memory framework that organizes clinical experiences into a dual-layer memory graph, capturing both the decision structure within each experience and the relational dependencies across experiences, and supporting applicability-aware retrieval and online feedback-driven calibration of node quality and edge weights. Across MedR-Bench and MedAgentsBench with two LLM backbones, GSEM achieves the highest average accuracy among all baselines, reaching 70.90\% and 69.24\% with DeepSeek-V3.2 and Qwen3.5-35B, respectively. Code is available at https://github.com/xhan1022/gsem.

Despite rapid progress in AI agents for enterprise automation and decision-making, their real-world deployment and further performance gains remain constrained by limited data quality and quantity, complex real-world reasoning demands, difficulties with self-play, and the lack of reliable feedback signals. To address these challenges, we propose a lightweight, model-agnostic framework for improving LLM-based enterprise agents via offline reinforcement learning (RL). The proposed Context Engineering via DT-MDP (DT-MDP-CE) framework comprises three key components: (1) A Digital-Twin Markov Decision Process (DT-MDP), which abstracts the agent's reasoning behavior as a finite MDP; (2) A robust contrastive inverse RL, which, armed with the DT-MDP, to efficiently estimate a well-founded reward function and induces policies from mixed-quality offline trajectories; and (3) RL-guided context engineering, which uses the policy obtained from the integrated process of (1) and (2), to improve the agent's decision-making behavior. As a case study, we apply the framework to a representative task in the enterprise-oriented domain of IT automation. Extensive experimental results demonstrate consistent and significant improvements over baseline agents across a wide range of evaluation settings, suggesting that the framework can generalize to other agents sharing similar characteristics in enterprise environments.

We present a controlled empirical comparison between autoregressive (AR) and masked diffusion (MDLM) language models. Both models are trained on identical data (50M tokens from TinyStories), identical compute budget (20,000 steps, batch size 32, sequence length 512), and identical hardware (NVIDIA H100 80GB), isolating the generation paradigm as the sole variable. We report three findings. First, both paradigms achieve comparable training throughput (~50K tokens/second), with MDLM requiring only 4.7% more wall-clock time. Second, AR converges faster and begins overfitting by step 14,000, while MDLM converges more slowly and is still improving at step 20,000, suggesting different compute-optimal training regimes. Third, quantitative diversity analysis over 1,000 generated samples reveals a structural diversity-fluency trade-off: AR produces fluent but repetitive outputs (99.8% begin with the same word), while MDLM generates more diverse narratives (93.4% unique 5-word openings, higher Distinct-n, lower Self-BLEU), at the cost of occasional grammatical inconsistencies. All code, trained checkpoints, and data pipelines are released for reproducibility.

Due to the prevalence of temporal data and its inherent dependencies in many real-world problems, time series classification is of paramount importance in various domains. However, existing models often struggle with series of variable length or high dimensionality. This paper introduces the MIHT (Multi-instance Hoeffding Tree) algorithm, an efficient model that uses multi-instance learning to classify multivariate and variable-length time series while providing interpretable results. The algorithm uses a novel representation of time series as "bags of subseries," together with an optimization process based on incremental decision trees that distinguish relevant parts of the series from noise. This methodology extracts the underlying concept of series with multiple variables and variable lengths. The generated decision tree is a compact, white-box representation of the series' concept, providing interpretability insights into the most relevant variables and segments of the series. Experimental results demonstrate MIHT's superiority, as it outperforms 11 state-of-the-art time series classification models on 28 public datasets, including high-dimensional ones. MIHT offers enhanced accuracy and interpretability, making it a promising solution for handling complex, dynamic time series data.

Multi-objective re-ranking has become a critical component of modern multi-stage recommender systems, as it tasked to balance multiple conflicting objectives such as accuracy, diversity, and fairness. Existing multi-objective re-ranking methods typically optimize aggregate objectives at the item level using static or handcrafted preference weights. This design overlooks that users inherently exhibit Pareto-optimal preferences at the intent level, reflecting personalized trade-offs among objectives rather than fixed weight combinations. Moreover, most approaches treat re-ranking task for each user as an isolated problem, and repeatedly learn the preferences from scratch. Such a paradigm not only incurs high computational cost, but also ignores the fact that users often share similar preference trade-off structures across objectives. Inspired by the existence of homogeneous multi-objective optimization spaces where Pareto-optimal patterns are transferable, we propose PreferRec, a novel framework that explicitly models and transfers Pareto preferences across users. Specifically, PreferRec is built upon three tightly coupled components: Preference-Aware Pareto Learning aims to capture user intrinsic trade-offs among multiple conflicting objectives at the intent level. By learning Pareto preference representations from re-ranking populations, this component explicitly models how users prioritize different objectives under diverse contexts. Knowledge-Guided Transfer facilitates efficient cross-user knowledge transfer by distilling shared optimization patterns across homogeneous optimization spaces. The transferred knowledge is then used to guide solution selection and personalized re-ranking, biasing the optimization process toward high-quality regions of the Pareto front while preserving user-specific preference characteristics.

Inasmuch as the removal of refusal behavior from instruction-tuned language models by directional abliteration requires the extraction of refusal-mediating directions from the residual stream activation space, and inasmuch as the construction of the contrast baseline against which harmful prompt activations are compared has been treated in the existing literature as an implementation detail rather than a methodological concern, the present work investigates whether a topically matched contrast baseline yields superior refusal directions. The investigation is carried out on the Qwen~3.5 2B model using per-category matched prompt pairs, per-class Self-Organizing Map extraction, and Singular Value Decomposition orthogonalization. It was found that topic-matched contrast produces no functional refusal directions at any tested weight level on any tested layer, while unmatched contrast on the same model, same extraction code, and same evaluation protocol achieves complete refusal elimination on six layers. The geometric analysis of the failure establishes that topic-matched subtraction cancels the dominant activation component shared between harmful and harmless prompts of the same subject, reducing the extracted direction magnitude below the threshold at which weight-matrix projection perturbs the residual stream. The implications for the design of contrast baselines in abliteration research are discussed.

Large language models (LLMs) achieve state-of-the-art (SOTA) performance across language tasks, but are costly to deploy due to their size and resource demands. Knowledge Distillation (KD) addresses this by training smaller Student models to mimic larger Teacher models, improving efficiency without significant performance loss. Dual-Space Knowledge Distillation with Cross-Model Attention (DSKD-CMA) has emerged as a SOTA method for KD between LLMs with distinct tokenizers, yet its internal workings remain largely opaque. In this work, we systematically analyse the attention mechanism of DSKD-CMA through manual token alignment probing and heatmap visualisations, revealing both strengths and limitations. Building on this, we introduce a novel method, DSKD-CMA-GA, based on Generative Adversarial (GA) learning, to address the mismatched distributions between the keys and queries computed from distinct models. Experiments show modest but consistent ROUGE-L gains in text generation quality, particularly on out-of-distribution data (+0.37 on average), narrowing the gap between cross- and same-tokenizer KD.

Animal vocalizations provide crucial insights for wildlife assessment, particularly in complex environments such as forests, aiding species identification and ecological monitoring. Recent advances in deep learning have enabled automatic species classification from their vocalizations. However, classifying species unseen during training remains challenging. To address this limitation, we introduce AnimalCLAP, a taxonomy-aware language-audio framework comprising a new dataset and model that incorporate hierarchical biological information. Specifically, our vocalization dataset consists of 4,225 hours of recordings covering 6,823 species, annotated with 22 ecological traits. The AnimalCLAP model is trained on this dataset to align audio and textual representations using taxonomic structures, improving the recognition of unseen species. We demonstrate that our proposed model effectively infers ecological and biological attributes of species directly from their vocalizations, achieving superior performance compared to CLAP. Our dataset, code, and models will be publicly available at https://dahlian00.github.io/AnimalCLAP_Page/.

Supervised machine learning describes the practice of fitting a parameterized model to labeled input-output data. Supervised machine learning methods have demonstrated promise in learning efficient surrogate models that can (partially) replace expensive high-fidelity models, making many-query analyses, such as optimization, uncertainty quantification, and inference, tractable. However, when training data must be obtained through the evaluation of an expensive model or experiment, the amount of training data that can be obtained is often limited, which can make learned surrogate models unreliable. However, in many engineering and scientific settings, cheaper \emph{low-fidelity} models may be available, for example arising from simplified physics modeling or coarse grids. These models may be used to generate additional low-fidelity training data. The goal of \emph{multifidelity} machine learning is to use both high- and low-fidelity training data to learn a surrogate model which is cheaper to evaluate than the high-fidelity model, but more accurate than any available low-fidelity model. This work proposes a new multifidelity training approach for Gaussian process regression which uses low-fidelity data to define additional features that augment the input space of the learned model. The approach unites desirable properties from two separate classes of existing multifidelity GPR approaches, cokriging and autoregressive estimators. Numerical experiments on several test problems demonstrate both increased predictive accuracy and reduced computational cost relative to the state of the art.

Translating natural language descriptions into viable code fixes remains a fundamental challenge in software engineering. While the proliferation of agentic large language models (LLMs) has vastly improved automated repository-level debugging, current frameworks hit a ceiling when dealing with sophisticated bugs like implicit type degradations and complex polymorphic control flows. Because these methods rely heavily on static analysis and superficial execution feedback, they lack visibility into intermediate runtime states. Consequently, agents are forced into costly, speculative trial-and-error loops, wasting computational tokens without successfully isolating the root cause. To bridge this gap, we propose DAIRA (Dynamic Analysis-enhanced Issue Resolution Agent), a pioneering automated repair framework that natively embeds dynamic analysis into the agent's reasoning cycle. Driven by a Test Tracing-Driven methodology, DAIRA utilizes lightweight monitors to extract critical runtime data -- such as variable mutations and call stacks -- and synthesizes them into structured semantic reports. This mechanism fundamentally shifts the agent's behavior from blind guesswork to evidence-based, deterministic deduction. When powered by Gemini 3 Flash Preview, DAIRA establishes a new state-of-the-art (SOTA) performance, achieving a 79.4% resolution rate on the SWE-bench Verified dataset. Compared to existing baselines, our framework not only conquers highly complex defects but also cuts overall inference expenses by roughly 10% and decreases input token consumption by approximately 25%.

While Vision-Language Models (VLMs) have achieved remarkable performance, their Euclidean embeddings remain limited in capturing hierarchical relationships such as part-to-whole or parent-child structures, and often face challenges in multi-object compositional scenarios. Hyperbolic VLMs mitigate this issue by better preserving hierarchical structures and modeling part-whole relations (i.e., whole scene and its part images) through entailment. However, existing approaches do not model that each part has a different level of semantic representativeness to the whole. We propose UNcertainty-guided Compositional Hyperbolic Alignment (UNCHA) for enhancing hyperbolic VLMs. UNCHA models part-to-whole semantic representativeness with hyperbolic uncertainty, by assigning lower uncertainty to more representative parts and higher uncertainty to less representative ones for the whole scene. This representativeness is then incorporated into the contrastive objective with uncertainty-guided weights. Finally, the uncertainty is further calibrated with an entailment loss regularized by entropy-based term. With the proposed losses, UNCHA learns hyperbolic embeddings with more accurate part-whole ordering, capturing the underlying compositional structure in an image and improving its understanding of complex multi-object scenes. UNCHA achieves state-of-the-art performance on zero-shot classification, retrieval, and multi-label classification benchmarks. Our code and models are available at: https://github.com/jeeit17/UNCHA.git.

Differentiable simulators enable gradient-based optimization of soft robots over material parameters, control, and morphology, but accurately modeling real systems remains challenging due to the sim-to-real gap. This issue becomes more pronounced when geometry is itself a design variable. System identification reduces discrepancies by fitting global material parameters to data; however, when constitutive models are misspecified or observations are sparse, identified parameters often absorb geometry-dependent effects rather than reflect intrinsic material behavior. More expressive constitutive models can improve accuracy but substantially increase computational cost, limiting practicality. We propose a residual acceleration field learning (RAFL) framework that augments a base simulator with a transferable, element-level corrective dynamics field. Operating on shared local features, the model is agnostic to global mesh topology and discretization. Trained end-to-end through a differentiable simulator using sparse marker observations, the learned residual generalizes across shapes. In both sim-to-sim and sim-to-real experiments, our method achieves consistent zero-shot improvements on unseen morphologies, while system identification frequently exhibits negative transfer. The framework also supports continual refinement, enabling simulation accuracy to accumulate during morphology optimization.

To safely operate, an autonomous vehicle must know the future behavior of a potentially high number of interacting agents around it, a task often posed as multi-agent trajectory prediction. Many previous attempts to model social interactions and solve the joint prediction task either add extensive computational requirements or rely on heuristics to label multi-agent behavior types. Braid theory, in contrast, provides a powerful exact descriptor of multi-agent behavior by projecting future trajectories into braids that express how trajectories cross with each other over time; a braid then corresponds to a specific mode of coordination between the multiple agents in the future. In past work, braids have been used lightly to reason about interacting agents and restrict the attention window of predicted agents. We show that leveraging more fully the expressivity of the braid representation and using it to condition the trajectories themselves leads to even further gains in joint prediction performance, with negligible added complexity either in training or at inference time. We do so by proposing a novel auxiliary task, braid prediction, done in parallel with the trajectory prediction task. By classifying edges between agents into their correct crossing types in the braid representation, the braid prediction task is able to imbue the model with improved social awareness, which is reflected in joint predictions that more closely adhere to the actual multi-agent behavior. This simple auxiliary task allowed us to obtain significant improvements in joint metrics on three separate datasets. We show how the braid prediction task infuses the model with future intention awareness, leading to more accurate joint predictions. Code is available at github.com/caiocj1/traj-pred-braid-theory.

Bayesian neural network (BNN) posteriors are often considered impractical for inference, as symmetries fragment them, non-identifiabilities inflate dimensionality, and weight-space priors are seen as meaningless. In this work, we study how overparametrization and priors together reshape BNN posteriors and derive implications allowing us to better understand their interplay. We show that redundancy introduces three key phenomena that fundamentally reshape the posterior geometry: balancedness, weight reallocation on equal-probability manifolds, and prior conformity. We validate our findings through extensive experiments with posterior sampling budgets that far exceed those of earlier works, and demonstrate how overparametrization induces structured, prior-aligned weight posterior distributions.

Ensuring consistency between research papers and their corresponding software implementations is fundamental to software reliability and scientific reproducibility. However, this problem remains underexplored, particularly in the domain of bioinformatics, where discrepancies between methodological descriptions in papers and their actual code implementations are prevalent. To address this gap, this paper introduces a new task, namely paper-code consistency detection, and curates a collection of 48 bioinformatics software projects along with their associated publications. We systematically align sentence-level algorithmic descriptions from papers with function-level code snippets. Combined with expert annotations and a hybrid negative sampling strategy, we construct the first benchmark dataset in the bioinformatics domain tailored to this task, termed BioCon. Based on this benchmark, we further propose a cross-modal consistency detection framework designed to model the semantic relationships between natural language descriptions and code implementations. The framework adopts a unified input representation and leverages pre-trained models to capture deep semantic alignment between papers and code. To mitigate the effects of class imbalance and hard samples, we incorporate a weighted focal loss to enhance model robustness. Experimental results demonstrate that our framework effectively identifies consistency between papers and code in bioinformatics, achieving an accuracy of 0.9056 and an F1 score of 0.8011. Overall, this study opens a new research direction for paper-code consistency analysis and lays the foundation for automated reproducibility assessment and cross-modal understanding in scientific software.

This work presents AdditiveLLM2 a multi-modal, domain adapted large language model built upon the instruction tuned variant of the Gemma 3 model using a relatively small dataset of around 50 million tokens. The dataset (AdditiveLLM2-OA) consists of open-access additive manufacturing journal articles with data extracted for the domain adaptive pretraining and visual instruction tuning processes. Various stages of the developed model are evaluated with the Additive-Manufacturing-Benchmark which consists of additive manufacturing domain specific tasks compiled published resources. AdditiveLLM2 exhibits proficiency in both language and vision based tasks, achieving accuracies upwards of 90% in general additive manufacturing knowledge. This domain adaptive pretraining and instruction tuning strategy outline an accessible specialization method for large language models to a domain such as additive manufacturing.

Data-driven discovery of partial differential equations (PDEs) offers a promising paradigm for uncovering governing physical laws from observational data. However, in practical scenarios, measurements are often contaminated by noise and limited by sparse sampling, which poses significant challenges to existing approaches based on numerical differentiation or integral formulations. In this work, we propose a Symbolic Graph Network (SGN) framework for PDE discovery under noisy and sparse conditions. Instead of relying on local differential approximations, SGN leverages graph message passing to model spatial interactions, providing a non-local representation that is less sensitive to high frequency noise. Based on this representation, the learned latent features are further processed by a symbolic regression module to extract interpretable mathematical expressions. We evaluate the proposed method on several benchmark systems, including the wave equation, convection-diffusion equation, and incompressible Navier-Stokes equations. Experimental results show that SGN can recover meaningful governing relations or solution forms under varying noise levels, and demonstrates improved robustness compared to baseline methods in sparse and noisy settings. These results suggest that combining graph-based representations with symbolic regression provides a viable direction for robust data-driven discovery of physical laws from imperfect observations. The code is available at https://github.com/CXY0112/SGN

Adapters are often selected and deployed based on nominal labels (e.g., instruction-tuned), which implicitly suggest what capability improves after adaptation. We test whether nominal training objectives reliably align with realized cross-task capability gains by evaluating the same LoRA adapter across tasks. Our strongest evidence is tied to strict, automatically verifiable instruction following as measured by IFEval: across multiple seeds, base models, and LoRA settings, nominal labels recurrently but not universally fail to predict improvements on this verifiable target, with clear configuration sensitivity including a near-zero or negative case. As an illustrative strongest-case example in a controlled instruction-versus-numeric setting, an instruction-tuned adapter substantially improves off-target NM-based numeric benchmark performance from 0.133 to 0.632 while not improving verifiable instruction following on IFEval (ILA: 0.313 to 0.271; PLA: 0.250 to 0.143; values rounded to three decimals). We refer to this nominal-versus-realized mismatch pattern as capability drift as a descriptive label. The mismatch is visible in the raw cross-task performance matrix; we use a drift score only as a compact summary in the same units as the underlying metrics, not as a new formal metric contribution. Evidence from broader instruction-following benchmarks is benchmark-dependent and mixed, reflecting heterogeneity in how instruction following is operationalized; we therefore do not treat cross-benchmark agreement as a premise. Overall, the practical takeaway is to perform routine cross-task evaluation before deployment and to avoid treating nominal labels as reliable capability proxies.

Evaluating badminton performance often requires expert coaching, which is rarely accessible for amateur players. We present BadminSense, a smartwatch-based system for fine-grained badminton performance analysis using wearable sensing. Through interviews with experienced badminton players, we identified four system design requirements with three implementation insights that guide the development of BadminSense. We then collected a badminton strokes dataset on 12 experienced badminton amateurs and annotated it with fine-grained labels, including stroke type, expert-assessed stroke rating, and shuttle impact location. Built on this dataset, BadminSense segments and classifies strokes, predicts stroke quality, and estimates shuttle impact location using vibration signal from an off-the-shelf smartwatch. Our evaluations show that BadminSense achieves a stroke classification accuracy of 91.43%, an average quality rating error of 0.438, and an average impact location estimation error of 12.9%. A real-world usability study further demonstrates BadminSense's potential to provide reliable and meaningful support for daily badminton practice.

Gastrointestinal (GI) tract image analysis plays a crucial role in medical diagnosis. This research addresses the challenge of accurately classifying and segmenting GI images for real-time applications, where traditional methods often struggle due to the diversity and complexity of abnormalities. The high computational demands of this domain require efficient and adaptable solutions. This PhD thesis presents a multifaceted approach to GI image analysis. Initially, texture-based feature extraction and classification methods were explored, achieving high processing speed (over 4000 FPS) and strong performance (F1-score: 0.76, Accuracy: 0.98) on the Kvasir V2 dataset. The study then transitions to deep learning, where an optimized model combined with data bagging techniques improved performance, reaching an accuracy of 0.92 and an F1-score of 0.60 on the HyperKvasir dataset, and an F1-score of 0.88 on Kvasir V2. To support real-time detection, a streamlined neural network integrating texture and local binary patterns was developed. By addressing inter-class similarity and intra-class variation through a learned threshold, the system achieved 41 FPS with high accuracy (0.99) and an F1-score of 0.91 on HyperKvasir. Additionally, two segmentation tools are proposed to enhance usability, leveraging Depth-Wise Separable Convolution and neural network ensembles for improved detection, particularly in low-FPS scenarios. Overall, this research introduces novel and adaptable methodologies, progressing from traditional texture-based techniques to deep learning and ensemble approaches, providing a comprehensive framework for advancing GI image analysis.

Building virtual cells with generative models to simulate cellular behavior in silico is emerging as a promising paradigm for accelerating drug discovery. However, prior image-based generative approaches can produce implausible cell images that violate basic physical and biological constraints. To address this, we propose to post-train virtual cell models with reinforcement learning (RL), leveraging biologically meaningful evaluators as reward functions. We design seven rewards spanning three categories-biological function, structural validity, and morphological correctness-and optimize the state-of-the-art CellFlux model to yield CellFluxRL. CellFluxRL consistently improves over CellFlux across all rewards, with further performance boosts from test-time scaling. Overall, our results present a virtual cell modeling framework that enforces physically-based constraints through RL, advancing beyond "visually realistic" generations towards "biologically meaningful" ones.

Recent advances in AI agents for software engineering and scientific discovery have demonstrated remarkable capabilities, yet their application to developing novel ranking models in commercial search engines remains unexplored. In this paper, we present an AI Co-Scientist framework that automates the full search ranking research pipeline: from idea generation to code implementation and GPU training job scheduling with expert in the loop. Our approach strategically employs single-LLM agents for routine tasks while leveraging multi-LLM consensus agents (GPT 5.2, Gemini Pro 3, and Claude Opus 4.5) for challenging phases such as results analysis and idea generation. To our knowledge, this is the first study in the ranking community to utilize an AI Co-Scientist framework for algorithmic research. We demonstrate that this framework discovered a novel technique for handling sequence features, with all model enhancements produced automatically, yielding substantial offline performance improvements. Our findings suggest that AI systems can discover ranking architectures comparable to those developed by human experts while significantly reducing routine research workloads.

Diffusion models deliver high-fidelity generation but remain slow at inference time due to many sequential network evaluations. We find that standard timestep conditioning becomes a key bottleneck for few-step sampling. Motivated by layer-dependent denoising dynamics, we propose Multi-layer Time Embedding Optimization (MTEO), which freeze the pretrained diffusion backbone and distill a small set of step-wise, layer-wise time embeddings from reference trajectories. MTEO is plug-and-play with existing ODE solvers, adds no inference-time overhead, and trains only a tiny fraction of parameters. Extensive experiments across diverse datasets and backbones show state-of-the-art performance in the few-step sampling and substantially narrow the gap between distillation-based and lightweight methods. Code will be available.

Distribution-to-distribution generative models support scientific imaging tasks ranging from modeling cellular perturbation responses to translating medical images across conditions. Trustworthy generation requires both reliability (generalization across labs, devices, and experimental conditions) and accountability (detecting out-of-distribution cases where predictions may be unreliable). Uncertainty quantification (UQ) based approaches serve as promising candidates for these tasks, yet UQ for distribution-to-distribution generative models remains underexplored. We present a unified UQ framework, Bayesian Stochastic Flow Matching (BSFM), that disentangles aleatoric and epistemic uncertainty. The Stochastic Flow Matching (SFM) component augments deterministic flows with a diffusion term to improve model generalization to unseen scenarios. For UQ, we develop a scalable Bayesian approach -- MCD-Antithetic -- that combines Monte Carlo Dropout with sample-efficient antithetic sampling to produce effective anomaly scores for out-of-distribution detection. Experiments on cellular imaging (BBBC021, JUMP) and brain fMRI (Theory of Mind) across diverse scenarios show that SFM improves reliability while MCD-Antithetic enhances accountability.

Recent advances in multimodal learning have motivated the integration of auxiliary modalities such as text or vision into time series (TS) forecasting. However, most existing methods provide limited gains, often improving performance only in specific datasets or relying on architecture-specific designs that limit generalization. In this paper, we show that multimodal models with naive fusion strategies (e.g., simple addition or concatenation) often underperform unimodal TS models, which we attribute to the uncontrolled integration of auxiliary modalities which may introduce irrelevant information. Motivated by this observation, we explore various constrained fusion methods designed to control such integration and find that they consistently outperform naive fusion methods. Furthermore, we propose Controlled Fusion Adapter (CFA), a simple plug-in method that enables controlled cross-modal interactions without modifying the TS backbone, integrating only relevant textual information aligned with TS dynamics. CFA employs low-rank adapters to filter irrelevant textual information before fusing it into temporal representations. We conduct over 20K experiments across various datasets and TS/text models, demonstrating the effectiveness of the constrained fusion methods including CFA. Code is publicly available at: https://github.com/seunghan96/cfa/.

Deploying large language models (LLMs) on edge devices requires extremely low-bit quantization. Ultra-low precision formats such as NVFP4 offer a promising solution for reducing memory footprint and accelerating computation. However, existing quantization methods typically rely on conventional rounding strategies and fail to account for the non-uniformity of the NVFP4 numerical grid, resulting in suboptimal rounding decisions and amplified quantization errors. To address this, we propose Format-Aware Adaptive Rounding (FAAR), a learnable rounding strategy tailored for the NVFP4 format. Unlike conventional quantization paradigms, FAAR explicitly incorporates the non-uniform NVFP4 grid into the optimization process. By adaptively adjusting rounding decisions guided by loss gradients, our method effectively approximates the theoretically optimal quantization. To complement FAAR, we introduce a 2-stages Format Alignment (2FA) fine-tuning scheme that aligns LLM parameters layer-by-layer to the NVFP4 numerical space, further narrowing the performance gap. Remarkably, this learnable optimization incurs a minimal training overhead of only 4 GPU hours on Llama3-1B. Extensive experiments demonstrate the effectiveness of our approach. Compared with Round-to-Nearest (RTN), our method reduces perplexity on WikiText-2 from 14.28 to 12.60 on Llama3-1B and from 23.06 to 21.27 on Qwen3-1.7B. Additionally, our method consistently outperforms state-of-the-art approaches across various zero-shot downstream tasks.

Current language model training commonly applies multi-task Supervised Fine-Tuning (SFT) using a homogeneous compute budget across all sub-datasets. This approach is fundamentally sub-optimal: heterogeneous learning dynamics cause faster-learning tasks to overfit early while slower ones remain under-fitted. To address this, we introduce mSFT, an iterative, overfitting-aware search algorithm for multi-task data mixtures. mSFT trains the model on an active mixture, identifies and excludes the earliest overfitting sub-dataset, and reverts to that specific optimal checkpoint before continuing. Extensive evaluations demonstrate that mSFT consistently outperforms 4 baselines across 10 benchmarks and 6 base models. Further analysis confirms mSFT maintains robust gains across diverse dataset sizes, task granularities, and is insensitive to its single new hyperparameter (compute budget). Notably, at low compute budget, mSFT can improve performance while lowering training FLOPs. Ultimately, mSFT establishes a practical overfitting-aware algorithm for multi-task SFT that maximizes the potential of models across diverse data mixtures.

Accurate prediction of synthetic lethality (SL) is important for guiding the development of cancer drugs and therapies. SL prediction faces significant challenges in the effective fusion of heterogeneous multi-source data. Existing multimodal methods often suffer from "modality laziness" due to disparate convergence speeds, which hinders the exploitation of complementary information. This is also one reason why most existing SL prediction models cannot perform well on both pan-cancer and single-cancer SL pair prediction. In this study, we propose SynLeaF, a dual-stage multimodal fusion framework for SL prediction across pan- and single-cancer contexts. The framework employs a VAE-based cross-encoder with a product of experts mechanism to fuse four omics data types (gene expression, mutation, methylation, and CNV), while simultaneously utilizing a relational graph convolutional network to capture structured gene representations from biomedical knowledge graphs. To mitigate modality laziness, SynLeaF introduces a dual-stage training mechanism employing featurelevel knowledge distillation with adaptive uni-modal teacher and ensemble strategies. In extensive experiments across eight specific cancer types and a pancancer dataset, SynLeaF achieves superior performance in 17 out of 19 scenarios. Ablation studies and gradient analyses further validate the critical contributions of the proposed fusion and distillation mechanisms to model robustness and generalization. To facilitate community use, a web server is available at https://synleaf.bioinformatics-lilab.cn.

Modern clinical practice increasingly depends on reasoning over heterogeneous, evolving, and incomplete patient data. Although recent advances in multimodal foundation models have improved performance on various clinical tasks, most existing models remain static, opaque, and poorly aligned with real-world clinical workflows. We present Cerebra, an interactive multi-agent AI team that coordinates specialized agents for EHR, clinical notes, and medical imaging analysis. These outputs are synthesized into a clinician-facing dashboard that combines visual analytics with a conversational interface, enabling clinicians to interrogate predictions and contextualize risk at the point of care. Cerebra supports privacy-preserving deployment by operating on structured representations and remains robust when modalities are incomplete. We evaluated Cerebra using a massive multi-institutional dataset spanning 3 million patients from four independent healthcare systems. Cerebra consistently outperformed both state-of-the-art single-modality models and large multimodal language model baselines. In dementia risk prediction, it achieved AUROCs up to 0.80, compared with 0.74 for the strongest single-modality model and 0.68 for language model baselines. For dementia diagnosis, it achieved an AUROC of 0.86, and for survival prediction, a C-index of 0.81. In a reader study with experienced physicians, Cerebra significantly improved expert performance, increasing accuracy by 17.5 percentage points in prospective dementia risk estimation. These results demonstrate Cerebra's potential for interpretable, robust decision support in clinical care.

Visualizations help communicate data insights, but deceptive data representations can distort their interpretation and propagate misinformation. While recent Vision Language Models (VLMs) perform well on many chart understanding tasks, their ability to detect misleading visualizations, especially when deception arises from subtle reasoning errors in captions, remains poorly understood. Here, we evaluate VLMs on misleading visualization-caption pairs grounded in a fine-grained taxonomy of reasoning errors (e.g., Cherry-picking, Causal inference) and visualization design errors (e.g., Truncated axis, Dual axis, inappropriate encodings). To this end, we develop a benchmark that combines real-world visualization with human-authored, curated misleading captions designed to elicit specific reasoning and visualization error types, enabling controlled analysis across error categories and modalities of misleadingness. Evaluating many commercial and open-source VLMs, we find that models detect visual design errors substantially more reliably than reasoning-based misinformation, and frequently misclassify non-misleading visualizations as deceptive. Overall, our work fills a gap between coarse detection of misleading content and the attribution of the specific reasoning or visualization errors that give rise to it.

Large Language Models (LLMs) deployed as autonomous agents commonly use Retrieval-Augmented Generation (RAG), feeding retrieved documents into the context window, which creates two problems: the risk of hallucination grows with context length, and token cost scales linearly with dataset size. We propose the Reasoner-Executor-Synthesizer (RES) architecture, a three-layer design that strictly separates intent parsing (Reasoner), deterministic data retrieval and aggregation (Executor), and narrative generation (Synthesizer). The Executor uses zero LLM tokens and passes only fixed-size statistical summaries to the Synthesizer. We formally prove that RES achieves O(1) token complexity with respect to dataset size, and validate this on ScholarSearch, a scholarly research assistant backed by the Crossref API (130M+ articles). Across 100 benchmark runs, RES achieves a mean token cost of 1,574 tokens regardless of whether the dataset contains 42,000 or 16.3 million articles. The architecture eliminates data hallucination by construction: the LLM never sees raw records. KEYWORDS LLM agents; agentic architecture; hallucination elimination; token optimization; context window; retrieval-augmented generation; deterministic execution; scholarly metadata; Crossref API; O(1) complexity.

We propose a Quantum Federated Autoencoder for Anomaly Detection, a framework that leverages quantum federated learning for efficient, secure, and distributed processing in IoT networks. By harnessing quantum autoencoders for high-dimensional feature representation and federated learning for decentralized model training, the approach transforms localized learning on edge devices without requiring transmission of raw data, thereby preserving privacy and minimizing communication overhead. The model leverages quantum advantage in pattern recognition to enhance detection sensitivity, particularly in complex and dynamic IoT network traffic. Experiments on a real-world IoT dataset show that the proposed method delivers anomaly detection accuracy and robustness comparable to centralized approaches, while ensuring data privacy.

With the rapid growth of interconnected devices, accurately detecting malicious activities in network traffic has become increasingly challenging. Most existing deep learning-based intrusion detection systems treat network flows as independent instances, thereby failing to exploit the relational dependencies inherent in network communications. To address this limitation, we propose Q-AGNN, a Quantum-Enhanced Attentive Graph Neural Network for intrusion detection, where network flows are modeled as nodes and edges represent similarity relationships. Q-AGNN leverages parameterized quantum circuits (PQCs) to encode multi-hop neighborhood information into a high-dimensional latent space, inducing a bounded quantum feature map that implements a second-order polynomial graph filter in a quantum-induced Hilbert space. An attention mechanism is subsequently applied to adaptively weight the quantum-enhanced embeddings, allowing the model to focus on the most influential nodes contributing to anomalous behavior. Extensive experiments conducted on four benchmark intrusion detection datasets demonstrate that Q-AGNN achieves competitive or superior detection performance compared to state-of-the-art graph-based methods, while consistently maintaining low false positive rates under hardware-calibrated noise conditions. Moreover, we also executed the Q-AGNN framework on actual IBM quantum hardware to demonstrate the practical operability of the proposed pipeline under real NISQ conditions. These results highlight the effectiveness of integrating quantum-enhanced representations with attention mechanisms for graph-based intrusion detection and underscore the potential of hybrid quantum-classical learning frameworks in cybersecurity applications.

Diffusion models have achieved state-of-the-art performance in generative modeling, but their success often relies heavily on classifier-free guidance (CFG), an inference-time heuristic that modifies the sampling trajectory. From a theoretical perspective, diffusion models trained with standard denoising score matching (DSM) are expected to recover the target data distribution, raising the question of why inference-time guidance is necessary in practice. In this work, we ask whether the DSM training objective can be modified in a principled manner such that standard reverse-time sampling, without inference-time guidance, yields effects comparable to CFG. We identify insufficient inter-class separation as a key limitation of standard diffusion models. To address this, we propose MCLR, a principled alignment objective that explicitly maximizes inter-class likelihood-ratios during training. Models fine-tuned with MCLR exhibit CFG-like improvements under standard sampling, achieving comparable qualitative and quantitative gains without requiring inference-time guidance. Beyond empirical benefits, we provide a theoretical result showing that the CFG-guided score is exactly the optimal solution to a weighted MCLR objective. This establishes a formal equivalence between classifier-free guidance and alignment-based objectives, offering a mechanistic interpretation of CFG.

Designing algorithms with provable guarantees that also work well in practice remains difficult, requiring both mathematical reasoning and careful implementation. Existing approaches that bridge worst-case theory and empirical performance, such as beyond-worst-case analysis and data-driven algorithm selection, typically assume prior distributional knowledge or restrict attention to a fixed pool of algorithms. Recent progress in LLMs suggests a new possibility: provable algorithm synthesis on the fly. To study this, we built Algorithmist, an autonomous researcher agent on top of GitHub Copilot that runs a multi-agent research-and-review loop, with separate stages for idea generation, algorithm and proof development, proof-guided implementation, and review of proofs, code, and their alignment. We evaluate Algorithmist on research-level tasks in private data analysis and clustering. When asked to design practical methods that jointly satisfy privacy, approximation, and interpretability requirements, it produced provably sound and empirically effective algorithms, together with research-style writeups and audited implementations. It also found improved algorithms in some settings, explained principled barriers in others, and uncovered a subtle proof bug in prior published work. More broadly, our results suggest a new paradigm in which LLM systems generate research-paper-quality algorithmic artifacts tailored to each dataset and deployment setting. They also point to a proof-first code-synthesis paradigm, in which code is developed alongside a structured natural-language proof intermediate representation and kept aligned with it throughout synthesis.

Full-waveform inversion (FWI) estimates physical parameters in the wave equation from limited measurements and has been widely applied in geophysical exploration, medical imaging, and non-destructive testing. Conventional FWI methods are limited by their notorious sensitivity to the accuracy of the initial models. Recent progress in continuous representation FWI (CR-FWI) demonstrates that representing parameter models with a coordinate-based neural network, such as implicit neural representation (INR), can mitigate the dependence on initial models. However, its underlying mechanism remains unclear, and INR-based FWI shows slower high-frequency convergence. In this work, we investigate the general CR-FWI framework and develop a unified theoretical understanding by extending the neural tangent kernel (NTK) for FWI to establish a wave-based NTK framework. Unlike standard NTK, our analysis reveals that wave-based NTK is not constant, both at initialization and during training, due to the inherent nonlinearity of FWI. We further show that the eigenvalue decay behavior of the wave-based NTK can explain why CR-FWI alleviates the dependency on initial models and shows slower high-frequency convergence. Building on these insights, we propose several CR-FWI methods with tailored eigenvalue decay properties for FWI, including a novel hybrid representation combining INR and multi-resolution grid (termed IG-FWI) that achieves a more balanced trade-off between robustness and high-frequency convergence rate. Applications in geophysical exploration on Marmousi, 2D SEG/EAGE Salt and Overthrust, 2004 BP model, and the more realistic 2014 Chevron models show the superior performance of our proposed methods compared to conventional FWI and existing INR-based FWI methods.

The multiplication of two sparse matrices, known as SpGEMM, is a key kernel in scientific computing and large-scale data analytics, underpinning graph algorithms, machine learning, simulations, and computational biology, where sparsity is often highly unstructured. The unstructured sparsity makes achieving high performance challenging because it limits both memory efficiency and scalability. In distributed memory, the cost of exchanging and merging partial products across nodes further constrains performance. These issues are exacerbated on modern heterogeneous supercomputers with deep, hierarchical GPU interconnects. Current SpGEMM implementations overlook the gap between intra-node and inter-node bandwidth, resulting in unnecessary data movement and synchronization not fully exploiting the fast intra-node interconnect. To address these challenges, we introduce Trident, a hierarchy-aware 2D distributed SpGEMM algorithm that uses communication-avoiding techniques and asynchronous communication to exploit the hierarchical and heterogeneous architecture of modern supercomputing interconnect. Central to Trident is the novel trident partitioning scheme, which enables hierarchy-aware decomposition and reduces internode communication by leveraging the higher bandwidth between GPUs within a node compared to across nodes. Here, we evaluate Trident on unstructured matrices, achieving up to $2.38\times$ speedup over a 2D SpGEMM with a corresponding geometric mean speedup of $1.54\times$. Trident reduces internode communication volume by up to $2\times$ on NERSC's Perlmutter supercomputer. Furthermore, we demonstrate the effectiveness of Trident in speeding up Markov Clustering, achieving up to $2\times$ speedup compared to competing strategies.

Multidisciplinary Software Development (MSD) requires domain experts and developers to collaborate across incompatible formalisms and separate artifact sets. Today, even with AI coding assistants like GitHub Copilot, this process remains inefficient; individual coding tasks are semi-automated, but the workflow connecting domain knowledge to implementation is not. Developers and experts still lack a shared view, resulting in repeated coordination, clarification rounds, and error-prone handoffs. We address this gap through a graph-based workflow optimization approach that progressively replaces manual coordination with LLM-powered services, enabling incremental adoption without disrupting established practices. We evaluate our approach on \texttt{spapi}, a production in-vehicle API system at Volvo Group involving 192 endpoints, 420 properties, and 776 CAN signals across six functional domains. The automated workflow achieves 93.7\% F1 score while reducing per-API development time from approximately 5 hours to under 7 minutes, saving an estimated 979 engineering hours. In production, the system received high satisfaction from both domain experts and developers, with all participants reporting full satisfaction with communication efficiency.

Current AI agent frameworks commit early to a single interaction protocol, a fixed tool integration strategy, and static user models, limiting their deployment across diverse interaction paradigms. To address these constraints, we introduce STEM Agent (Self-adapting, Tool-enabled, Extensible, Multi-agent), a modular architecture inspired by biological pluripotency in which an undifferentiated agent core differentiates into specialized protocol handlers, tool bindings, and memory subsystems that compose into a fully functioning AI system. The framework unifies five interoperability protocols (A2A, AG-UI, A2UI, UCP, and AP2) behind a single gateway, introduces a Caller Profiler that continuously learns user preferences across more than twenty behavioral dimensions, externalizes all domain capabilities through the Model Context Protocol (MCP), and implements a biologically inspired skills acquisition system in which recurring interaction patterns crystallize into reusable agent skills through a maturation lifecycle analogous to cell differentiation. Complementing these capabilities, the memory system incorporates consolidation mechanisms, including episodic pruning, semantic deduplication, and pattern extraction, designed for sub-linear growth under sustained interaction. A comprehensive 413-test suite validates protocol handler behavior and component integration across all five architectural layers, completing in under three seconds.

We propose a new architectural change, and post-training pipeline, for making LLMs more verbose reasoners by teaching a model to truncate forward passes early. We augment an existing transformer architecture with an early-exit mechanism at intermediate layers and train the model to exit at shallower layers when the next token can be predicted without deep computation. After a calibration stage, we incentivise the model to exit as early as possible while maintaining task performance using reinforcement learning. We provide preliminary results to this effect for small reasoning models, showing that they learn to adaptively reduce computations across tokens. We predict that, applied at the right scale, our approach can minimise the amount of excess computation that reasoning models have at their disposal to perform non-myopic planning using their internal activations, reserving this only for difficult-to-predict tokens.

Time toxicity, the cumulative healthcare contact days from clinical trial participation, is an important but labor-intensive metric to extract from protocol documents. We developed TimeTox, an LLM-based pipeline for automated extraction of time toxicity from Schedule of Assessments tables. TimeTox uses Google's Gemini models in three stages: summary extraction from full-length protocol PDFs, time toxicity quantification at six cumulative timepoints for each treatment arm, and multi-run consensus via position-based arm matching. We validated against 20 synthetic schedules (240 comparisons) and assessed reproducibility on 644 real-world oncology protocols. Two architectures were compared: single-pass (vanilla) and two-stage (structure-then-count). The two-stage pipeline achieved 100% clinically acceptable accuracy ($\pm$3 days) on synthetic data (MAE 0.81 days) versus 41.5% for vanilla (MAE 9.0 days). However, on real-world protocols, the vanilla pipeline showed superior reproducibility: 95.3% clinically acceptable accuracy (IQR $\leq$ 3 days) across 3 runs on 644 protocols, with 82.0% perfect stability (IQR = 0). The production pipeline extracted time toxicity for 1,288 treatment arms across multiple disease sites. Extraction stability on real-world data, rather than accuracy on synthetic benchmarks, is the decisive factor for production LLM deployment.

Aphasia, a language impairment primarily resulting from stroke or brain injury, profoundly disrupts communication and everyday functioning. Despite advances in speech therapy, barriers such as limited therapist availability and the scarcity of personalized, culturally relevant tools continue to hinder optimal rehabilitation outcomes. This paper reviews recent developments in neurocognitive research and language technologies that contribute to the diagnosis and therapy of aphasia. Drawing on findings from our ethnographic field study, we introduce two digital therapy prototypes designed to reflect local linguistic diversity and enhance patient engagement. We also show how insights from neuroscience and the local context guided the design of these tools to better meet patient and therapist needs. Our work highlights the potential of adaptive, AI-enhanced assistive technologies to complement conventional therapy and broaden access to therapy. We conclude by outlining future research directions for advancing personalized and scalable aphasia rehabilitation.

Advanced Persistent Threats (APTs) are stealthy, multi-stage attacks that require adaptive and timely defense. While deep reinforcement learning (DRL) enables autonomous cyber defense, its decisions are often opaque and difficult to trust in operational environments. This paper presents DeepXplain, an explainable DRL framework for stage-aware APT defense. Building on our prior DeepStage model, DeepXplain integrates provenance-based graph learning, temporal stage estimation, and a unified XAI pipeline that provides structural, temporal, and policy-level explanations. Unlike post-hoc methods, explanation signals are incorporated directly into policy optimization through evidence alignment and confidence-aware reward shaping. To the best of our knowledge, DeepXplain is the first framework to integrate explanation signals into reinforcement learning for APT defense. Experiments in a realistic enterprise testbed show improvements in stage-weighted F1-score (0.887 to 0.915) and success rate (84.7% to 89.6%), along with higher explanation confidence (0.86), improved fidelity (0.79), and more compact explanations (0.31). These results demonstrate enhanced effectiveness and trustworthiness of autonomous cyber defense.

Recent progress in multimodal large language models has led to strong performance on reasoning tasks, but these improvements largely rely on high-quality annotated data or teacher-model distillation, both of which are costly and difficult to scale. To address this, we propose an unsupervised self-evolution training framework for multimodal reasoning that achieves stable performance improvements without using human-annotated answers or external reward models. For each input, we sample multiple reasoning trajectories and jointly model their within group structure. We use the Actor's self-consistency signal as a training prior, and introduce a bounded Judge based modulation to continuously reweight trajectories of different quality. We further model the modulated scores as a group level distribution and convert absolute scores into relative advantages within each group, enabling more robust policy updates. Trained with Group Relative Policy Optimization (GRPO) on unlabeled data, our method consistently improves reasoning performance and generalization on five mathematical reasoning benchmarks, offering a scalable path toward self-evolving multimodal models. The code are available at https://github.com/OPPO-Mente-Lab/LLM-Self-Judge.

Knowledge distillation has emerged as a powerful technique for compressing large language models (LLMs) into efficient, deployable architectures while preserving their advanced capabilities. Recent advances in low-rank knowledge distillation, particularly methods like Low-Rank Clone (LRC), have demonstrated remarkable empirical success, achieving comparable performance to full-parameter distillation with significantly reduced training data and computational overhead. However, the theoretical foundations underlying these methods remain poorly understood. In this paper, we establish a rigorous theoretical framework for low-rank knowledge distillation in language models. We prove that under mild assumptions, low-rank projection preserves the optimization dynamics, yielding explicit convergence rates of $O(1/\sqrt{T})$. We derive generalization bounds that characterize the fundamental trade-off between model compression and generalization capability, showing that the generalization error scales with the rank parameter as $O(r(m+n)/\sqrt{n})$. Furthermore, we provide an information-theoretic analysis of the activation cloning mechanism, revealing its role in maximizing the mutual information between the teacher's and student's intermediate representations. Our theoretical results offer principled guidelines for rank selection, mathematically suggesting an optimal rank $r^* = O(\sqrt{n})$ where $n$ is the sample size. Experimental validation on standard language modeling benchmarks confirms our theoretical predictions, demonstrating that the empirical convergence, rank scaling, and generalization behaviors align closely with our bounds.

Inferring the physical mechanisms that govern earthquake sequences from indirect geophysical observations remains difficult, particularly across tectonically distinct environments where similar seismic patterns can reflect different underlying processes. Current interpretations rely heavily on the expert synthesis of catalogs, spatiotemporal statistics, and candidate physical models, limiting reproducibility and the systematic transfer of insight across settings. Here we present TRACE (Trans-perspective Reasoning and Automated Comprehensive Evaluator), a multi-agent system that combines large language model planning with formal seismological constraints to derive auditable, physically grounded mechanistic inference from raw observations. Applied to the 2019 Ridgecrest sequence, TRACE autonomously identifies stress-perturbation-induced delayed triggering, resolving the cascading interaction between the Mw 6.4 and Mw 7.1 mainshocks; in the Santorini-Kolumbo case, the system identifies a structurally guided intrusion model, distinguishing fault-channeled episodic migration from the continuous propagation expected in homogeneous crustal failure. By providing a generalizable logical infrastructure for interpreting heterogeneous seismic phenomena, TRACE advances the field from expert-dependent analysis toward knowledge-guided autonomous discovery in Earth sciences.

Recent progress in reinforcement learning with verifiable rewards (RLVR) offers a practical path to self-improvement of language models, but existing methods face a key trade-off: endogenous self-play can drift over iterations, while corpus-grounded approaches rely on curated data environments. We present \textbf{WIST}, a \textbf{W}eb-grounded \textbf{I}terative \textbf{S}elf-play \textbf{T}ree framework for domain-targeted reasoning improvement that learns directly from the open web without requiring any pre-arranged domain corpus. WIST incrementally expands a domain tree for exploration, and retrieves and cleans path-consistent web corpus to construct a controllable training environment. It then performs Challenger--Solver self-play with verifiable rewards, and feeds learnability signals back to update node posteriors and guide subsequent exploration through an adaptive curriculum. Across four backbones, WIST consistently improves over the base models and typically outperforms both purely endogenous self-evolution and corpus-grounded self-play baselines, with the Overall gains reaching \textbf{+9.8} (\textit{Qwen3-4B-Base}) and \textbf{+9.7} (\textit{OctoThinker-8B}). WIST is also domain-steerable, improving \textit{Qwen3-8B-Base} by \textbf{+14.79} in medicine and \textit{Qwen3-4B-Base} by \textbf{+5.28} on PhyBench. Ablations further confirm the importance of WIST's key components for stable open-web learning. Our Code is available at https://github.com/lfy-123/WIST.

Deterministic pre-execution safety gates evaluate whether individual agent actions are compatible with their assigned roles. While effective at per-action authorization, these systems are structurally blind to distributed attacks that decompose harmful intent across multiple individually-compliant steps. This paper introduces Session Risk Memory (SRM), a lightweight deterministic module that extends stateless execution gates with trajectory-level authorization. SRM maintains a compact semantic centroid representing the evolving behavioral profile of an agent session and accumulates a risk signal through exponential moving average over baseline-subtracted gate outputs. It operates on the same semantic vector representation as the underlying gate, requiring no additional model components, training, or probabilistic inference. We evaluate SRM on a multi-turn benchmark of 80 sessions containing slow-burn exfiltration, gradual privilege escalation, and compliance drift scenarios. Results show that ILION+SRM achieves F1 = 1.0000 with 0% false positive rate, compared to stateless ILION at F1 = 0.9756 with 5% FPR, while maintaining 100% detection rate for both systems. Critically, SRM eliminates all false positives with a per-turn overhead under 250 microseconds. The framework introduces a conceptual distinction between spatial authorization consistency (evaluated per action) and temporal authorization consistency (evaluated over trajectory), providing a principled basis for session-level safety in agentic systems.

Online learning algorithms often faces a fundamental trilemma: balancing regret guarantees between adversarial and stochastic settings and providing baseline safety against a fixed comparator. While existing methods excel in one or two of these regimes, they typically fail to unify all three without sacrificing optimal rates or requiring oracle access to problem-dependent parameters. In this work, we bridge this gap by introducing COMPASS-Hedge. Our algorithm is the first full-information method to simultaneously achieve: i) Minimax-optimal regret in adversarial environments; ii) Instance-optimal, gap-dependent regret in stochastic environments; and iii) $\tilde{\mathcal{O}}(1)$ regret relative to a designated baseline policy, up to logarithmic factors. Crucially, COMPASS-Hedge is parameter-free and requires no prior knowledge of the environment's nature or the magnitude of the stochastic sub optimality gaps. Our approach hinges on a novel integration of adaptive pseudo-regret scaling and phase-based aggression, coupled with a comparator-aware mixing strategy. To the best of our knowledge, this provides the first "best-of-three-world" guarantee in the full-information setting, establishing that baseline safety does not have to come at the cost of worst-case robustness or stochastic efficiency.

Diffusion-based image super-resolution (SR), which aims to reconstruct high-resolution (HR) images from corresponding low-resolution (LR) observations, faces a fundamental trade-off between inference efficiency and reconstruction quality. The state-of-the-art residual-shifting diffusion framework achieves efficient 4-step inference, yet suffers from severe performance degradation in compact sampling trajectories. This is mainly attributed to two core limitations: the inherent suboptimality of unconstrained random Gaussian noise in intermediate steps, which leads to error accumulation and insufficient LR prior guidance, and the initialization bias caused by naive bicubic upsampling. In this paper, we propose LPNSR, a prior-enhanced efficient diffusion framework to address these issues. We first mathematically derive the closed-form analytical solution of the optimal intermediate noise for the residual-shifting diffusion paradigm, and accordingly design an LR-guided multi-input-aware noise predictor to replace random Gaussian noise, embedding LR structural priors into the reverse process while fully preserving the framework's core efficient residual-shifting mechanism. We further mitigate initial bias with a high-quality pre-upsampling network to optimize the diffusion starting point. With a compact 4-step trajectory, LPNSR can be optimized in an end-to-end manner. Extensive experiments demonstrate that LPNSR achieves state-of-the-art perceptual performance on both synthetic and real-world datasets, without relying on any large-scale text-to-image priors. The source code of our method can be found at https://github.com/Faze-Hsw/LPNSR.

While Landauer's principle establishes the fundamental thermodynamic floor for information erasure and Fisher Information provides a metric for local curvature in parameter space, these classical frameworks function effectively only as approximations within regimes of sparse rule-constraints. They fail to explain the super-linear, and often explosive, computational and energy costs incurred when maintaining symbolic interpretability during the reconfiguration of advanced intelligent systems. This paper introduces the property of intelligence inertia and its underlying physical principles as foundational characteristics for quantifying the computational weight of intelligence. We demonstrate that this phenomenon is not merely an empirical observation but originates from the fundamental non-commutativity between rules and states, a root cause we have formally organized into a rigorous mathematical framework. By analyzing the growing discrepancy between actual adaptation costs and static information-theoretic estimates, we derive a non-linear cost formula that mirrors the Lorentz factor, characterizing a relativistic J-shaped inflation curve -- a "computational wall" that static models are blind to. The validity of these physical principles is examined through a trilogy of decisive experiments: (1) a comparative adjudication of this J-curve inflation against classical Fisher Information models, (2) a geometric analysis of the "Zig-Zag" trajectory of neural architecture evolution, and (3) the implementation of an inertia-aware scheduler wrapper that optimizes the training of deep networks by respecting the agent's physical resistance to change. Our results suggest a unified physical description for the cost of structural adaptation, offering a first-principle explanation for the computational and interpretability-maintenance overhead in intelligent agents.

We isolate and empirically characterize first-mover bias -- a path-dependent concentration of feature importance caused by sequential residual fitting in gradient boosting -- as a specific mechanistic cause of the well-known instability of SHAP-based feature rankings under multicollinearity. When correlated features compete for early splits, gradient boosting creates a self-reinforcing advantage for whichever feature is selected first: subsequent trees inherit modified residuals that favor the incumbent, concentrating SHAP importance on an arbitrary feature rather than distributing it across the correlated group. Scaling up a single model amplifies this effect -- a Large Single Model with the same total tree count as our method produces the worst explanations of any approach tested. We demonstrate that model independence is sufficient to resolve first-mover bias in the linear regime, and remains the most effective mitigation under nonlinear data-generating processes. Both our proposed method, DASH (Diversified Aggregation of SHAP), and simple seed-averaging (Stochastic Retrain) restore stability by breaking the sequential dependency chain, confirming that the operative mechanism is independence between explained models. At rho=0.9, both achieve stability=0.977, while the single-best workflow degrades to 0.958 and the Large Single Model to 0.938. On the Breast Cancer dataset, DASH improves stability from 0.32 to 0.93 (+0.61) against a tree-count-matched baseline. DASH additionally provides two diagnostic tools -- the Feature Stability Index (FSI) and Importance-Stability (IS) Plot -- that detect first-mover bias without ground truth, enabling practitioners to audit explanation reliability before acting on feature rankings. Software and reproducible benchmarks are available at https://github.com/DrakeCaraker/dash-shap.

Multimodal emotion recognition in conversations (MERC) aims to identify and understand the emotions expressed by speakers during utterance interaction from multiple modalities (e.g., text, audio, images, etc.). Existing studies have shown that GCN can improve the performance of MERC by modeling dependencies between speakers. However, existing methods usually use fixed parameters to process multimodal features for different emotion types, ignoring the dynamics of fusion between different modalities, which forces the model to balance performance between multiple emotion categories, thus limiting the model's performance on some specific emotions. To this end, we propose a dynamic fusion-aware graph convolutional neural network (DF-GCN) for robust recognition of multimodal emotion features in conversations. Specifically, DF-GCN integrates ordinary differential equations into graph convolutional networks (GCNs) to {capture} the dynamic nature of emotional dependencies within utterance interaction networks and leverages the prompts generated by the global information vector (GIV) of the utterance to guide the dynamic fusion of multimodal features. This allows our model to dynamically change parameters when processing each utterance feature, so that different network parameters can be equipped for different emotion categories in the inference stage, thereby achieving more flexible emotion classification and enhancing the generalization ability of the model. Comprehensive experiments conducted on two public multimodal conversational datasets {confirm} that the proposed DF-GCN model delivers superior performance, benefiting significantly from the dynamic fusion mechanism introduced.

Large language models (LLMs) assisted literature retrieval may lead to erroneous references, but these errors have not been rigorously quantified. Therefore, we quantitatively assess errors in reference retrieval of widely used free-version LLM platforms and identify the factors associated with retrieval errors. We evaluated 2,000 references retrieved by 5 LLMs (Grok-2, ChatGPT GPT-4.1, Google Gemini Flash 2.5, Perplexity AI, and DeepSeek GPT-4) for 40 randomly-selected original articles (10 per journal) published Jan. 2024 to July 2025 from British Medical Journal (BMJ), Journal of the American Medical Association, and The New England Journal of Medicine (NEJM). Primary outcomes were a multimetric score ratio combining validity of digital object identifier, PubMed ID, Google-Scholar link, and relevance; and complete miss rate (proportion of references failing all applicable metrics). Multivariable regression was used to examine independent associations. LLM platforms completely failed to retrieve correct reference data 47.8% of the time. The average score ratio of the 5 LLM platforms was 0.29 (standard deviation, 0.35; range, 0-1.25), with a higher score ratio indicating a higher accuracy in retrieving relevant references and correct bibliographic data. The highest and lowest accuracies were achieved by Grok (0.57) and Genimi (0.11), respectively. Compared with BMJ, NEJM articles had lower score ratios and higher complete miss rates. Multivariable analysis shows LLM platforms and journals were independently associated with score ratios and complete miss rate, respectively. We show modest overall performance of LLMs and significant variability in retrieval accuracy across platforms and journals. LLM platforms and journals are associated with LLM's performance in retrieving medical literature. Bibliographic data should be carefully reviewed when using LLM-assisted literature retrieval.

Photovoltaic (PV) power forecasting in edge-enabled grids requires balancing forecasting accuracy, robustness under weather-driven distribution shifts, and strict latency constraints. Local specialized models are efficient for routine conditions but often degrade under rare ramp events and unseen weather patterns, whereas always relying on cloud-side large models incurs substantial communication delay and cloud overhead. To address this challenge, we propose a risk-aware cloud-edge collaborative framework for latency-sensitive PV forecasting. The framework integrates a site-specific expert predictor for routine cases, a lightweight edge-side model for enhanced local inference, and a cloud-side large retrieval model that provides matched historical context when needed through a retrieval-prediction pipeline. A lightweight screening module estimates predictive uncertainty, out-of-distribution risk, weather mutation intensity, and model disagreement, while a Lyapunov-guided router selectively escalates inference to the edge-small or cloud-assisted branches under long-term latency, communication, and cloud-usage constraints. The outputs of the activated branches are combined through adaptive fusion. Experiments on two real-world PV datasets demonstrate a favorable overall trade-off among forecasting accuracy, routing quality, robustness, and system efficiency.

Neutrino oscillations encode fundamental information about neutrino masses and mixing parameters, offering a unique window into physics beyond the Standard Model. Estimating these parameters from oscillation probability maps is, however, computationally challenging due to the maps' high dimensionality and nonlinear dependence on the underlying physics. Traditional inference methods, such as likelihood-based or Monte Carlo sampling approaches, require extensive simulations to explore the parameter space, creating major bottlenecks for large-scale analyses. In this work, we introduce a data-driven framework that reformulates atmospheric neutrino oscillation parameter inference as a supervised regression task over structured oscillation maps. We propose a hierarchical transformer architecture that explicitly models the two-dimensional structure of these maps, capturing angular dependencies at fixed energies and global correlations across the energy spectrum. To improve physical consistency, the model is trained using a surrogate simulation constraint that enforces agreement between the predicted parameters and the reconstructed oscillation patterns. Furthermore, we introduce a neural network-based uncertainty quantification mechanism that produces distribution-free prediction intervals with formal coverage guarantees. Experiments on simulated oscillation maps under Earth-matter conditions demonstrate that the proposed method is comparable to a Markov Chain Monte Carlo baseline in estimation accuracy, with substantial improvements in computational cost (around 240$\times$ fewer FLOPs and 33$\times$ faster in average processing time). Moreover, the conformally calibrated prediction intervals remain narrow while achieving the target nominal coverage of 90%, confirming both the reliability and efficiency of our approach.

While prior red-teaming efforts have focused on eliciting harmful text outputs from large language models (LLMs), such approaches fail to capture agent-specific vulnerabilities that emerge through multi-step tool execution, particularly in rapidly growing ecosystems such as the Model Context Protocol (MCP). To address this gap, we propose a trajectory-aware evolutionary search method, T-MAP, which leverages execution trajectories to guide the discovery of adversarial prompts. Our approach enables the automatic generation of attacks that not only bypass safety guardrails but also reliably realize harmful objectives through actual tool interactions. Empirical evaluations across diverse MCP environments demonstrate that T-MAP substantially outperforms baselines in attack realization rate (ARR) and remains effective against frontier models, including GPT-5.2, Gemini-3-Pro, Qwen3.5, and GLM-5, thereby revealing previously underexplored vulnerabilities in autonomous LLM agents.

Recent advances in Retrieval-Augmented Generation (RAG) have revolutionized knowledge-intensive tasks, yet traditional RAG methods struggle when the search space is unknown or when documents are semi-structured or structured. We introduce a novel end-to-end Graph RAG framework that leverages both Labeled Property Graph (LPG) and Resource Description Framework (RDF) architectures to overcome these limitations. Our approach enables dynamic document retrieval without the need to pre-specify the number of documents and eliminates inefficient reranking. We propose an innovative method for converting documents into RDF triplets using JSON key-value pairs, facilitating seamless integration of semi-structured data. Additionally, we present a text to Cypher framework for LPG, achieving over 90% accuracy in real-time translation of text queries to Cypher, enabling fast and reliable query generation suitable for online applications. Our empirical evaluation demonstrates that Graph RAG significantly outperforms traditional embedding-based RAG in accuracy, response quality, and reasoning, especially for complex, semi-structured tasks. These findings establish Graph RAG as a transformative solution for next-generation retrieval-augmented systems.

Chinchilla Approach 2 is among the most widely used methods for fitting neural scaling laws. Its parabolic approximation introduces systematic biases in compute-optimal allocation estimates, even on noise-free synthetic data. Applied to published Llama 3 IsoFLOP data at open frontier compute scales, these biases imply a parameter underallocation corresponding to 6.5% of the $3.8\times10^{25}$ FLOP training budget and \$1.4M (90% CI: \$412K-\$2.9M) in unnecessary compute at 50% H100 MFU. Simulated multimodal model misallocations show even greater opportunity costs due to higher loss surface asymmetry. Three sources of this error are examined: IsoFLOP sampling grid width (Taylor approximation accuracy), uncentered IsoFLOP sampling, and loss surface asymmetry ($α\neq β$). Chinchilla Approach 3 largely eliminates these biases but is often regarded as less data-efficient, numerically unstable, prone to local minima, and harder to implement. Each concern is shown to be unfounded or addressable, especially when the partially linear structure of the objective is exploited via Variable Projection, enabling unbiased inference on all five loss surface parameters through a two-dimensional optimization that is well-conditioned, analytically differentiable, and amenable to dense, or even exhaustive, grid search. It may serve as a more convenient replacement for Approach 2 or a more scalable alternative for adaptations of Approach 3 to richer scaling law formulations.

The rapid growth in the volume, variety, and velocity of geospatial data has created data ecosystems that are highly distributed, heterogeneous, and semantically inconsistent. Existing data catalogs, portals, and infrastructures still rely largely on keyword-based search with limited semantic support, which often fails to capture user intent and leads to weak retrieval performance. To address these challenges, this study proposes a knowledge graph-driven multi-agent framework for intelligent geospatial data discovery, powered by large language models. The framework introduces a unified geospatial metadata ontology as a semantic mediation layer to align heterogeneous metadata standards across platforms and constructs a geospatial metadata knowledge graph to explicitly model datasets and their multidimensional relationships. Building on the structured representation, the framework adopts a multi-agent collaborative architecture to perform intent parsing, knowledge graph retrieval, and answer synthesis, forming an interpretable and closed-loop discovery process from user queries to results. Results from representative use cases and performance evaluation show that the framework substantially improves intent matching accuracy, ranking quality, recall, and discovery transparency compared with traditional systems. This study advances geospatial data discovery toward a more semantic, intent-aware, and intelligent paradigm, providing a practical foundation for next-generation intelligent and autonomous spatial data infrastructures and contributing to the broader vision of Autonomous GIS.

Large language model (LLM) services are mostly centralized, leading to scalability bottlenecks and underutilization of substantial scattered GPU resources. While decentralization offers a promising alternative, existing frameworks primarily focus on cooperation among GPU providers while overlooking their inherent competitive dynamics, imposing substantial constraints such as excessive platform-level oversight or rigid requirements to execute all assigned requests using fixed software stacks on fixed hardware configurations. We argue that such assumptions are unrealistic in real-world decentralized environments. To this end, we propose WWW$.$Serve, a decentralized framework for interconnecting LLM services worldwide. It allows participants to flexibly determine their participation policies and resource commitments, and supports self-organizing request dispatch, enabling the network to autonomously allocate requests without centralized coordination. Empirically, we show that WWW$.$Serve improves global SLO (service-level-objective) attainment by up to 1.5x and lowers latency by 27.6%. Its performance approaches, and in some cases surpasses, centralized scheduling, while fully preserving the benefits of decentralization. These results highlight WWW$.$Serve as a promising foundation for real-world, decentralized LLM serving.

We introduce Exponential Family Discriminant Analysis (EFDA), a unified generative framework that extends classical Linear Discriminant Analysis (LDA) beyond the Gaussian setting to any member of the exponential family. Under the assumption that each class-conditional density belongs to a common exponential family, EFDA derives closed-form maximum-likelihood estimators for all natural parameters and yields a decision rule that is linear in the sufficient statistic, recovering LDA as a special case and capturing nonlinear decision boundaries in the original feature space. We prove that EFDA is asymptotically calibrated and statistically efficient under correct specification, and we generalise it to $K \geq 2$ classes and multivariate data. Through extensive simulation across five exponential-family distributions (Weibull, Gamma, Exponential, Poisson, Negative Binomial), EFDA matches the classification accuracy of LDA, QDA, and logistic regression while reducing Expected Calibration Error (ECE) by $2$-$6\times$, a gap that is structural: it persists for all $n$ and across all class-imbalance levels, because misspecified models remain asymptotically miscalibrated. We further prove and empirically confirm that EFDA's log-odds estimator approaches the Cramér-Rao bound under correct specification, and is the only estimator in our comparison whose mean squared error converges to zero. Complete derivations are provided for nine distributions. Finally, we formally verify all four theoretical propositions in Lean 4, using Aristotle (Harmonic) and OpenGauss (Math, Inc.) as proof generators, with all outputs independently machine-checked by AXLE (Axiom).

Direct Preference Optimization (DPO) guides large language models (LLMs) to generate recommendations aligned with user historical behavior distributions by minimizing preference alignment loss. However, our systematic empirical research and theoretical analysis reveal that DPO tends to amplify spurious correlations caused by environmental confounders during the alignment process, significantly undermining the generalization capability of LLM-based generative recommendation methods in out of distribution (OOD) scenarios. To mitigate this issue, we propose CausalDPO, an extension of DPO that incorporates a causal invariance learning mechanism. This method introduces a backdoor adjustment strategy during the preference alignment phase to eliminate interference from environmental confounders, explicitly models the latent environmental distribution using a soft clustering approach, and enhances robust consistency across diverse environments through invariance constraints. Theoretical analysis demonstrates that CausalDPO can effectively capture users stable preference structures across multiple environments, thereby improving the OOD generalization performance of LLM-based recommendation models. We conduct extensive experiments under four representative distribution shift settings to validate the effectiveness of CausalDPO, achieving an average performance improvement of 17.17% across four evaluation metrics.

As long-context language modeling becomes increasingly important, the cost of maintaining and attending to large Key/Value (KV) caches grows rapidly, becoming a major bottleneck in both training and inference. While prior works such as Multi-Query Attention (MQA) and Multi-Latent Attention (MLA) reduce memory by sharing or compressing KV features, they often trade off representation quality or incur runtime overhead. We propose Memory-Keyed Attention (MKA), a hierarchical attention mechanism that integrates multi-level KV caches (local, session, and long-term) and learns to route attention across them dynamically. We further introduce Route-Fused MKA (FastMKA), a broadcast-routed variant that fuses memory sources before attention computation for improved efficiency. Experiments on different sequence lengths show that FastMKA achieves a favorable accuracy-efficiency trade-off: comparable perplexity to MLA while achieving up to 5x faster training throughput and 1.8x lower evaluation latency. These results highlight MKA as a practical and extensible framework for efficient long-context attention.

State-space models (SSMs) offer efficient alternatives to attention with linear-time recurrence. Mamba2, a recent SSM-based language model, uses selective input gating and a multi-head structure, enabling parallel computation and strong benchmark performance. However, its multi-head recurrence operates independently without structured utilization or analysis. In this work, we propose a novel method called Hierarchical ADaptive filter bank for Efficient SSMs (HADES), a Graph Signal Processing (GSP)-inspired framework that reinterprets Mamba2 as an adaptive filter bank on a line graph. Our hierarchical architecture introduces two filter types: shared filters for global low-pass behavior and expert filters for local high-pass behavior, achieved through structured bias on the parameter Δ. HADES achieves comparable performance to baseline models including Mamba2 across various benchmarks in language modeling, commonsense reasoning, and long-context retrieval, while using only 58.9% of the original parameters. In this regard, HADES bridges GSP and neural sequence modeling, enabling efficient, hierarchical, and interpretable filtering within state-space models.

Data imputation is a cornerstone technique for handling missing values in real-world datasets, which are often plagued by missingness. Despite recent progress, prior studies on Large Language Models-based imputation remain limited by scalability challenges, restricted cross-model comparisons, and evaluations conducted on small or domain-specific datasets. Furthermore, heterogeneous experimental protocols and inconsistent treatment of missingness mechanisms (MCAR, MAR, and MNAR) hinder systematic benchmarking across methods. This work investigates the robustness of Large Language Models for missing data imputation in tabular datasets using a zero-shot prompt engineering approach. To this end, we present a comprehensive benchmarking study comparing five widely used LLMs against six state-of-the-art imputation baselines. The experimental design evaluates these methods across 29 datasets (including nine synthetic datasets) under MCAR, MAR, and MNAR mechanisms, with missing rates of up to 20\%. The results demonstrate that leading LLMs, particularly Gemini 3.0 Flash and Claude 4.5 Sonnet, consistently achieve superior performance on real-world open-source datasets compared to traditional methods. However, this advantage appears to be closely tied to the models' prior exposure to domain-specific patterns learned during pre-training on internet-scale corpora. In contrast, on synthetic datasets, traditional methods such as MICE outperform LLMs, suggesting that LLM effectiveness is driven by semantic context rather than purely statistical reconstruction. Furthermore, we identify a clear trade-off: while LLMs excel in imputation quality, they incur significantly higher computational time and monetary costs. Overall, this study provides a large-scale comparative analysis, positioning LLMs as promising semantics-driven imputers for complex tabular data.

Requirements engineering in Industry 4.0 faces critical challenges with heterogeneous, unstructured documentation spanning technical specifications, supplier lists, and compliance standards. While retrieval-augmented generation (RAG) shows promise for knowledge-intensive tasks, no prior work has evaluated RAG on authentic industrial RE workflows using comprehensive production-grade performance metrics. This paper presents a comprehensive empirical evaluation of RAG for industrial requirements engineering automation using authentic automotive manufacturing documentation comprising 669 requirements across four specification standards (MBN 9666-1, MBN 9666-2, BQF 9666-5, MBN 9666-9) spanning 2015-2023, plus 49 supplier qualifications with extensive supporting documentation. Through controlled comparisons with BERT-based and ungrounded LLM approaches, the framework achieves 98.2% extraction accuracy with complete traceability, outperforming baselines by 24.4% and 19.6%, respectively. Hybrid semantic-lexical retrieval achieves MRR of 0.847. Expert quality assessment averaged 4.32/5.0 across five dimensions. The evaluation demonstrates 83% reduction in manual analysis time and 47% cost savings through multi-provider LLM orchestration. Ablation studies quantify individual component contributions. Longitudinal analysis reveals a 55% reduction in requirement volume coupled with 1,800% increase in IT security focus, identifying 10 legacy suppliers (20.4%) requiring requalification, representing potential $2.3M in avoided contract penalties.

Every wildfire prediction model deployed today shares a dangerous property: none of these methods provides formal guarantees on how much fire spread is missed. Despite extensive work on wildfire spread prediction using deep learning, no prior study has applied distribution-free safety guarantees to this domain, leaving evacuation planners reliant on probability thresholds with no formal assurance. We address this gap by presenting, to our knowledge, the first application of conformal risk control (CRC) to wildfire spread prediction, providing finite-sample guarantees on false negative rate (FNR <= 0.05). We expose a stark failure: across three model families of increasing complexity (tabular: LightGBM, AUROC 0.854; convolutional: Tiny U-Net, AUROC 0.969; and graph-based: Hybrid ResGNN-UNet, AUROC 0.964), standard thresholds capture only 7-72% of true fire spread. CRC eliminates this failure uniformly. Our central finding is that model architecture determines evacuation efficiency, while CRC determines safety: both spatial models with CRC achieve approximately 95% fire coverage while flagging only approximately 15% of total pixels, making them 4.2x more efficient than LightGBM, while the graph model's additional complexity over a simple U-Net yields no meaningful efficiency gain. We propose a shift-aware three-way CRC framework that assigns SAFE/MONITOR/EVACUATE zones for operational triage, and characterize a fundamental limitation of prevalence-weighted bounds under extreme class imbalance (approximately 5% fire prevalence). All models, calibration code, and evaluation pipelines are released for reproducibility.

Accurate prediction of RNA secondary structure underpins transcriptome annotation, mechanistic analysis of non-coding RNAs, and RNA therapeutic design. Recent gains from deep learning and RNA foundation models are difficult to interpret because current benchmarks may overestimate generalization across RNA families. We present the Comprehensive Hierarchical Annotation of Non-coding RNA Groups (CHANRG), a benchmark of 170{,}083 structurally non-redundant RNAs curated from more than 10 million sequences in Rfam~15.0 using structure-aware deduplication, genome-aware split design and multiscale structural evaluation. Across 29 predictors, foundation-model methods achieved the highest held-out accuracy but lost most of that advantage out of distribution, whereas structured decoders and direct neural predictors remained markedly more robust. This gap persisted after controlling for sequence length and reflected both loss of structural coverage and incorrect higher-order wiring. Together, CHANRG and a padding-free, symmetry-aware evaluation stack provide a stricter and batch-invariant framework for developing RNA structure predictors with demonstrable out-of-distribution robustness.

Decoder-only language models are stateless: hidden representations are discarded after every forward pass and nothing persists across sessions. Jeong (2026a) showed that trained memory adapters give a frozen encoder-decoder backbone persistent latent-space memory, building on the lateral-memory framework of Jeong (2026b,c). Here we ask whether the same principle transfers to the decoder-only setting, where no cross-attention pathway exists and memory must enter through self-attention alone. We adapt six methods -- prefix, parallel cross-attention, KV extension, Hebbian memory, context-gated branch, and slot-based sparse write -- to a frozen GPT-2, training only a small adapter $θ_{mem}$. The write rule is shared; only the read injection changes from decoder cross-attention to self-attention KV prefix or parallel branch. On LoCoMo we find a striking inductive-bias dichotomy: at $1\times$ capacity, three methods with strong architectural priors -- cross-attention (M.2), Hebbian (M.4), and slot write (M.6) -- achieve retained-memory scores of $7-18\%$ and knowledge gains $ΔK$ of $7-10$, while the other three fail ($< 0.4\%$). At $10\times$ capacity all six converge, showing the gap is architectural, not fundamental. Together with the encoder-decoder results of Jeong (2026a) and the brain-inspired modules of Jeong (2026b,c), these findings establish persistent latent-space memory as a general paradigm spanning major transformer families.

Despite the strong predictive performance achieved by machine learning models across many application domains, assessing their trustworthiness through reliable estimates of predictive confidence remains a critical challenge. This issue arises in scenarios where the likelihood of error inferred from learned representations follows a bimodal distribution, resulting from the coexistence of confident and ambiguous predictions. Standard regression approaches often struggle to adequately express this predictive uncertainty, as they implicitly assume unimodal Gaussian noise, leading to mean-collapse behavior in such settings. Although Mixture Density Networks (MDNs) can represent different distributions, they suffer from severe optimization instability. We propose a family of distribution-aware loss functions integrating normalized RMSE with Wasserstein and Cramér distances. When applied to standard deep regression models, our approach recovers bimodal distributions without the volatility of mixture models. Validated across four experimental stages, our results show that the proposed Wasserstein loss establishes a new Pareto efficiency frontier: matching the stability of standard regression losses like MSE in unimodal tasks while reducing Jensen-Shannon Divergence by 45% on complex bimodal datasets. Our framework strictly dominates MDNs in both fidelity and robustness, offering a reliable tool for aleatoric uncertainty estimation in trustworthy AI systems.

Recent work on chain-of-thought (CoT) faithfulness reports single aggregate numbers (e.g., DeepSeek-R1 acknowledges hints 39% of the time), implying that faithfulness is an objective, measurable property of a model. This paper provides evidence that it is not. Three classifiers (a regex-only detector, a regex-plus-LLM pipeline, and a Claude Sonnet 4 judge) are applied to 10,276 influenced reasoning traces from 12 open-weight models spanning 9 families and 7B to 1T parameters. On identical data, these classifiers produce faithfulness rates of 74.4%, 82.6%, and 69.7%. Per-model gaps range from 2.6 to 30.6 percentage points; all pairwise McNemar tests are significant (p < 0.001). The disagreements are systematic: Cohen's kappa ranges from 0.06 ("slight") for sycophancy hints to 0.42 ("moderate") for grader hints, and the asymmetry is pronounced: for sycophancy, 883 cases are classified as faithful by the pipeline but unfaithful by the Sonnet judge, while only 2 go the other direction. Classifier choice can also reverse model rankings: Qwen3.5-27B ranks 1st under the pipeline but 7th under Sonnet; OLMo-3.1-32B moves from 9th to 3rd. Different classifiers operationalize faithfulness at different levels of stringency (lexical mention versus epistemic dependence), yielding divergent measurements on the same behavior. These results indicate that published faithfulness numbers cannot be meaningfully compared across studies using different classifiers, and that future evaluations should report sensitivity ranges across multiple classification methodologies.

Systematic literature reviews are essential for synthesizing scientific evidence but are costly, difficult to scale and time-intensive, creating bottlenecks for evidence-based policy. We study whether large language models can automate the complete systematic review workflow, from article retrieval, article screening, data extraction to report synthesis. Applied to epidemiological reviews of nine WHO-designated priority pathogens and validated against expert-curated ground truth, our open-source agentic pipeline (AgentSLR) achieves performance comparable to human researchers while reducing review time from approximately 7 weeks to 20 hours (a 58x speed-up). Our comparison of five frontier models reveals that performance on SLR is driven less by model size or inference cost than by each model's distinctive capabilities. Through human-in-the-loop validation, we identify key failure modes. Our results demonstrate that agentic AI can substantially accelerate scientific evidence synthesis in specialised domains.

Forward-backward (FB) representations provide a powerful framework for learning the successor representation (SR) in continuous spaces by enforcing a low-rank factorization. However, a fundamental spectral mismatch often exists between the high-rank transition dynamics of continuous environments and the low-rank bottleneck of the FB architecture, making accurate low-rank representation learning difficult. In this work, we analyze temporal abstraction as a mechanism to mitigate this mismatch. By characterizing the spectral properties of the transition operator, we show that temporal abstraction acts as a low-pass filter that suppresses high-frequency spectral components. This suppression reduces the effective rank of the induced SR while preserving a formal bound on the resulting value function error. Empirically, we show that this alignment is a key factor for stable FB learning, particularly at high discount factors where bootstrapping becomes error-prone. Our results identify temporal abstraction as a principled mechanism for shaping the spectral structure of the underlying MDP and enabling effective long-horizon representations in continuous control.

Decision support systems are essential for maintaining grid stability in low-carbon power systems, such as wind power plants, by providing real-time alerts to control room operators regarding potential events, including Wind Power Ramp Events (WPREs). These early warnings enable the timely initiation of more detailed system stability assessments and preventive actions. However, forecasting these events is challenging due to the inherent class imbalance in WPRE datasets, where ramp events are less frequent (typically less than 15\% of observed events) compared to normal conditions. Ignoring this characteristic undermines the performance of conventional machine learning models, which often favor the majority class. This paper introduces a novel methodology for WPRE forecasting as a multivariate time series classification task and proposes a data preprocessing strategy that extracts features from recent power observations and masks unavailable ramp information, making it integrable with traditional real-time ramp identification tools. Particularly, the proposed methodology combines majority-class undersampling and ensemble learning to enhance wind ramp event forecasting under class imbalance. Numerical simulations conducted on a real-world dataset demonstrate the superiority of our approach, achieving over 85% accuracy and 88% weighted F1 score, outperforming benchmark classifiers.

Recurrent neural networks (RNNs) and self-attention are both widely used sequence-mixing layers that maintain an internal memory. However, this memory is constructed using two orthogonal mechanisms: RNNs compress the entire past into a fixed-size state, whereas self-attention's state stores every past time step growing its state (the KV cache) linearly with the sequence length. This results in orthogonal strengths and weaknesses. Self-attention layers excel at retrieving information in the context but have large memory and computational costs, while RNNs are more efficient but degrade over longer contexts and underperform for precise recall tasks. Prior work combining these mechanisms has focused primarily on naively interleaving them to reduce computational cost without regard to their complementary mechanisms. We propose the Hybrid Associative Memory (HAM) layer, which combines self-attention and RNNs while leveraging their individual strengths: the RNN compresses the entire sequence, while attention supplements it *only* with information that is difficult for the RNN to predict, which is hence the most valuable information to explicitly store. HAM layers enable data-dependent growth of the KV cache, which can be precisely controlled by the user with a single, continuous threshold. We find that this fine-grained control of the KV cache growth rate has a smooth trade-off with loss and performance. Empirically, we show that our hybrid architecture offers strong, competitive performance relative to RNNs and Transformers even at substantially lower KV-cache usage.

Configuration tuning for better performance is crucial in quality assurance. Yet, there has long been a mystery on tuners' effectiveness, due to the black-box nature of configurable systems. Prior efforts predominantly adopt static domain analysis (e.g., static taint analysis), which often lacks generalizability, or dynamic data analysis (e.g., benchmarking performance analysis), limiting explainability. In this work, we embrace Fitness Landscape Analysis (FLA) as a bridge between domain knowledge and difficulty of the tuning. We propose Domland, a two-pronged methodology that synergizes the spatial information obtained from FLA and domain-driven analysis to systematically capture the hidden characteristics of configuration tuning cases, explaining how and why a tuner might succeed or fail. This helps to better interpret and contextualize the behavior of tuners and inform tuner design. To evaluate Domland, we conduct a case study of nine software systems and 93 workloads, from which we reveal several key findings: (1) configuration landscapes are inherently system-specific, with no single domain factor (e.g., system area, programming language, or resource intensity) consistently shaping their structure; (2) the core options (e.g., pic-struct of x264), which control the main functional flows, exert a stronger influence on landscape ruggedness (i.e. the difficulty of tuning) compared to resource options (e.g., cpu-independent of x264); (3) Workload effects on landscape structure are not uniformly tied to type or scale. Both contribute to landscape variations, but their impact is system-dependent.

We introduce Delta-Aware Quantization (DAQ), a data-free post-training quantization framework that preserves the knowledge acquired during post-training. Standard quantization objectives minimize reconstruction error but are agnostic to the base model, allowing quantization noise to disproportionately corrupt the small-magnitude parameter deltas ($ΔW$) that encode post-training behavior -- an effect we analyze through the lens of quantization as implicit regularization. DAQ replaces reconstruction-based objectives with two delta-aware metrics -- Sign Preservation Rate and Cosine Similarity -- that directly optimize for directional fidelity of $ΔW$, requiring only the base and post-trained weight matrices. In a pilot FP8 study, DAQ recovers style-specific capabilities lost under standard quantization while maintaining general performance.

Accurately predicting the state-of-health (SOH) and remaining useful life (RUL) of lithium-ion batteries is crucial for ensuring the safe and efficient operation of electric vehicles while minimizing associated risks. However, current deep learning methods are limited in their ability to selectively extract features and model time dependencies for these two parameters. Moreover, most existing methods rely on traditional recurrent neural networks, which have inherent shortcomings in long-term time-series modeling. To address these issues, this paper proposes a multi-task targeted learning framework for SOH and RUL prediction, which integrates multiple neural networks, including a multi-scale feature extraction module, an improved extended LSTM, and a dual-stream attention module. First, a feature extraction module with multi-scale CNNs is designed to capture detailed local battery decline patterns. Secondly, an improved extended LSTM network is employed to enhance the model's ability to retain long-term temporal information, thus improving temporal relationship modeling. Building on this, the dual-stream attention module-comprising polarized attention and sparse attention to selectively focus on key information relevant to SOH and RUL, respectively, by assigning higher weights to important features. Finally, a many-to-two mapping is achieved through the dual-task layer. To optimize the model's performance and reduce the need for manual hyperparameter tuning, the Hyperopt optimization algorithm is used. Extensive comparative experiments on battery aging datasets demonstrate that the proposed method reduces the average RMSE for SOH and RUL predictions by 111.3\% and 33.0\%, respectively, compared to traditional and state-of-the-art methods.

Machine learning systems deployed in medical devices require governance frameworks that ensure safety while enabling continuous improvement. Regulatory bodies including the FDA and European Union have introduced mechanisms such as the Predetermined Change Control Plan (PCCP) and Post-Market Surveillance (PMS) to manage iterative model updates without repeated submissions. This paper presents AI/ML Evaluation and Governance Infrastructure for Safety (AEGIS), a governance framework applicable to any healthcare AI system. AEGIS comprises three modules, i.e., dataset assimilation and retraining, model monitoring, and conditional decision, that operationalize FDA PCCP and EU AI Act Article 43(4) provisions. We implement a four-category deployment decision taxonomy (APPROVE, CONDITIONAL APPROVAL, CLINICAL REVIEW, REJECT) with an independent PMS ALARM signal, enabling detection of the critical state in which no deployable model exists while the released model is simultaneously at risk. To illustrate how AEGIS can be instantiated across heterogeneous clinical contexts, we provide two examples: sepsis prediction from electronic health records and brain tumor segmentation from medical imaging. Both cases use identical governance architecture, differing only in configuration. Across 11 simulated iterations on the sepsis example, AEGIS yielded 8 APPROVE, 1 CONDITIONAL APPROVAL, 1 CLINICAL REVIEW, and 1 REJECT decision, exercising all four categories. ALARM signals were co-issued at iterations 8 and 10, including the critical state where no deployable model exists and the released model is simultaneously failing. AEGIS detected drift before observable performance degradation. These results demonstrate that AEGIS translates regulatory change-control concepts into executable governance procedures, supporting safe continuous learning for adaptive medical AI across diverse clinical applications.

The recent advancements introduced by Large Language Models (LLMs) have transformed how Artificial Intelligence (AI) can support complex, real world tasks, pushing research outside the text boundaries towards multi modal contexts and leading to Multimodal Large Language Models (MLMs). Given the current adoption of LLM based assistants in solving technical or domain specific problems, the natural continuation of this trend is to extend the input domains of these assistants exploiting MLMs. Ideally, these MLMs should be used as real time assistants in procedural tasks, hopefully integrating a view of the environment where the user being assisted is, or even better sharing the same point of view via Virtual Reality (VR) or Augmented Reality (AR) supports, to reason over the same scenario the user is experiencing. With this work, we aim at evaluating the quality of currently openly available MLMs to provide this kind of assistance on technical tasks. To this end, we annotated a data set of furniture assembly with step by step labels and manual references: the Manual to Action Dataset (M2AD). We used this dataset to assess (1) to which extent the reasoning abilities of MLMs can be used to reduce the need for detailed labelling, allowing for more efficient, cost effective annotation practices, (2) whether MLMs are able to track the progression of assembly steps (3) and whether MLMs can refer correctly to the instruction manual pages. Our results showed that while some models understand procedural sequences, their performance is limited by architectural and hardware constraints, highlighting the need for multi image and interleaved text image reasoning.

While climate models provide insights for climate decision-making, their use is constrained by significant computational and technical demands. Although machine learning (ML) emulators offer a way to bypass the high computational costs, their effective use remains challenging. The hurdles are diverse, ranging from limited accessibility and a lack of specialized knowledge to a general mistrust of ML methods that are perceived as insufficiently physical. Here, we introduce a framework to overcome these barriers by integrating both climate science and machine learning perspectives. We find that designing easy-to-adopt emulators that address a clearly defined task and demonstrating their reliability offers a promising path for bridging the gap between our two fields.

We present Future-KL Influenced Policy Optimization (FIPO), a reinforcement learning algorithm designed to overcome reasoning bottlenecks in large language models. While GRPO style training scales effectively, it typically relies on outcome-based rewards (ORM) that distribute a global advantage uniformly across every token in a trajectory. We argue that this coarse-grained credit assignment imposes a performance ceiling by failing to distinguish critical logical pivots from trivial tokens. FIPO addresses this by incorporating discounted future-KL divergence into the policy update, creating a dense advantage formulation that re-weights tokens based on their influence on subsequent trajectory behavior. Empirically, FIPO enables models to break through the length stagnation seen in standard baselines. Evaluated on Qwen2.5-32B, FIPO extends the average chain-of-thought length from roughly 4,000 to over 10,000 tokens and increases AIME 2024 Pass@1 accuracy from 50.0% to a peak of 58.0% (converging at approximately 56.0\%). This outperforms both DeepSeek-R1-Zero-Math-32B (around 47.0%) and o1-mini (approximately 56.0%). Our results suggest that establishing dense advantage formulations is a vital path for evolving ORM-based algorithms to unlock the full reasoning potential of base models. We open-source our training system, built on the verl framework.

Data assimilation (DA) for systems governed by partial differential equations (PDE) aims to reconstruct full spatiotemporal fields from sparse high-fidelity (HF) observations while respecting physical constraints. While full-grid low-fidelity (LF) simulations provide informative priors in multi-fidelity settings, recovering an HF field consistent with both sparse observations and the governing PDE typically requires per-instance test-time optimization, which becomes a major bottleneck in time-critical applications. To alleviate this, amortized reconstruction using generative models has recently been proposed; however, such approaches rely on full-field HF supervision during training, which is often impractical in real-world settings. From a more realistic perspective, we propose the Physics-Informed Conditional Schrödinger Bridge (PICSB), which transports an informative LF prior toward an observation-conditioned HF posterior without any additional inference-time guidance. To enable learning without HF endpoints, PICSB employs an iterative surrogate-endpoint refresh scheme, and directly incorporates PDE residuals into the training objective while enforcing observations via hard conditioning throughout sampling. Experiments on fluid PDE benchmarks demonstrate that PICSB enables extremely fast spatiotemporal field reconstruction while maintaining competitive accuracy under sparse HF supervision.

Direct numerical simulations of turbulent reacting flows involving millions of grid points and detailed chemical mechanisms with hundreds of species and thousands of reactions are computationally prohibitive. To address this challenge, we present two data-driven chemical mechanism reduction formulations based on graph neural networks (GNNs) with message-passing transformer layers that learn nonlinear dependencies among species and reactions. The first formulation, GNN-SM, employs a pre-trained surrogate model to guide reduction across a broad range of reactor conditions. The second formulation, GNN-AE, uses an autoencoder formulation to obtain highly compact mechanisms that remain accurate within the thermochemical regimes used during training. The approaches are demonstrated on detailed mechanisms for methane (53 species, 325 reactions), ethylene (96 species, 1054 reactions), and iso-octane (1034 species, 8453 reactions). GNN-SM achieves reductions comparable to the established graph-based method DRGEP while maintaining accuracy across a wide range of thermochemical states. In contrast, GNN-AE achieves up to 95% reduction in species and reactions and outperforms DRGEP within its target conditions. Overall, the proposed framework provides an automated, machine-learning-based pathway for chemical mechanism reduction that can complement traditional expert-guided analytical approaches.

Graph-structured data typically exhibits complex topological heterogeneity, making it difficult to model accurately within a single Riemannian manifold. While emerging mixed-curvature methods attempt to capture such diversity, they often rely on implicit, task-driven routing that lacks fundamental geometric grounding. To address this challenge, we propose a Geometric Mixture-of-Experts framework (GeoMoE) that adaptively fuses node representations across diverse Riemannian spaces to better accommodate multi-scale topological structures. At its core, GeoMoE leverages Ollivier-Ricci Curvature (ORC) as an intrinsic geometric prior to orchestrate the collaboration of specialized experts. Specifically, we design a graph-aware gating network that assigns node-specific fusion weights, regularized by a curvature-guided alignment loss to ensure interpretable and geometry-consistent routing. Additionally, we introduce a curvature-aware contrastive objective that promotes geometric discriminability by constructing positive and negative pairs according to curvature consistency. Extensive experiments on six benchmark datasets demonstrate that GeoMoE outperforms state-of-the-art baselines across diverse graph types.

Despite the rapid advancement of Virtual Try-On (VTON) and Try-Off (VTOFF) technologies, existing VTON methods face challenges with fine-grained detail preservation, generalization to complex scenes, complicated pipeline, and efficient inference. To tackle these problems, we propose OmniDiT, an omni Virtual Try-On framework based on the Diffusion Transformer, which combines try-on and try-off tasks into one unified model. Specifically, we first establish a self-evolving data curation pipeline to continuously produce data, and construct a large VTON dataset Omni-TryOn, which contains over 380k diverse and high-quality garment-model-tryon image pairs and detailed text prompts. Then, we employ the token concatenation and design an adaptive position encoding to effectively incorporate multiple reference conditions. To relieve the bottleneck of long sequence computation, we are the first to introduce Shifted Window Attention into the diffusion model, thus achieving a linear complexity. To remedy the performance degradation caused by local window attention, we utilize multiple timestep prediction and an alignment loss to improve generation fidelity. Experiments reveal that, under various complex scenes, our method achieves the best performance in both the model-free VTON and VTOFF tasks and a performance comparable to current SOTA methods in the model-based VTON task.

Group dance generation from music requires synchronizing multiple dancers while maintaining spatial coordination, making it highly relevant to applications such as film production, gaming, and animation. Recent group dance generation models have achieved promising generation quality, but they remain difficult to deploy in interactive scenarios due to bidirectional attention dependencies. As the number of dancers and the sequence length increase, the attention computation required for aligning music conditions with motion sequences grows quadratically, leading to reduced efficiency and increased risk of motion collisions. Effectively modeling dense spatial-temporal interactions is therefore essential, yet existing methods often struggle to capture such complexity, resulting in limited scalability and unstable multi-dancer coordination. To address these challenges, we propose ST-GDance++, a scalable framework that decouples spatial and temporal dependencies to enable efficient and collision-aware group choreography generation. For spatial modeling, we introduce lightweight distance-aware graph convolutions to capture inter-dancer relationships while reducing computational overhead. For temporal modeling, we design a diffusion noise scheduling strategy together with an efficient temporal-aligned attention mask, enabling stream-based generation for long motion sequences and improving scalability in long-duration scenarios. Experiments on the AIOZ-GDance dataset show that ST-GDance++ achieves competitive generation quality with significantly reduced latency compared to existing methods.

Emergency vehicle (EV) response time is a critical determinant of survival outcomes, yet deployed signal preemption strategies remain reactive and uncontrollable. We propose a return-conditioned framework for emergency corridor optimization based on the Decision Transformer (DT). By casting corridor optimization as offline, return-conditioned sequence modeling, our approach (1) eliminates online environment interaction during policy learning, (2) enables dispatch-level urgency control through a single target-return scalar, and (3) extends to multi-agent settings via a Multi-Agent Decision Transformer (MADT) with graph attention for spatial coordination. On the LightSim simulator, DT reduces average EV travel time by 37.7% relative to fixed-timing preemption on a 4x4 grid (88.6 s vs. 142.3 s), achieving the lowest civilian delay (11.3 s/veh) and fewest EV stops (1.2) among all methods, including online RL baselines that require environment interaction. MADT further improves on larger grids, overtaking DT with 45.2% reduction on 8x8 via graph-attention coordination. Return conditioning produces a smooth dispatch interface: varying the target return from 100 to -400 trades EV travel time (72.4-138.2 s) against civilian delay (16.8-5.4 s/veh), requiring no retraining. A Constrained DT extension adds explicit civilian disruption budgets as a second control knob.

We present LoD-Loc v3, a novel method for generalized aerial visual localization in dense urban environments. While prior work LoD-Loc v2 achieves localization through semantic building silhouette alignment with low-detail city models, it suffers from two key limitations: poor cross-scene generalization and frequent failure in dense building scenes. Our method addresses these challenges through two key innovations. First, we develop a new synthetic data generation pipeline that produces InsLoD-Loc - the largest instance segmentation dataset for aerial imagery to date, comprising 100k images with precise instance building annotations. This enables trained models to exhibit remarkable zero-shot generalization capability. Second, we reformulate the localization paradigm by shifting from semantic to instance silhouette alignment, which significantly reduces pose estimation ambiguity in dense scenes. Extensive experiments demonstrate that LoD-Loc v3 outperforms existing state-of-the-art (SOTA) baselines, achieving superior performance in both cross-scene and dense urban scenarios with a large margin. The project is available at https://nudt-sawlab.github.io/LoD-Locv3/.

Tropical cyclones (TCs) pose severe threats to life, infrastructure, and economies in tropical and subtropical regions, underscoring the critical need for accurate and timely forecasts of both track and intensity. Recent advances in AI-based weather forecasting have shown promise in improving TC track forecasts. However, these systems are typically trained on coarse-resolution reanalysis data (e.g., ERA5 at 0.25 degree), which constrains predicted TC positions to a fixed grid and introduces significant discretization errors. Moreover, intensity forecasting remains limited especially for strong TCs by the smoothing effect of coarse meteorological fields and the use of regression losses that bias predictions toward conditional means. To address these limitations, we propose BaguanCyclone, a novel, unified framework that integrates two key innovations: (1) a probabilistic center refinement module that models the continuous spatial distribution of TC centers, enabling finer track precision; and (2) a region-aware intensity forecasting module that leverages high-resolution internal representations within dynamically defined sub-grid zones around the TC core to better capture localized extremes. Evaluated on the global IBTrACS dataset across six major TC basins, our system consistently outperforms both operational numerical weather prediction (NWP) models and most AI-based baselines, delivering a substantial enhancement in forecast accuracy. Remarkably, BaguanCyclone excels in navigating meteorological complexities, consistently delivering accurate forecasts for re-intensification, sweeping arcs, twin cyclones, and meandering events. Our code is available at https://github.com/DAMO-DI-ML/Baguan-cyclone.

Social insect colonies and ensemble machine learning methods represent two of the most successful examples of decentralized information processing in nature and computation respectively. Here we develop a rigorous mathematical framework demonstrating that ant colony decision-making and random forest learning are isomorphic under a common formalism of \textbf{stochastic ensemble intelligence}. We show that the mechanisms by which genetically identical ants achieve functional differentiation -- through stochastic response to local cues and positive feedback -- map precisely onto the bootstrap aggregation and random feature subsampling that decorrelate decision trees. Using tools from Bayesian inference, multi-armed bandit theory, and statistical learning theory, we prove that both systems implement identical variance reduction strategies through decorrelation of identical units. We derive explicit mappings between ant recruitment rates and tree weightings, pheromone trail reinforcement and out-of-bag error estimation, and quorum sensing and prediction averaging. This isomorphism suggests that collective intelligence, whether biological or artificial, emerges from a universal principle: \textbf{randomized identical agents + diversity-enforcing mechanisms $\rightarrow$ emergent optimality}.

The increasing global aging population has intensified the demand for reliable health monitoring systems, particularly those capable of detecting critical events such as falls among elderly individuals. Traditional fall detection approaches relying on single-modality acceleration data suffer from high false alarm rates, while conventional machine learning methods require extensive hand-crafted feature engineering. This paper proposes a novel multi-modal deep learning framework, MultiModalFallDetector, designed for real-time elderly fall detection using wearable sensors. Our approach integrates multiple innovations: a multi-scale CNN-based feature extractor capturing motion dynamics at varying temporal resolutions; fusion of tri-axial accelerometer, gyroscope, and four-channel physiological signals; incorporation of a multi-head self-attention mechanism for dynamic temporal weighting; adoption of Focal Loss to mitigate severe class imbalance; introduction of an auxiliary activity classification task for regularization; and implementation of transfer learning from UCI HAR to SisFall dataset. Extensive experiments on the SisFall dataset, which includes real-world simulated fall trials from elderly participants (aged 60-85), demonstrate that our framework achieves an F1-score of 98. 7, Recall of 98. 9, and AUC-ROC of 99. 4, significantly outperforming baseline methods including traditional machine learning and standard deep learning approaches. The model maintains sub- 50ms inference latency on edge devices, confirming its suitability for real-time deployment in geriatric care settings.

Intercavity polaritons, hybrid quasiparticles with spatially separated photonic and excitonic components, provide a platform to engineer structured light-matter states. We show that resonant driving of the middle polariton branch leads to a qualitatively distinct dynamical regime in which coherent Rabi oscillations are suppressed, and the system evolves monotonically toward its steady state. Including interactions, we demonstrate that this regime supports Bogoliubov excitations with a phonon-like dispersion at low momenta. These collective modes inherit interactions from the excitonic fraction, while preserving the intrinsically intercavity nature of the quasiparticles.

Understanding how electric fields destabilize biological membranes is important for electroporation-based technologies and bioelectronic interfaces. However, theoretical descriptions of this phenomenon remain fragmented. Existing theories treat either electrostatics in membranes of finite thickness or electrohydrodynamic flows at idealized zero-thickness interfaces, leaving unresolved a unified description that simultaneously incorporates finite membrane thickness, surface charge, and bulk electrohydrodynamics. Here, we apply a recently-developed, dimension-reduction framework that captures the coupled electrohydrodynamic and mechanical effects governing height fluctuations of a charged lipid bilayer of thickness $δ$ in an electrolyte characterized by Debye screening length $λ$. We derive voltage- and charge-dependent renormalizations of the effective surface tension and bending rigidity, along with a dispersion relation governing undulatory instabilities. A wide range of prior theoretical results arise as limiting cases of our more general theory when finite-thickness effects are neglected or screening is asymptotically strong. The key new contribution arises from traction moments generated across the finite membrane thickness, which are absent in zero-thickness descriptions. Under physiological screening ($δ/λ\sim 4$), these contributions account for more than $>70\%$ of the total electrostatic correction to both surface tension and bending rigidity. The theory further reveals that surface charges can stabilize the membrane at physiological ionic strengths, increasing the effective tension and shifting electroporation thresholds in a manner that depends on charge asymmetry between the leaflets.

We ask under what conditions a finite brickwork circuit of random gates retains local information about the initial state. To answer this question we measure the averaged Frobenius distance between the reduced states obtained by evolving two arbitrary initial states and tracing out a portion of the system. By characterising this distance exactly at all times we find that the information is retained if the environment -- the subsystem traced out -- is smaller than half of the system and washed away otherwise. We also find that, while the dynamics of the Frobenius distance depends on the specific initial states chosen, this dependence becomes increasingly weak for large scales and eventually the Frobenius distance attains a universal form as a function of time. Finally, we show that by introducing weak enough boundary dissipation, one can observe a phase transition between a memory preserving phase and one where the information is completely lost.

We study the Kondo effect induced by a topological soliton in a one-dimensional Dirac system with the sign-changing mass term. The soliton hosts a localized zero mode whose spatially extended wavefunction leads to a momentum-dependent exchange coupling with itinerant electrons. We show that this structure generates a nontrivial form factor that suppresses high-energy scattering processes, resulting in an energy-dependent effective Kondo coupling. As a consequence, the real-space structure of the soliton directly controls the emergent Kondo scale. This work establishes a mechanism by which topological defects control many-body energy scales through their wavefunction structure, suggesting a general principle for engineering many-body energy scales via topology.

We introduce a zero-bias superconducting voltage amplifier that harvests energy from a thermal gradient by exploiting negative differential resistance (NDR) in an asymmetric tunnel junction. The device is based on an asymmetric superconductor-insulator-superconductor (SIS) junction with an energy-gap ratio of $Δ_1/Δ_2 = 0.5$, connected in series with a load resistor. Owing to the superconducting bipolar thermoelectric effect, the current-voltage characteristic of the junction exhibits a region of NDR, in which the net current flows opposite to the applied voltage. This mechanism enables voltage amplification in the absence of any external electrical bias, relying solely on the temperature difference between the electrodes ($T_H \simeq 1$ K, $T_B \simeq 20$ mK). Numerical simulations predict a voltage gain of 20 dB, a 1 dB compression point at an input amplitude of 2 $μ$V, and a total harmonic distortion below $-50$ dB. The input-referred noise is approximately 1 nV/$\sqrt{Hz}$, with an associated thermal load on the order of nanowatts. The frequency response is broadband from near DC, with a $-3$ dB cutoff around 180 MHz, set by the RC time constant of the junction. Using Al-, Al-Cu-, and AlO$_x$-based technologies, the amplifier is compatible with conventional superconducting circuit fabrication processes. These findings demonstrate that thermoelectric superconducting junctions can deliver bias-free voltage amplification from near DC up to about 200 MHz, making them promising candidates for transition-edge sensor readout, quantum circuit instrumentation, and low-frequency cryogenic signal processing.

The deterministic realization of quantum light sources operating at telecom wavelengths is essential for long-distance fiber-based quantum communication and distributed quantum computing. In this work, we demonstrate that telecom O-band emission can be achieved from site-controlled InGaAs/GaAs quantum dots (QDs). Our concept utilizes a buried AlAs/Al$_2$O$_3$ stressor layer with the unique feature that induces a well-defined and controllable tensile strain field at the growth surface, enabling both a redshift of QD emission to the $\sim$1.3~μm range and site-selective nucleation at the mesa centers. This concept eliminates not only the need for strain-reducing layers (SRLs), which are known to degrade optical coherence, but also provides spatial control and spectral tunability. The grown telecom QDs show pure single-photon emission with $g^{(2)}(τ) = (5.0 \pm 1.0) \times 10^{-2}$ at 4 K and $(2.8 \pm 0.3) \times 10^{-1}$ at 77~K, demonstrating the quantum nature and thermal stability of the emitters. The emission characteristics of complex excitonic states are analyzed using 8-band $k \cdot p$ and configuration-interaction modeling, which quantitatively reproduces the experimental observations. Finally, we present a theory-supported strategy to further redshift the emission toward the center of the O-band and beyond by employing a multi-buried-stressor approach. This combined framework of experiment and theory establishes the buried stressor concept as a scalable route toward highly coherent, position-controlled O-band quantum emitters compatible with industrial photonic integration.

Fracture networks are ubiquitous in nature, spanning scales from millimeter-sized cracks in botanical peels to hundred-kilometer-long lineae on planetary satellites. The propagation of a crack is a complex, nonlinear phenomenon governed by the interplay of mechanical properties, rheological behavior, and system geometry. While fracture mechanics has long addressed structural failure, the relationship among fracture, elasticity, and nonlinear geometry has recently revived as a focal point in condensed matter and biophysics. However, a unified framework that systematically explains how surface geometry prescribes the transition between disparate fracture morphologies remains elusive. Here we show that shell curvature provides a geometric blueprint for fracture, governing the evolution of complex crack networks through induced stress anisotropy. By internally pressurizing thin, bilayer spheroidal shells, we demonstrate that a rich diversity of crack morphologies across lateral, longitudinal, and random orientations depends on the curvature ratio between the pole and the equator. We find that these patterns arise from the nonlinear mechanics of the shell, which can be leveraged to effectively control crack growth. Our results establish a direct link between structural curvature and fractures, providing a predictive framework that integrates nonlinear geometry with the classical Griffith and von Mises criteria. Beyond our model system, we find that the disparate fracture patterns observed in ripening muskmelons and in the icy crust of Europa follow the same geometric principles. We expect that this unified understanding of crack morphogenesis will inform the design principles of novel functional materials that are resilient to fracture and provide insights into the mechanical performance of curved biological and geophysical architectures.

Over the past decade, substantial progress has been made in clarifying a central question of the Fermi-Pasta-Ulam-Tsingou problem: whether weakly nonlinear lattice systems thermalize and, if so, through what mechanisms. The current understanding is as follows. (a) Classical lattice systems fall into two universal classes. In the first, the Hamiltonian has extended normal modes. For sufficiently large systems, the thermalization time scales as $T_{\rm eq}\sim g^{-γ}$ with $γ=2$, where $g$ denotes the effective nonlinear strength, i.e., the perturbation strength or degree of non-integrability. Thus, in the thermodynamic limit, these systems inevitably thermalize. Typical examples include common one-, two-, and three-dimensional lattice models. In the second class, all normal modes are localized. Here the relaxation time is essentially independent of system size. Although one may still formally write $T_{\rm eq}\sim g^{-γ}$, the exponent $γ$ diverges as $g\to0$, implying that arbitrarily weak nonlinear perturbations cannot induce thermalization. For sufficiently small $g$, such systems may therefore be viewed, in a theoretical sense, as thermal insulators. (b) In systems of the first class, disorder does not obstruct thermalization. Rather, by breaking translational symmetry and relaxing wave-vector resonance constraints, it increases the number of quasi-resonant processes and can therefore accelerate thermalization. (c) In systems of the second class, when both on-site potentials and disorder are present, all normal modes become localized in sufficiently large systems, suppressing thermalization. The perturbative framework underlying these conclusions will also be presented systematically, with particular emphasis on the thermalization criterion based on resonance-network connectivity, an approach rooted in weak wave turbulence theory.

Proliferation is a defining feature of life. Through growth, division, and death, living systems consume energy and inject mass, breaking conservation laws and driving collective phenomena from biofilm formation to embryonic development. Yet, while active matter physics has advanced our understanding of self-propelled agents, quantitative frameworks for proliferating systems are still emerging, and most work focuses on simplified settings. Here, we study \textit{E.coli} bacteria growing inside a network of single-file microchannels as a minimal model of structured environments. Competition for free volume drives the spontaneous emergence of coherent growth patterns that persist across generations but vanish when the channel links exceed the typical cell size at birth. Despite the strongly out-of-equilibrium character of the dynamics, the observed phenomenology can be quantitatively captured by an effective equilibrium description in which the flow state at each node is represented by a spin variable with ferromagnetic interactions. Simulations of growing elastic cells show that this coupling arises from internal stress accumulated at network nodes due to dynamical constraints. Our results reveal a surprising correspondence between proliferating active matter and equilibrium statistical mechanics, highlighting open fundamental questions and offering a first step toward describing growth phenomena in real-world complex environments.

In their Letter Fujiwara et al. (10.48550/arXiv.2511.09888. 2025) report a high-quality experiment demonstrating the synchronization of the CDW sliding in NbSe3 whiskers (nanowires) with surface acoustic waves (SAWs). The SAWs are induced in the conventional LiNbO3 piezoelectric substrates through application of rf voltage to an interdigital transducer (IDT). When a SAW mode is excited, Shapiro steps (ShSs) appear on the I-V curves, while no ShSs are observed when the frequency is shifted from the resonance. This gives clear evidence of synchronization of the CDW with the acoustic modes.

We investigate the Goos--Hänchen (GH) shifts and group delay time of Dirac fermions traversing a rectangular electrostatic potential barrier in silicene. By analyzing their dependence on the incident angle, barrier height, barrier width, and incident energy, we demonstrate that the GH shifts exhibit pronounced oscillations arising from quantum interference within the barrier region. The amplitude and number of oscillation peaks increase with increasing energy, barrier width, and incidence angle, resulting in enhanced lateral beam displacement. Meanwhile, the group delay time exhibits resonant features associated with the formation of quasi-bound states, increasing with barrier width, energy, and incidence angle, while decreasing with increasing barrier height. These results clarify how barrier-induced quantum interference controls both the lateral and temporal dynamics of Dirac fermions in silicene, highlighting the potential role of electrostatic barriers in enabling tunable transport in two-dimensional Dirac materials.

We introduce a fractional generalization of Tsallis entropy by acting with a $q$-Caputo operator on the generating family $\sum_i p_i^{\,x}$ evaluated at $x=1$. Concretely, we define $S_{q}^α$ through the $q$-Caputo differintegral of order $0<α<1$ and derive a closed series representation in terms of the $q$-Gamma function. The construction is anchored at the evaluation point, which ensures well-behaved limits: as $α\!\to\!1$ we recover the standard Tsallis entropy $S_q$. Finally we perform a numerical calculation to show the regions where the obtained $q$-fractional entropy $S^α_q$ can be non-negative (or negative) through the fractional parameter $α$ and the non extensive index $q$.

Altermagnets have recently emerged as a new platform for spintronics applications, offering spin-split electronic bands despite vanishing net magnetization. Here, we investigate spin-current generation in Dirac altermagnets and identify Klein tunneling as an efficient mechanism for enhancing spin transport. Using a low-energy Dirac model combined with scattering theory, we demonstrate that Klein tunneling in altermagnets is strongly spin-dependent and can be used to effectively control the electronic spin-current polarization by, for instance, adjusting the height, width and orientation of the potential barrier. Finally, we explore how the l-wave symmetry of the Dirac altermagnet shapes the spin-current polarization and transmission, focusing especially on the d- and g-wave cases. Particularly promising results are obtained for the g-wave Dirac altermagnet, as it is found that the presence of a potential barrier can significantly boost the spin-current polarization, even when the intrinsic polarization due to the spin-split band structure is vanishingly small. For a barrier implemented via electrostatic gating, such a mechanism would in turn allow the spin-current polarization to be switched on and off via a gate voltage.

We investigate emergent dynamics due to non-reciprocity in the $\mathcal{O}(2)$ model. The lattice XY model, where non-reciprocity stems from vision cone like couplings, can be described by a continuum description in which non-reciprocity translates into a new term depending on the rotational of the orientation field. We argue that non-reciprocity is akin to activity and we highlight the connection between our hydrodynamic equation and the constant density Toner-Tu framework. The active force advects and reshapes patterns, a generic feature found in many non-reciprocal systems. We show how $1d$ excitations in the non-reciprocal $\mathcal{O}(2)$ model can be described by a generalized Burgers equation, derived from our continuum model. We then extend the results to $2d$ perturbations. As such, we establish the first principles of excitation trajectory control in a non-reciprocal $\mathcal{O}(2)$ medium. Concretely, we explain how tuning the degree of non-reciprocity and the orientation of the background medium impacts the time evolution of excitations. We also showcase how initially different excitations lead to very different dynamical behavior. Non-reciprocity also affects the stability of defect-free excitations with non-zero winding numbers and, unlike in its equilibrium $O(2)$ counterpart, enables the system, above a certain threshold, to relax to its ground state.

In this work, we investigate Gaussian Mixture Models ({\it abbrv} GMM) and the related problem of non parametric maximum likelihood estimation ({\it abbrv} NPMLE) from the perspective of statistical mechanics. In particular, we establish stability guarantees for the NPMLE procedure that extend well beyond the state of the art. Crucially, we obtain guarantees on the Kullback-Leibler divergence between NPMLE estimators and the ground truth, a type of result which has been known to be challenging in the literature on this problem. In particular, we provide high probability upper bounds on the KL divergence between the NPMLE and the true density that are of the order of $\min\big\{\frac{(\log n)^{d+2}}{n} , \frac{\log n}{\sqrt n}\big\}$, which cover a wide range of scenarios for the comparative sizes of $n$ and $d$. We obtain similar guarantees for approximate solutions to the NPMLE problem, addressing realistic situations wherein optimization algorithms need to be stopped in finite time, allowing access only to approximations to the true NPMLE. A technical cornerstone of our approach is an analysis of the function class complexity of logarithms of gaussian mixture densities, which is able to handle their unboundedness, and could be of wider interest. We also establish correspondences between stability phenomena in the NPMLE problem and concepts from chaos and multiple valleys in random energy landscapes of statistical mechanics models. We believe that these correspondences may be useful for a wide variety of random optimization problems in statistics and machine learning, especially the connections to the the technical ingredients of concentration phenomena and Langevin dynamics for these models.

We define a generalized Golomb-Dickman constant $λ_θ$ as the limiting expected proportion of the longest cycle in random permutations under the Ewens measure with parameter $θ> 0$. Exploiting the independence properties of Kingman's Poisson process construction of the Poisson-Dirichlet distribution, we obtain an explicit integral representation for $λ_θ$ in terms of the exponential integral. The dependence of $λ_θ$ on $θ$ reflects the transition between regimes dominated by long cycles (small $θ$) and those with many small cycles (large $θ$). Our result can be viewed as an extension of the classical calculations of Shepp and Lloyd to the Ewens setting by relatively elementary means. A figure and a table of numerical values of $λ_θ$ are included.

Moiré superlattices have emerged as a premier platform for simulating the Hubbard model, yet achieving high tunability in square-lattice systems remains a key challenge. We demonstrate that $Γ$-valley twisted square homobilayers provide a faithful and highly tunable realization of $t-t'-U$ Hubbard model, extending the recent proposal in M-valley systems. We show that at small twist angles, an emergent layer-exchange symmetry decouples electronic states into flat bands residing on two nested square sublattices. An interlayer displacement field breaks this symmetry to induce controllable inter-sublattice hybridization, enabling wide-range experimental tuning of the effective hopping ratio $t'/t$. By establishing a direct correspondence between $Γ$- and M-valley systems, we provide a unified framework for understanding displacement-field tunability in square moiré physics. These findings establish $Γ$-valley twisted bilayers as a versatile platform for simulating the square-lattice Hubbard model and exploring its rich landscape of correlated phenomena.

A voltage biased Zener diode always exhibit positive differential resistance, thus cannot be used as an element to provide amplification of a signal. We show how to induce negative differential resistance in the reverse bias regime of a 12V Zener diode by noise feedback. We use this to build a voltage amplifier in the audio frequency range, which we characterize by providing bandwidth, gain, power consumption, gain compression and output noise spectral density.

Two-dimensional Rokhsar-Kivelson (RK) dimer models on bipartite lattices are generally limited to translation-symmetry-broken dimer crystals. We introduce a tensor-product regularisation of the dimer Hilbert space that yields a qubit Hamiltonian interpolating from the RK model to the $π$-flux toric code, thereby accessing a deconfined $\mathbb{Z}_2$ topological liquid. In this framework, the $\mathbb{Z}_2$ liquid descends from a multicritical $U(1)$ spin liquid through condensation of a charge-2 Higgs field, thus avoiding confinement. Using iDMRG together with low-energy field theory, we determine a phase diagram containing two continuous quantum phase transitions -- a $3\mathrm{D}$ XY$^{\ast}$ transition between the $\mathbb{Z}_2$ liquid and the columnar/plaquette-VBS, and a quantum Lifshitz transition between two dimer crystals -- alongside a first-order transition between the staggered crystal and the $\mathbb{Z}_2$ liquid. Our field theory suggests a deconfined multicritical point described by an Abelian Higgs model with dynamical critical exponent, $z=2$, where the three transitions meet, highlighting the interplay of fractionalisation and emergent gauge fluctuations.

Phase-coherent superconducting proximity in topological materials requires clean superconductor-topological material (SC-TM) interfaces, yet conventional top-contact fabrication often degrades them through oxidation, polymer residue, and process-induced disorder. Here we introduce a pre-patterned superconducting bottom-contact architecture in which MoRe/Au electrodes are defined before van der Waals crystal transfer, thereby avoiding on-flake lithography after transfer. In WTe2- and Bi1.5Sb0.5Te1.7Se1.3-based Josephson junctions, this architecture yields systematically larger I_c R_N and longer-ranged coupling than conventional top contacts. Cross-sectional STEM/EDS reveals atomically abrupt, chemically well-separated interfaces. These results establish pre-patterned SC-TM contacts as a practical route to reproducible, micrometer-scale Josephson platforms in van der Waals topological materials.

Gallium oxide (Ga$_2$O$_3$) is a wide-bandgap semiconductor with exceptional electrical and optical properties, making it a promising material for optoelectronic and sensing applications. In this work, we demonstrate for the first time the formation of plasmonic silver (Ag) and gold (Au) nanoparticles embedded in Ga$_2$O$_3$ thin films via ion implantation. Ga$_2$O$_3$ films deposited by RF sputtering on sapphire substrates were implanted with Ag or Au ions at 150 keV and a nominal fluence of 5 $\times$ 10$^{16}$ ions/cm$^2$, followed by thermal annealing between 200 and 700 °C. Rutherford backscattering spectrometry (RBS) measurements revealed saturation effects during implantation, resulting in lower incorporated fluences, as well as out-diffusion with post-implantation annealing. Transmission electron microscopy confirmed the formation of metallic nanoparticles with a distribution consistent with the metal profiles measured by RBS. Optical absorption measurements showed a pronounced localized surface plasmon resonance (LSPR) band in the Ag-implanted films, visible even in the as-implanted state and red-shifting with increasing annealing temperature, while Au-implanted films exhibited a distinct LSPR peak only after annealing at $\geq$500 °C. The observed LSPR shifts with annealing are attributed primarily to changes in the Ga$_2$O$_3$ matrix rather than a change in nanoparticle size. These results establish ion implantation as a viable approach for integrating plasmonic nanostructures into Ga$_2$O$_3$.

We analyze the metastability of Bose-Hubbard condensates for finite-size one-dimensional ring lattices and open chains, using a semiclassical tomographic perspective that emphasizes the relation of the many-body spectrum to the underlying classical phase-space structures. This constitutes an arena for inspection of quantum ergodicity and localization, in far-from-equilibrium scenarios of experimental interest. Both local aspects (via Bogoliubov analysis) and global aspects (by inspecting the mixed regular-chaotic dynamics) are addressed. We also clarify how chaos is diminished in the limit of the Gross-Pitaevskii equation.

Magnetic nanoflowers (MNFs), which exhibit large intrinsic magnetic losses and high specific absorption rates under clinically relevant alternating magnetic fields, highlight strong potential as efficient mediators for magnetic hyperthermia. In this work, we provide a versatile platform for creating dynamic, field-responsive microstructures based on MNFs through a flexible, low-cost, and template-free self-assembly strategy driven by tunable interparticle interactions, external magnetic fields, and spatial confinement. By controlling ionic strength, particle coverage, surface charge, particle concentration, and confinement, MNFs spontaneously assemble in aqueous solution into magnetic micropillars and microfilaments without predefined scaffolds, with low ionic strength favoring reversible assemblies and intermediate salt concentrations yielding stable, irreversible structures. The size, geometry, and dynamic response of these architectures can be precisely tuned, enabling field-induced behaviors such as cilia-like rotations, oscillations, and torque-driven detachment of micropillars into free-standing, swarming microfilaments. L-dopamine (L-DOPA) was used for surface modification in this work, as it is a biocompatible ligand offering catechol, amine, and carboxylate groups. Resulting MNFs@L-DOPA have negative surface charge and show assembly behavior qualitatively similar to the uncoated system under magnetic actuation. Together, these results establish practical guidelines for the template-free design of biomimetic, functional magnetic and elongated microarchitectures, highlighting their potential for microfluidic manipulation and bio-microrobotic applications.

Eigenstate thermalization refers to the property that an energy eigenstate of a many-body system is indistinguishable from a thermal equilibrium ensemble at the same energy as far as expectation values of local observables are concerned. In systems with degeneracies, the choice of an energy eigenbasis is not unique and the fraction of basis states exhibiting eigenstate thermalization can vary. We present a simple example where this fraction vanishes in the thermodynamic limit for one basis choice, but remains nonzero for another choice. In other words, the weak eigenstate thermalization hypothesis is satisfied in the first, but violated in the second basis. We furthermore prove that degeneracies must abound whenever a system is simultaneously symmetric under spatial translations and reflection. Finally, we derive general bounds on how strongly eigenstate thermalization may depend on the choice of the basis, and we reveal some interesting implications regarding the temporal relaxation properties of such systems.

There has been recent interest in the relaxational modes of small-scale fully connected systems of aligning self-propelled particles (Spera et al., Phys. Rev. Lett. {\bf 132}: 078301 (2024)). We revisit the classical connection between Fokker-Planck and Schrödinger equations to address this by means of exact diagonalization, allowing for rigorous analytical insight into the full spectrum. This allows us to extract exact results which we compare to the existing result from linearized statistical field theory. We derive asymptotically correct analytical results that improve upon the prior approximations. We show that this methodology can fruitfully be extended to the case of non-reciprocal interactions which gives rise to a non-Hermitian Schrödinger problem akin to those in open quantum mechanics. While the non-reciprocity can be chosen such as not to alter the stationary distribution, it fundamentally changes the nature of the steady state which we quantify via the entropy production. We discuss the case of low particle numbers as well as the emergence of mean-field dynamics at large numbers.

Cavity exciton--polaritons are quasiparticles that form when quantum well excitons hybridize with a cavity mode. Here, we carry out photon correlation measurements under continuous wave resonant laser excitation to demonstrate quantum correlations between cavity--polaritons. Our experiments reveal an unexpectedly strong dependence of polariton interactions on cavity--exciton detuning. When the polaritons are predominantly exciton-like, we observe a transition from photon antibunching to bunching as the laser is tuned across the polariton resonance, in agreement with a simple Kerr-nonlinearity model. When the lower-branch polariton energy is tuned to induce a two-polariton Feshbach resonance with the biexciton mode, the degree of polariton antibunching becomes independent of the laser detuning: we explain our finding by invoking a dissipative blockade mechanism arising from large biexciton broadening. Our experiments demonstrate that the strong polariton blockade regime would be achieved by reducing the polariton decay rate by a factor of 10.

In this study, motivated by recent advancements in non-Hermitian physics, we explore new characteristics of Dirac semimetals (DSMs) using the spectral singularities by means of scattering techniques, with the goal of uncovering additional unique properties. To achieve this, we investigate how the axion texture of a DSM affects its topological properties by analyzing its interaction with electromagnetic waves. We examine the transverse electric (TE) mode configuration, where the magneto-electric effect induces a dichroic property in these materials. This behavior is particularly interesting and commonly seen in potential DSM candidates. Consequently, we report for the first time that a dichroic DSM generates 12 unique topological laser types. We discover that surface currents are generated by topological terms on the surface of the DSM slab. Furthermore, we examine how the θ term associated with axions in topological materials contributes to these topological properties. Our study reveals distinct topological role of the θ term more clearly than ever before. Our results confirm that the topological properties of DSMs with a single Dirac cone remain stable under external influences and that a topologically robust DSM laser can be developed accordingly

In single-particle tracking experiments measuring anomalous diffusion dynamics, understanding ergodicity is crucial, as it ensures that the time average of an observable matches the ensemble average, and can thus be fitted with known ensemble-averaged observables. A commonly used criterion for assessing the ergodicity of a stochastic process is based on the comparison of the mean-squared displacement (MSD) with the time-averaged MSD (TAMSD). This approach has been widely applied and proves effective in cases of weak ergodicity breaking across various systems in both theoretical and experimental studies. However, there is relatively little discussion regarding the theoretical justification and limitations of this definition. Here, we demonstrate that this widely accepted criterion to some extent contradicts the classical definition of ergodicity as well as physical intuition, leading to spurious (non-)ergodicity results when applied to several well-known stochastic models. To address this limitation, we propose using the mean-squared increment (MSI) instead of the MSD for comparison of ensemble- and time-averaged observables. Several well-established examples demonstrate that our MSI-TAMSD criterion not only effectively reveals weak ergodicity breaking, equivalent to the MSD-TAMSD approach, but also provides a more accurate characterization of the genuine (non-)ergodicity of systems where the MSD-TAMSD method fails. Additionally, for systems exhibiting "ultraweak" ergodicity breaking, the MSI can reveal the asymptotic stationarity and ergodic nature of the process' increments. Our findings emphasize the important role of the MSI observable for SPT experiments and anomalous diffusion studies.

In this study, we investigate the electronic structure, charge transport dynamics, and optical properties of a quantum dot light-emitting diode (QD-LED) featuring a double nanorod (NR) emission layer composed of CdSe-ZnS core-shell structures. Utilizing a rigorous self-consistent numerical approach, we solve the coupled Schrodinger-Poisson equations iteratively to obtain accurate wave functions, energy levels, and potential profiles under varying external bias voltages. Detailed analyses reveal voltage-dependent electron localization dynamics, demonstrating a systematic transition of electrons between distinct NR regions via quantum tunneling. Charge density and electrostatic potential distributions are modeled comprehensively, employing the asymmetric Erlang distribution to characterize interface effects. By calculating current-voltage (I-V) characteristics and photoluminescence spectra, we demonstrate that external voltage serves as a robust tuning parameter for modulating emission energies and intensities, underscoring the potential of these NR-LED systems for tunable optoelectronic and photonic applications.

The collective action of actively contractile units embedded in elastic biopolymer networks plays a crucial role in regulating the network's macroscopic mechanical response. Here, we investigate how the macroscopic boundary stress in model elastic fiber networks depends on the number and nature of embedded contractile units, each exerting an isotropic force dipole, as well as on the bending stiffness of fibers. We find that the macroscopic stress increases nonlinearly with the number of dipoles due to mutual stiffening of initially soft, bending-dominated networks. Using effective medium theory, we relate this enhanced contractility to an increase in the effective average network coordination number due to constraints imposed by the force dipoles. By comparing three distinct force dipole models that differ in their local structures, we demonstrate that the specific manner in which an active unit constrains the network strongly influences the onset and nature of the stiffening transition. Our results highlight that not only the quantity but also the local geometry of force-generating units critically determines the macroscopic mechanical behavior. This framework provides a physical basis for understanding how biological systems-such as molecular motors in the cytoskeleton, or adherent cells in the extracellular matrix-can modulate network-scale nonlinear elastic properties through local tuning of active force-generating units.

Solitons are universal nonlinear excitations that appear in settings as varied as optics, water waves, and quantum gases [1-5]. While reduced models of soliton dynamics are well established, their validity and dynamical behaviour in strongly nonlinear regimes with frequent interactions remain largely unexplored experimentally. Progress has been constrained by the difficulty of simultaneously achieving precise control of dispersion and nonlinearity, together with the temporal and spatial resolution required for dynamical observations. Here we overcome these difficulties by producing acoustic solitons in integrated phononic waveguides. We exploit the interplay between waveguide dispersion and mechanical Kerr nonlinearity to generate 'dark' solitons that persist over metre-scale propagation distances. The slow phonon velocity allows direct imaging of hundreds of dark soliton collisions -- two orders of magnitude more than have previously been accessible [6, 7] -- as well as soliton fission and the melting of a soliton Wigner crystal. Furthermore, the unprecedented dynamical resolution allows us to verify two long-predicted aspects of dark soliton behaviour: the existence of a collisional phase shift and two depth-dependent collision regimes [8, 9]. These results not only illuminate fundamental nonlinear energy transport processes, but also show a path towards acoustic versions of soliton-enabled technologies such as frequency combs and mode-locked lasers [1, 2, 10].

The scheme of experiment for studying electron scattering in the field of a magnetic monopole in two dimensional electron gas is proposed. The differential scattering cross section is obtained in the eikonal approximation. For unpolarized initial electron, the differential cross section coincides with that in the field of an infinitely long solenoid, up to redefinition of a magnetic flux. It is shown that the scattered electron becomes polarized even for unpolarized initial electron. Besides, in an experimental setup similar to the Hall experiment, the spin polarization arises in the direction perpendicular to the electron current.

Platinum oxides are vital catalysts, but their limited thermal stability hinders applications. Recent studies have uncovered a structural transition in two-dimensional platinum oxides that significantly enhances their thermal resilience by several hundred Kelvin. Herein, we demonstrate that this enhanced stability stems from the mechanical robustness of the elastic network at the atomic scale. Prior to the transition, an over-constrained lattice generates localized states of self-stress through an incommensurate Moiré pattern with the platinum substrate, reducing thermal endurance. After the transition, the oxide shifts to a mechanically flexible structure with balanced degrees of freedom and constraints. The isostatic network, together with the platinum substrate, forms a commensurate Moiré superlattice that relaxes elastic energy and enhances stability. These findings highlight the fundamental role of network connectivity in governing thermal stability, and provide a design principle for catalysts in extreme environments.

Symmetric mass generation (SMG) transitions defy the conventional Landau-Ginzburg-Wilson paradigm by opening a many-body gap without spontaneous symmetry breaking or topological order, attracting intense interest across particle physics and condensed matter physics. Here, we utilize unbiased quantum Monte Carlo simulations to investigate the equilibrium and nonequilibrium critical dynamics of the SMG transition in a bilayer honeycomb lattice model. We unambiguously confirm the existence of an SMG transition at $J_{\text{c}}=2.584(8)$ that separates the Dirac semimetal phase from a symmetry-preserving SMG phase. High-precision extraction of the critical exponents reveals a novel universality class that profoundly departs from mean-field theory. We then extend our study to the nonequilibrium regime, exploring the driven dynamics of the SMG transition. Notably, despite the breakdown of the prerequisites for the celebrated Kibble-Zurek mechanism, the nonequilibrium SMG transition still follows the generalized finite-time scaling. By bridging equilibrium criticality and nonequilibrium dynamics, our work uncovers the universal critical properties of SMG transitions, providing a solid theoretical basis for future experimental studies of SMG physics.

Understanding heat transfer across solid-liquid interfaces is central to thermal management and energy technologies, yet whether the interfacial thermal conductance (ITC) depends on the timescale of heating remains unclear. Here we use square-pulsed source thermoreflectance, which combines time-resolved detection with broadband modulation, to probe Al-water and Al-octane interfaces. We observe a reproducible increase of the apparent ITC with modulation frequency. A control Al-silica interface shows no measurable frequency dependence, indicating that the effect is specific to liquids rather than a generic feature shared by all amorphous materials. We explain the data with a two-channel liquid picture in which diffusional and phonon-like modes exchange energy weakly over a finite nonequilibrium length. From the relative magnitude of the thermal penetration depth and the nonequilibrium length, we identify three transport regimes. These findings challenge the common assumption of fully equilibrated liquid modes and provide experimental constraints for modeling dynamic energy exchange at liquid interfaces.

The Landau quantization space is based on the respective motion of the electron and hole in a magnetic field and can provide a new route to understand the bound exciton behaviors observed in the experiments. In this paper, we study the two-dimensional exciton properties of monolayer WSe$_2$ in both the real space and Landau quantization space. Focusing on the excitons of zero center-of-mass momentum, we calculate its energy spectrum in both spaces, with the results agreeing well with each other. We then obtain the diamagnetic coefficients and root-mean-square radius, which are consistent with the $s$ state results from the experiments. More importantly, in the exciton state $nl$, we find that the dominant electron-hole pair component may shift with the magnetic field and the Coulomb interactions, and reveal that the magnetic field will drive the dominant component to be the free electron-hole pair $\{n_e=n+l-1,n_h=n-1\}$, whereas the Coulomb interactions drives it to be the pair of the lower index.

We study a mesoscopic elasto-plastic model of amorphous matter with varying dimensionless compression modulus, $K/μ$, where $K$ and $μ$ are the compression and shear moduli. We study both cyclic shear with amplitude $Γ$ and forward steady shear. In cyclic shear, the terminal behavior is, in order of increasing $Γ$: i) trivially elastic, ii) hysteretic but with microscopically reversible limit cycles, iii) diffusive with no return to previously visited configurations. We show that the transition between i) and ii) at the onset point $Γ_0$ is determined by the Eshelby back stress, $σ_0$, which depends on the Poisson ratio. Systems with smaller $K/μ$ (more compressible) are effectively harder with a higher $Γ_0$ and a correspondingly larger purely elastic regime in cyclic loading. In forward shear, $σ_0$ plays a similar role where lower $K/μ$ results in a higher steady state flow stress, $σ_y$. We show that increasing $K/μ$ increases the amplitude of stress redistribution after a local yielding event without changing the net stress relaxation and relate this to the assumptions in mean-field descriptions of amorphous solids. A striking feature of the model is the emergence of a complex hardening behavior in the absence of any ad-hoc hardening parameters: a transition between a kinematic and an isotropic hardening behavior precisely at $Γ_0$ associated with the hysteresis transition. The enhanced plastic response for incompressible systems is also seen in amorphous alloys where it is usually attributed to excess free volume, while in the present model, it arises from the dependence of the Eshelby backstress on the Poisson ratio. Our results should have important implications for amorphous metallic alloys or other glassy systems where $K/μ$ can vary with composition, age, quench procedure, or mechanical processing history.

We introduce the non-Hermitian mosaic Maryland model, where a discrete modulation period and a non-Hermitian phase are incorporated into the potential, rendering the originally exactly solvable system generally non-integrable. This model provides a unique platform to investigate how structural modulation governs localization in complex quasiperiodic potentials. Using Avila's global theory, we analytically derive the exact Lyapunov exponent and obtain explicit formulas for the complex mobility edges. Remarkably, for modulation periods kappa >= 2, the system intrinsically hosts kappa-1 robust extended bands that persist independently of the potential strength and non-Hermiticity. We further characterize the topological nature of these phases via the spectral winding number. Unlike the standard Maryland model, the mosaic modulation induces mobility edges, and the resulting phase transitions are continuous, reflecting the non-integrable nature of the system. Numerical calculations of the inverse participation ratio and fractal dimension confirm the analytical predictions for the asymptotic form of the mobility edges in the large non-Hermiticity limit. This work establishes structural design as a powerful degree of freedom for engineering wave transport and enhancing the robustness of extended states in non-Hermitian systems.

The non-Hermitian skin effect (NHSE), characterized by a macroscopic accumulation of eigenstates at the edge of a system with open boundaries, is often ascribed to a non-trivial point-gap topology of the Bloch Hamiltonian. We revisit this connection and show that the eigenspectrum of non-normal operators is highly sensitive to boundary conditions and generic perturbations, and therefore does not constitute a stable object encoding topological information. Instead, topological properties are reflected in the singular-value spectrum of finite systems and, in the semi-infinite limit, correspond to boundary-localized eigenmodes implied by the index of the corresponding Toeplitz operator. For a Hatano-Nelson ladder, where point-gap winding and non-normality can be varied independently, we demonstrate that the NHSE can occur without point-gap winding and, conversely, that point-gap winding can persist without the NHSE. These results establish that the NHSE originates from spectral instability and non-reciprocity rather than topology, and that the commonly assumed relation between spectral winding and boundary localization relies on translational invariance and is therefore not generic.

Control of oxygen vacancies by electrical currents in complex oxides such as YBa$_2$Cu$_3$O$_{7-δ}$ (YBCO) has attracted considerable interest due to the relative simplicity of its implementation and its potential for both fundamental studies and the tuning of superconducting device properties. However, the structural evolution and depth-dependent effects associated with current-based techniques remain largely unexplored, particularly with respect to the connection between optical signatures and the spatial distribution of oxygen vacancies. Here, we combine nanoprobe X-ray Diffraction (NanoXRD), Cu K-edge X-ray Absorption Near-Edge Structure (XANES), X-ray Photoelectron Spectroscopy (XPS), electrical transport, and optical measurements to reveal modifications induced in YBCO microbridges by pulsed electromigration. We observe a c-axis expansion correlated with spectroscopic features of oxygen depletion in the Cu-O chains, and we confirm that oxygen redistribution, crystallographic changes, and copper coordination evolve consistently across techniques. Notably, the spatial profile of unit-cell expansion closely follows the optical contrast observed after electromigration, demonstrating that the different signatures capture the same underlying oxygen reordering. We further show that optical microscopy cannot reliably capture bipolar electromigration involving strong resistance modifications, as surface deoxygenation appears largely irreversible. Taken together, our findings provide a significant step toward a microscopic understanding of current-assisted oxygen migration in YBCO and establish a framework for effectively exploiting vacancy control in high-temperature superconducting devices.

We review recent theoretical and experimental developments in magnetic Weyl semimetals, focusing on the electromagnetic responses emerging from the interplay of their electronic band topology and magnetism. We begin by introducing the fundamental topological properties of the electrons in Weyl semimetals, and provide an overview of the characteristic phenomena arising from their band topology, such as the anomalous Hall effect and chiral magnetic effect. The materials exhibiting the magnetic Weyl semimetal state, with ferromagnetic ordering, antiferromagnetic ordering, etc., are listed. The possible mechanisms for their magnetism are discussed in connection with the Weyl electrons. Non-uniform magnetic textures and magnetization dynamics are expected to exhibit a topological interplay with the Weyl electrons, manifesting as spinmotive force and spin torques. We also review the magnetotransport phenomena such as domain wall magnetoresistance, studied by mesoscopic scale calculations. Finally, we mention the spin transport properties studied in magnetic Weyl semimetals. The topological nature of Weyl electrons reviewed here is important not only for fundamental physics, but also for the potential application to low-dissipative electronics and spintronics devices.

Classical thermodynamics contains familiar geometric relations associated with cyclic processes, most notably the identification of mechanical work with the area enclosed by a trajectory in the $(P,V)$ plane. We show that the area laws for work and reversible heat arise as projections of a single canonical two--form defined on the equilibrium thermodynamic manifold, providing a unified description of thermodynamic cycles in both the $(P,V)$ and $(T,S)$ representations. The same structure yields a direct link between cycle geometry and thermodynamic response: the work generated by infinitesimal cycles is set locally by the mixed curvature $U_{SV}$ of the equilibrium energy surface, which can be expressed in terms of measurable susceptibilities. This identifies thermodynamic work as a local geometric field over state space rather than solely a global property of cyclic processes. More broadly, the framework connects classical cycle geometry to stochastic thermodynamic trajectories, providing a geometric interpretation of nonequilibrium work relations such as the Jarzynski equality.

Recent advances in on-surface chemistry, combined with scanning probe microscopy, have enabled the synthesis of correlated molecules on surfaces and the characterization of their chemical and electronic properties with unprecedented spatial resolution. Low-energy magnetic excitations of individual molecules are frequently investigated by scanning tunneling spectroscopy (STS) and often appear as symmetric step-like features in the differential conductance as a function of bias voltage. The interpretation of such steps is well established within cotunneling theory and effective model Hamiltonians (e.g., Hubbard- and spin-based models). Here, we extend the cotunneling formalism to general multireference systems. We show that multireference character, together with orbital-dependent and strongly asymmetric tip/substrate couplings, can produce pronounced asymmetric line shapes in inelastic STS. These results provide an alternative microscopic mechanism for the asymmetric peaks and dips near the Fermi level frequently observed in STS experiments.

We formulate a geometric framework for quasistatic thermodynamics in open quantum systems by parameterizing the dynamics on a control manifold. In the quasistatic limit, the system follows a manifold of stationary states, and the work performed over a cycle is given by the flux of a curvature two-form, $W \sim \int Ω$, defined by the parametric response of the stationary state, establishing an open-system analog of classical thermodynamic area laws. For thermal stationary states, the curvature is isotropic and depends only on the instantaneous energy scale, yielding a population-driven geometry in which environmental parameters reshape how work is distributed across the control manifold. Beyond this limit, nonequilibrium stationary states can retain coherence in the energy representation; using a fixed-basis Lindblad model, we show that this coherence reshapes the curvature, making it anisotropic and sign-changing, so that work depends sensitively on the placement and orientation of the cycle. Quantum coherence therefore partitions the control manifold into regions of opposite curvature, producing geometric cancellation of work and allowing the net work over a cycle to be reduced or reversed despite dissipative dynamics. Thermodynamic work thus emerges as a curvature flux whose structure is set by thermodynamic response in classical systems and by basis misalignment between the Hamiltonian eigenbasis and the environment-selected pointer basis in open quantum systems.

Deterministic classical cellular automata can be in two phases, depending on how irreversible the dynamical rules are. In the strongly irreversible phase, trajectories with different initial conditions coalesce quickly, while in the weakly irreversible phase, trajectories with different initial conditions can remain different for a time exponential in the system volume. The transition between these phases is referred to as the damage-spreading transition (the "damaged" sites are those that differ between the trajectories). We develop a theory for this transition. In the simplest and most generic setting, the transition is known to be related to directed percolation, one of the best-studied nonequilibrium phase transitions. However, we show that full theory of the damage-spreading critical point is richer than directed percolation, and contains an infinite hierarchy of sectors of local observables. Directed percolation describes the first level of the hierarchy. The higher observables include "overlaps" for multiple trajectories, and may be labeled by set partitions. (These higher observables arise naturally if, for example, we consider decay of entropy under the irreversible dynamics.) The full hierarchy yields a hierarchy of nonequilibrium fixed points for reaction-diffusion-type processes, all of which contain directed percolation as a subsector, but which possess additional universal critical exponents. We analyze these higher fixed points using a field theory formulation and renormalization group arguments, and using simulations in 1+1 dimensions.

We study the simplest terms that need to be included in active field theories to couple them to external potentials. To do so, we consider active Brownian particles and implement a systematic perturbative expansion in the particle persistence time. The result is a non-trivial coupling between density and potential gradients, which accounts for the nonequilibrium features of active particles in the presence of an external potential, from boundary accumulation to far-field density modulation. We show how the method can be applied to particles interacting via pairwise forces and to spatial modulations of the propulsion speed.

Non-Hermitian systems provide a powerful platform for engineering and controlling nonequilibrium phenomena beyond Hermitian settings, with the presence of non-Hermitian skin effect broadening the scope of dynamical control. Here, we develop a general theory of non-Hermitian systems driven exclusively at their boundaries, providing a unified description of the driving-frequency dependence of bulk spectra and dynamics in the thermodynamic limit. Our framework extends non-Bloch band theory to time-periodic systems at arbitrary boundary driving frequencies. Applying it to representative models, we demonstrate boundary-driving-induced parity-time symmetry breaking, with the driving frequency serving as a control knob and the driving amplitude providing an additional handle in finite-size systems. These results establish boundary Floquet driving as a versatile mechanism for controlling bulk properties of non-Hermitian systems and open new routes for dynamical engineering in driven open systems.

We study sum rules that control the Regge limit of one-dimensional conformal field theory (CFT) correlators and relate them to dual bulk scattering processes at high energies in $\mathrm{AdS}_2$. By imposing the condition that no scattering takes place in the bulk, these sum rules single out special solutions to crossing symmetry that describe long-range models, which can be understood as free fields in AdS with boundary interactions tuned to criticality. We test these sum rules perturbatively in several distinct theories, namely the 1d long-range versions of the Ising, $O(N)$ and Lee--Yang models, and find that they correctly predict the CFT data characterising these theories. Along the way we compute for the first time the leading contributions of quadruple-twist operators to the long range Ising correlator and analyse their role in the new sum rules. Finally, we explore the consequences of imposing these sum rules in a numerical bootstrap framework and find that they lead to substantial reductions in the allowed parameter space.

The arrow of time is usually attributed to two mechanisms: decoherence through environmental entanglement, and chaos through nonlinear dynamics. Here we demonstrate a third route, Precision-Induced Irreversibility (PIR), requiring neither. No entanglement. No nonlinearity. Just three ingredients: amplification, non-normality, and finite dynamic range, whose interplay yields an operational arrow of time; remove any one and reversibility can be restored. Non-Hermitian evolution remains mathematically invertible, yet beyond a sharp temporal predictability horizon scaling linearly with available precision, distinct states collapse onto identical representations. Echo-fidelity tests confirm this transition across arbitrary-precision calculations and hardware, revealing where formal invertibility and physical reversibility diverge.

Global entanglement in quantum many-body systems is inherently nonlocal, raising the question of whether it can be inferred from local observations. We investigate this problem in monitored quantum circuits, where projective measurements generate classical records distributed across spacetime. Using graph neural networks (GNNs), we represent individual quantum trajectories as directed spacetime graphs and reconstruct the half-chain entanglement entropy from local measurement data alone. Because information propagates through the network via local message passing, the architecture directly controls the spacetime region over which correlations can be aggregated. By systematically varying this accessible scale -- through network depth and hierarchical spacetime coarse-graining -- we probe how much measurement information is required to reconstruct global entanglement. We find that prediction accuracy improves as the accessible spacetime region grows and that results from different architectures collapse when expressed in terms of an effective spacetime scale combining depth and coarse-graining. These results demonstrate that the information required to reconstruct global entanglement is organized in spacetime scales and show that graph-based learning architectures provide a controlled operational framework for probing how global quantum correlations emerge from local measurement data.

We predict a supercurrent-driven Néel spin-orbit torque in a superconductor/$d$-wave altermagnet heterostructure, associated with the emergence of spin-triplet correlations. The effect can be understood as a consequence of the supercurrent-induced spin polarization, owing to the interplay between spin-orbit coupling and momentum-dependent spin splitting, as found, for example, in altermagnets. Remarkably, the supercurrent can be tuned by the Néel-vector direction, and the supercurrent-induced torque can both propel magnetic domain walls and reverse the Néel-vector orientation within a domain wall. These findings establish superconductor/altermagnet heterostructures as a versatile platform for the dissipationless control of the Néel vector, with potential applications in racetrack memory, dissipationless superconducting electronics, and unconventional computing.

We establish an exact Wilson-loop conjugation identity, $W(-δ)=W(δ)^*$, for the many-body overlap Wilson-loop $W(δ)$ accumulated along a $U(1)$ flux-threading (twist) cycle parametrized by $θ\in[0,2π]$, where $δ$ denotes the bond-dimerization parameter. A minimal sufficient condition is the existence of a composite antiunitary mapping acting on the flux-threaded Hamiltonian family that implements $(δ,θ)\mapsto(-δ,-θ)$. As a concrete demonstration, we construct such a mapping microscopically in a dimerized staggered Hubbard ring at half filling. We then verify the conjugation identity using the density-matrix renormalization group (DMRG) for gapped, nondegenerate ground states along the twist cycle. Importantly, the identity persists in depinned gapped regimes where the Berry-phase $γ\equiv-\arg W$ is not symmetry-quantized; as a corollary, $γ(-δ)=-γ(δ)$ (mod $2π$). More generally, the same conjugation relation applies to any lattice model whose flux-threaded Hamiltonian family is closed under an orientation reversal of the bond pattern (a suitable permutation of link-hopping parameters) combined with a reversal of the flux orientation.

Ultrathin films of widely different materials exhibit a dramatic enhancement of projectile penetration resistance under high--velocity impact. Despite extensive simulations and experiments, a unifying physical explanation has remained elusive. Here we show that the thickness dependence of the specific penetration energy obeys a universal law, $E_p^*(h)=E_{p,\infty}^*+B h^{-3}$, independent of chemical composition and degree of disorder. The inverse--cube scaling is traced back to a finite--size correction to the effective shear modulus arising from the suppression of long--wavelength nonaffine deformation modes in confined solids. The scaling quantitatively describes impact data for multilayer graphene, graphene oxide, and polymer thin films, revealing a common elastic origin for nanoscale impact resistance.

We report on the fabrication and characterization of patterned high-mobility two-dimensional electron gases (2DEG) formed on SrTiO$_3$ (STO) substrate surfaces by hydrogen plasma exposure. The resulting devices consistently showed high electron mobilities up to 7400 cm$^2$/V$\cdot$s. A large range of electron density was systematically explored by controlled aging of the sample between cooldowns, including the expected range for the STO 2DEG superconducting dome. No superconducting transition was observed down to the base temperature of approximately 10 mK. This suggests suppression of superconductivity in high mobility quasi-two-dimensional SrTiO$_3$ electron gas, likely linked to vertical confinement and electronic orbital rearrangement. We systematically explored electrostatic gate modulation in this 2DEG system and its scaling with electron density and side gate geometry. In contrast with our initial expectation, we observed an improvement of achievable total modulation for larger side gate to channel separation. At low electron density, stochastic channel pinch-off events were observed, creating quasi-ballistic constrictions with irregular conductance quantization. This epitaxy-free and high mobility oxide material platform offers a promising new route towards patterning quantum devices.

Previous experimental reports of long-range interactions in ionic liquids (ILs) stand in contradiction with theoretical predictions and numerical simulations. To provide insights into the literature discrepancies regarding the experimental ranges of electrostatic screening, claimed with orders of magnitude larger, the interactions between pairs of mica and borosilicate surfaces confining ILs are investigated by two complementary advanced Surface Force Apparatuses. Regardless of differences in confinement geometries (crossed-cylinders, sphere-flat), radii of curvature (cm-mm), and measurement techniques (stepwise vs continuous approach), two ever present force regimes are evidenced. At small surface separations, oscillatory forces reflect IL structuration and layering, while outside this gap, the interaction is monotonic repulsive. In both regimes the spatial extent and force magnitude depend critically on motion conditions, as demonstrated by achieving velocities as low as 9 pm/s with equilibration times up to 90 s. At large separations, fast surface displacements generate long-range interactions (over tens of ion size) creating the illusion of anomalous underscreening, whereas increasingly slow ones shrink both magnitude and range of the repulsion with decay-lengths converging ultimately to a screening length consistent with Poisson-Boltzmann theory with finite ion sizes. The transition from apparent long-range to short-range screening unfolds over nearly two orders of magnitude in time, revealing slow relaxation dynamics reminiscent of aging phenomena. These findings definitely resolve a decade-old controversy on force measurements and reveal rich out-of-equilibrium dynamics. The hydrodynamic contribution to the net force is admittedly crucial to be reduced especially when relaxations span decades in time, but approaching thermodynamic equilibrium during measurements proves essential.

Quantum anomalies arise when symmetries of a classical theory cannot be preserved upon quantization, leading to unconventional topological responses. A prominent example is the parity anomaly of a single two-dimensional Dirac fermion, which enforces a half-quantized Hall response. Anomaly inflow mechanism allows this effect to be observed at the surfaces of three-dimensional topological insulators, however, its realization in a genuinely two-dimensional system has remained elusive. Here we report the observation of a parity-anomalous Hall response at the critical point of a quantum Hall topological phase transition in a synthetic two-dimensional system of ultracold dysprosium atoms. By coupling a continuous spatial dimension to a finite synthetic dimension encoded in atomic spin states, we engineer tunable Chern bands with C = 0 and 1. At the transition, the bulk gap closes at a single Dirac point, where we observe a robust half-quantized Hall drift despite strong non-adiabatic excitations. We show that this response originates from the global structure of the band topology, is protected by an emergent parity symmetry at criticality, and disappears when parity is explicitly broken. Our work establishes synthetic quantum systems as a powerful platform to probe quantum anomalies and their interplay with topology and non-equilibrium dynamics.

Recently, a class of spin chains known as ``free fermions in disguise'' (FFD) has been discovered, which possess hidden free-fermion spectra even though they are not solvable via the standard Jordan-Wigner transformation. In this work, we extend this FFD framework to open quantum systems governed by the Gorini-Kossakowski-Sudarshan-Lindblad (GKSL) equation. We establish a general class of exactly solvable open quantum systems within the FFD framework: if the Liouvillian frustration graph is claw-free and has a simplicial clique, the Liouvillian possesses a hidden free-fermion spectrum. In particular, the (even-hole, claw)-free condition automatically guarantees this, enabling exact computation of the Liouvillian gap and an infinite-temperature autocorrelation function. Our results provide the first realization of the FFD mechanism in open quantum systems.

Nonlinear response functions, formulated as multipoint correlation functions or Volterra kernels, encode the dynamical and spectroscopic properties of physical systems and underpin a wide range of nonlinear transport and optical phenomena. However, their evaluation rapidly becomes prohibitive at high orders because of combinatorial (often factorial) scaling or severe numerical errors. Here, we establish a systematic and efficient framework to compute nonlinear response functions directly from real-time dynamics, without explicitly constructing multipoint correlators or relying on numerically unstable finite-difference methods for order-resolved extraction. Our approach is based on the Gateaux derivative with respect to the external field in function space, which yields a closed hierarchy of tangent equations of motion (TEOM). Propagating the TEOM alongside the original dynamics isolates each perturbative order with high accuracy, providing a term-by-term decomposition of physical contributions. The computational cost scales exponentially with response order in the fully general setting and reduces to polynomial complexity when all perturbation directions are identical; both regimes avoid the factorial scaling of explicit multipoint-correlator evaluations. We demonstrate the power of TEOM by computing frequency-resolved fifth-order response functions for a solid-state electron model and by obtaining nonlinear response functions up to the 49th order with controlled accuracy in a classical Duffing oscillator. We further show that our time-evolution formulation allows optical conductivities to be evaluated directly while remaining numerically stable even near zero frequency. TEOM can be incorporated seamlessly into existing real-time evolution methods, yielding a general framework for computing nonlinear response functions in quantum and classical dynamical systems.

We investigate how non-Markovian mechanical dissipation affects exceptional points in linearized optomechanical systems with red-sideband drive. For a chosen non-Ohmic mechanical bath, we derive analytical conditions for the memory-renormalized exceptional point by employing a pseudomode mapping, thereby demonstrating that structured environments displace the mode coalescence away from the Markovian prediction. Crucially, we reveal that failing to account for this memory-induced shift suppresses the divergent Petermann factor by orders of magnitude, showing that accurate bath modeling is essential for the successful operation of exceptional-point-based devices whenever reservoir-induced memory is non-negligible. We finally show that non-Markovianity modifies the cavity reflection spectrum, manifesting as a shallower optomechanically-induced-transparency dip, providing therefore an experimentally-accessible signature of structured mechanical environments.

Graphene quantum point contacts (QPCs) in the quantum Hall regime host competing transport mechanisms including chiral edge propagation, valley degeneracy, and gate-induced mode mixing. Their interplay is not visible in conductance alone. Shot noise directly probes the statistics of transmission eigenvalues, revealing microscopic mode partitioning that conductance cannot access. We develop a hybrid framework combining tight-binding simulations of gate-defined graphene QPCs with random matrix theory (RMT) to predict shot noise and Fano factor signatures across different quantum Hall regimes, validated against experimental conductance maps of hBN-encapsulated graphene Hall bars. Three distinct regimes are identified: adiabatic propagation, sharp mode filtering, and multi-mode mixing driven by localized states beneath the split gate. For higher Landau levels ($N_L > 0$), complete mode mixing produces the universal chaotic-cavity limit $F \simeq 1/4$. Strikingly, the zeroth Landau level ($N_L = 0$) converges to $F = 1/3$. This distinct value originates in the sublattice polarization of the $N_L = 0$ edge state: coupling to mixed-sublattice localized states beneath the gate is suppressed, confining transport to an effective single channel ($N = 1$). Complete mixing within this single channel yields a flat transmission eigenvalue distribution and hence exactly $F = 1/3$ from single-channel RMT, numerically coincident with but mechanistically distinct from pseudo-diffusive zero-field graphene transport. The $F = 1/3$ versus $F = 1/4$ crossover is a Landau-level-resolved noise signature absent in conductance, providing a direct discriminator between single-channel and multi-channel chaotic transport in graphene QPCs.

Traditional percolation theory assumes static microscopic rules, limiting its ability to describe real-world complex systems where macroscopic order actively regulates local interactions. Here, we introduce feedback percolation, an unified framework that dynamically couples the microscopic activation probability to the macroscopic size of the giant component. We show that this simple feedback mechanism produces a rich variety of behaviors both analytically and numerically. Depending on the feedback functions, the system exhibits explosive discontinuous jumps, hybrid transitions, limit-cycle oscillations, and routes to chaos, absent in classical percolation. Our findings establish that macroscopic feedback provides a unifying physical mechanism for phenomena ranging from self-regulating oscillations to systemic infrastructure collapse.

The two-dimensional Heisenberg spin-glass model is investigated by means of a semiclassical expansion around classical states. At leading order, we obtain an effective quadratic spin-wave Hamiltonian and study the localization properties of its spectrum and eigenfunctions. We find that the nature of the spin-wave excitations, whether they are hydrodynamic or localized modes, depends crucially on the relevance/irrelevance - in the renormalization group sense - of the correlations induced by the underlying classical order in the spin-wave Hamiltonian matrix elements: low-energy excitations around magnetically ordered states are delocalized, whereas those around spin-glass ordered states are localized, albeit weakly. Remarkably, in the magnetically ordered case, spin-wave delocalization is robust with respect to the presence of disorder, even in two spatial dimensions. We interpret this phenomenology by relating the spontaneous breaking of spin-rotation symmetry in the original Heisenberg model to the symmetry and universality class of the resulting quadratic spin-wave Hamiltonian. We conjecture that the hydrodynamic picture can be recovered through the inclusion of interactions among the spin-wave excitations at higher order in the semiclassical expansion, favoring the onset of ergodic behavior.

We comment on the recent work by Yamaguchi and Barré [Phys. Rev. E 107, 054203 (2023)], which uses linear stability analysis of the Vlasov equation to characterize phase transitions in a generalized Hamiltonian Mean Field (gHMF) model. By performing extensive molecular dynamics simulations with $N=10^8$ particles, we demonstrate that the bifurcation analysis of the initial stationary distribution is insufficient to predict either the location or the nature of the phase transition to a quasi-stationary state (qSS). Specifically, we show that for bimodal momentum distributions, the instability threshold identified by the authors does not correspond to a ferromagnetic transition; instead, the system remains in a paramagnetic state characterized by magnetization oscillations with a zero time-average. We find that the true paramagnetic-ferromagnetic transition is discontinuous (first-order) and occurs at significantly larger coupling strengths, characterized by a clear coexistence of states. These results indicate that linear bifurcation and symmetry-breaking phase transitions are distinct phenomena in long-range interacting systems, and that the former lacks the predictive power to describe the long-time fate of the system.

We study experimentally the nonlinear propagation of short pulses of forward volume spin waves in nanometer-thick YIG films. We show that nonlinearity of the spin system can efficiently counteract dispersion broadening of the pulses, leading to the formation of envelope solitons. We demonstrate that in microscopic YIG systems, microwave powers of the order of one milliwatt are sufficient to reach the soliton formation threshold. At powers slightly above this threshold, we achieve transmission of 3-ns spin-wave pulses over distances of up to 50 micrometers without increase in their temporal width. Our results demonstrate a promising way towards high-rate transmission of information in microscopic spin-wave circuits unaffected by detrimental dispersion effects.

Landau theory usually treats free-energy coefficients as intrinsic parameters fixed by thermodynamic variables. We show that externally written microscale fields can survive coarse graining and enter the free-energy functional as spatially prescribed coefficient fields. This defines a nonintrinsic sector of Landau theory. The key condition is a hierarchy of correlation, writing, and frustration lengths. We identify ion-patterned FeRh as a plausible realization.

We investigate the equilibrium magnetization in ferromagnetic nanoscrews (NSw) using micromagnetic simulations. These systems consist of elongated three-dimensional magnetic membranes with helicoidal geometry, combining curvature, torsion ($\mathrm{w}$), and eccentricity ($ε$) along their length. We focus on the influence of these geometric parameters, together with membrane thickness and inner diameter, on remanent states and coercive fields. Our results, obtained over a broad range of eccentricities and torsions, reveal bistable magnetic behavior, with vortex-domain-wall propagation during magnetization reversal. We identify four degenerate configurations of a remarkably stable mixed remanent state. The coercive field is found to increase with eccentricity for structures with a major axis (larger inner diameter) approximately 30\% larger than the minor axis (smaller inner diameter), while remaining largely insensitive to variations in torsion. These findings are interpreted in terms of geometry-induced modifications of surface magnetostatic charges on the membrane mantle. Overall, our results demonstrate that nanoscrews exhibit robust bistability under systematic geometric deformation, together with enhanced coercivity, highlighting their potential for applications in three-dimensional nanomagnetism.

We solve numerically exactly a simple toy model to quantum general relativity or more properly to path integral on a curved space. We consider the thermal equilibrium of a quantum many body problem on the sphere, the surface of constant positive curvature. We use path integral Monte Carlo to measure the kinetic energy, the internal energy and the static structure of a bosons, fermions and anyons fluid at low temperatures on the sphere. For bosons we also measure the superfluid fraction and compare its behavior at the critical temperature with the universal jump predicted by Nelson and Kosterlitz in flat space in the thermodynamic limit at the superfluid phase transition. For fermions and anyons it is necessary to use the restricted path integral recipe in order to overcome the sign problem. Even if this recipe is exact for the non interacting fluid it reduces to just an approximation for an interacting system. And we make the example of the electron gas at low temperature. Snapshots of the many body path configuration during the evolution of the computer experiment show that the ``speed'' of the single particle path near the poles slows down as a consequence of the ``hairy ball theorem'' of Poincaré. The influence of curvature on the thermodynamic and structural properties of the many body fluid is also studied.

We present an in-depth analysis of the orbital magnetism by means of the so-called modern theory based on the Berry phase across distinct classes of materials-d transition metals, sp metals, and transition metal dichalcogenides-highlighting the microscopic nature of band structure characteristics. We adopt a gauge-covariant formulation of the modern theory proposed in [Lopez et al. Phys. Rev. B 85, 014435 (2012)], which enables the calculation of orbital magnetism in a controlled manner in any chosen gauge of Wannier functions and gives the total contribution as a gauge-invariant measurable. This captures consistently the contributions due to the anomalous position, velocity, and orbital angular momentum of Wannier basis, as well as the contributions due to Hamiltonian such that their sum is gauge-invariant. For d transition metals, we find that the atom-centered approximation captures the majority of the total contribution given by modern theory, which we attribute to localized nature of d electrons. However, 5d metals tend to exhibit larger deviation between the two methods than 3d metals do, as 5d electrons are more delocalized than 3d electrons. On the other hand, sp metals exhibit a strong deviation between the two methods, where large kinetic energy of sp electrons is important. Finally, in 1H-MoS2, we find that the valley orbital moment far exceeds the atomic limit of d electrons due to coherent hybridization between valence and conduction bands in direct band gaps. Our work elucidates the interplay of the chemical nature of electronic orbitals and the effect of band structures in a consistent manner and highlights the role of Berry phase in orbital magnetism. The results suggest a promising direction of orbitronics beyond controlling atomic orbitals, in which the orbital magnetism can be greatly enhanced by exploiting Berry phase.

We analyze the mixed Hessian of the dispersionless Toda $τ$-function for the $s$-fold symmetric one-harmonic polynomial conformal map. The inverse branch exhibits two distinct thresholds: an analytic threshold $ζ_c$, where the dominant square-root singularity reaches the circle of convergence, and a later geometric threshold $ζ_{\mathrm{univ}}>ζ_c$, where the map ceases to be univalent. We prove that the first spectral instability occurs already at $ζ_c$. In each symmetry sector, the weighted subcritical realization has exactly one logarithmically diverging eigenvalue, whereas the remaining spectrum stays bounded and, after removal of the singular direction, converges to that of a compact limiting remainder. We further continue the corresponding scalar Gram functions beyond $ζ_c$, showing that they admit a generalized hypergeometric description, a Cauchy--Stieltjes representation, and, for $1\le p\le s$, a realization as Weyl functions of bounded Jacobi operators. In particular, these scalar quantities remain finite at $ζ_{\mathrm{univ}}$. This identifies analytic criticality, rather than loss of univalence, as the first spectral threshold of the Toda Hessian.

We present a simple method to study the dynamics of planar Kahan-Hirota-Kimura (KHK) maps preserving rational fibrations. Using this approach, we show that integrable KHK maps may exhibit complex dynamics, even when obtained from vector fields with trivial behavior. As an application, we study the KHK map associated with a quadratic planar vector field with an isochronous center. This map preserves the original first integral and admits the vector field as a Lie symmetry. Moreover, for a dense set of values of the integration step, it is globally periodic and exhibits all possible periods except 2. We also provide evidence of non-integrability for KHK maps associated with other quadratic vector fields possessing isochronous centers. To overcome this issue, we introduce the notion of pseudo-KHK maps, as alternative integrable discretizations for vector fields with isochronous centers. These maps are constructed to preserve the first integrals of the original vector field and to ensure that the vector field itself is a Lie symmetry of the map. The construction can be extended to isochronous centers of degree greater than two.

This research paper explores the potential of nonlinear magnetic levitation systems for energy harvesting by developing a modified system that incorporates a more realistic energy harvesting circuit, enabling a better representation of practical operating conditions. Methodologically, approximate solutions for the system dynamics were obtained using the method of multiple scales, complemented by numerical simulations to capture parameter variations visualized through phase planes and parameter variation plots. The results demonstrate that by adjusting capacitance to induce internal and primary resonances, an extended detuning formulation (utilizing parameters sigma3 and sigma4) is innovatively introduced to capture the coupled dynamic interaction between the harvesting circuit and the mechanical system under diverse conditions. Periodic variations in circuit charge and intermediate magnet dis placement were thoroughly analyzed. Comparisons with legacy models demonstrate that the proposed coupling mechanism effectively suppresses undesirable nonlinear behaviors, such as chaos and multi-stability, resulting in a more stable and predictable energy harvesting process. Ultimately, a critical trade-off between energy harvesting efficiency and dynamical stability is identified, providing a valuable new design perspective for practical energy harvesting systems.

We show a unified approach to the Ablowitz-Kaup-Newell-Segur (AKNS) hierarchy and the unreduced derivative nonlinear Schrödinger (DNLS) hierarchies (including the Kaup-Newell, Chen-Lee-Liu, Gerdjikov-Ivanov and a generalized DNLS), together with their multi-component extensions, in the framework of the anti-self-dual Yang-Mills (ASDYM) reduction. By restricting the gauge group to GL(2), the Kadomtsev-Petviashvili (KP) and modified KP (mKP) hierarchies are formulated in the ASDYM reduction via squared eigenfunction symmetry constraints. In this case, the bilinearization of the generalized DNLS equations can also be understood through this reduction. Finally, Gram-type exact solutions for the relevant equations are presented in terms of quasi-determinants.

We investigate parametric modulation of the nonlinear coupling in the Henon Heiles system, which directly modifies intrinsic resonance structure in a manner complementary to additive forcing. Canonical perturbation theory in extended phase space yields normal forms predicting resonance tongues scaling as $\sqrt{\varepsilon}$ near commensurate frequencies. Melnikov analysis quantifies separatrix splitting and chaos onset, confirmed by symplectic simulations showing transition from localized resonances to global transport via overlap. High-frequency averaging reveals potential stiffening that suppresses chaos. parametric modulation of nonlinear coupling provides an alternative route for generating combination resonances and influencing chaotic dynamics

We study the thermalization slowing down of Fermi-Past-Ulam-Tsingou (FPUT) chains and of Toda chains with nonintegrable boundaries. We focus on the transition from FPUT to harmonic chains, from FPUT to Toda chains with fixed boundaries, and from nonintegrable open boundary Toda to integrable fixed boundary Toda. We compute the Lyapunov spectrum and analyze its scaling properties upon approaching integrable limits. We analyze the scaling of the largest Laypunov exponent, the rescaled Lyapunov spectrum, and the Kolmogorov-Sinai entropy. Using additional analytic arguments we demonstrate evidence that all three cases are operating in the regime of a Long Range Network of nonintegrable perturbations.

It is shown that a generalization of the Painlevé-II equation (P-II) to a system of coupled equations with symmetry breaking terms is integrable. A Lax pair for this system is used to relate the asymptotic behavior of the solutions at different infinities via an asymptotically exact WKB approach. The analysis relies on an exact solution of the quantum mechanical Demkov-Osherov model (DOM). An application to the problem of unstable vacuum decay during a second order phase transition provides precise scaling of the number of excitations, including subdominant contributions.

Chimera states are one of the most intriguing phenomena in nonlinear dynamics, characterized by the coexistence of coherent and incoherent behavior in systems of coupled identical oscillators. Despite extensive studies and numerous observations in different settings, the development of reliable and systematic methods to classify chimera states and distinguish them from other dynamical patterns remains a challenging task. Existing approaches are often limited in scope and lack robustness. In this work, we propose a method based on Fourier analysis combined with statistical classification to characterize chimera behavior. The method is applied to a system of topological signals coupled via the Dirac operator, where it successfully captures the rich dynamical regimes exhibited by the model. We demonstrate that the proposed approach is robust with respect to variations in network topology and system parameters. Beyond the specific model considered, the framework provides a general and automated tool for distinguishing different dynamical regimes in complex systems.

The proposed algorithm seeks to provide a novel data-driven framework for the discovery of stochastic differential equations (SDEs) by application of the Weak-formulation to stochastic SINDy. This Weak formulation of the algorithm provides a noise-robust methodology that avoids traditional noisy derivative computation using finite differences. An additional novelty is the adoption of spatial Gaussian test functions in place of temporal test functions, wherein, the use of the kernel weight $K_j(X_{t_n})$ guarantees unbiasedness in expectation and prevents the structural regression bias that is otherwise pertinent temporal test functions. The proposed framework converts the SDE identification problem into two SINDy based linear sparse identification problems. We validate the algorithm on three SDEs, for which we recover all active non-linear terms with coefficient errors below 4\%, stationary-density total-variation distances below 0.01, and autocorrelation functions that reproduce true relaxation timescales across all three benchmarks faithfully.

Acoustic holography provides a practical means of flexibly controlling acoustic wavefronts. However, high-fidelity shaping of acoustic fields remains constrained by the numerical-physical gap inherent in conventional phase-only designs. These approaches realize a two-dimensional phase-delay profile as a three-dimensional thickness-varying lens, while neglecting wave-matter interactions arising from the lens structure. Here, we introduce an end-to-end, physics-aware differentiable structural optimization framework that directly incorporates three-dimensional lens geometries into the acoustic simulation and optimization loop. Using a novel differentiable relaxation, termed Differentiable Hologram Lens Approximation (DHLA), the lens geometry is treated as a differentiable design variable, ensuring intrinsic consistency between numerical design and physical realization. The resulting Thickness-Only Acoustic Holograms (TOAHs) significantly outperform state-of-the-art phase-only acoustic holograms (POAHs) in field reconstruction fidelity and precision under complex conditions. We further demonstrate the application of the framework to spatially selective neuromodulation in a neuropathic pain mouse model, highlighting its potential for non-invasive transcranial neuromodulation. In summary, by reconciling numerical design with physical realization, this work establishes a robust strategy for high-fidelity acoustic wavefront shaping in complex environments.

Density functional theory (DFT) offers an exceptional balance between accuracy and efficiency, but practical density functional approximations face an unavoidable trade-off among simplicity, accuracy, and transferability. A systematic protocol is therefore needed to develop functionals that are reliably most accurate within a chosen application domain. Here we present such a protocol by combining constraint enforcement, flexible functional forms, and modern optimization. Applying this strategy to the range-separated hybrid (RSH) meta-GGA framework, we obtain the carefully optimized and appropriately constrained hybrid (COACH) functional. Across broad molecular benchmarks, COACH improves both accuracy and transferability relative to leading RSH meta-GGAs, including \omegaB97M-V, while retaining the computational practicality of its rung. Finally, our analysis of the remaining trade-offs and saturation behavior suggests that further systematic progress will likely require the incorporation of genuinely nonlocal information.

Achieving climate neutrality in Europe requires rapid electrification alongside carbon management strategies for residual emissions. Existing analyses of the European energy system often focus on collocated carbon capture and geological sequestration, with limited attention to the interactions among carbon capture and utilization, transport, sequestration, and diverse carbon dioxide removal (CDR) options. Moreover, existing literature focuses on discussing the optimal, neglecting that near-optimal solutions might provide very different system configurations at a marginal higher cost. Here, we integrate afforestation, biochar, enhanced rock weathering, and perennialization into a sector-coupled European energy system model (PyPSA-Eur) clustered to 39 nodes with 750 aggregated time steps. We explore their contributions using a Modelling to Generate Alternatives (MGA) approach. The approach combines minimization, maximization, and random vectors to explore the near-optimal solution space for up to 5% increased total system costs. Our results show that, in a carbon-neutral system, multiple configurations of carbon management options can achieve net-zero emissions with only marginal cost increases. We find that a 5% total system cost increase is sufficient to accommodate the full spectrum from zero to full deployment of the individual CDR options, as well as a wide range of synthetic fuel use across different fuel types. Increased reliance on CDR options offers no clear cost advantage compared to greater utilization of synthetic fuels.

Laser-enhanced contact optimization (LECO) is increasingly used method to reduce contact resistance and recover fill factor in advanced crystalline silicon solar cells. However, the industrial transferability is limited because the same localized activation that improves carrier transport can also create kinetically unstable interface states. LECO can be viewed as a coupled Multiphysics process that links microstructural evidence to device-level signature that uses instantaneous regime map together with a reliability classification based on time-dependent drift. Thus, a predictive workflow is outlined in the review that couples (i) transient electrothermal modeling to reduced state metrics, (ii) effective diffusion depth and local areal energy density, and (iii) propagated calibrated thresholds across recipe space. The framework separates stable optimization from marginal activation and latent damage and explains why fine-line scaling and copper-containing stacks tighten stability margins through current localization and diffusion-barrier constraints. It sum up the next generation AI guided optimization digital twin.

This work provides a brief review of the numerical methods for modelling the optical propagation of an ultrashort, intense laser pulse in plasmas and ionizable gases. These methods are implemented in an open-source simulation toolkit Axiprop, which is now actively used for design studies in the context of laser plasma acceleration of electrons (LPA). We present two examples of such studies: optical plasma waveguide generation, and phase-locked flying-focus LPA. These tools enable the identification of complex effects impacting both energy deposition during waveguide formation, and the pulse propagation dynamics that governs wakefield structure, ultimately determining the properties of the accelerated electron bunches. The developed tools and presented simulation designs are relevant to experiments on laser plasma interactions at modern high power laser facilities.

We derive an exact reduction of the screened second-order exchange (SOSEX) energy in the uniform electron gas to a triple integral for a specific class of single-pole screened interaction. The reduction proceeds by rescaling the frequency variable to factorize the propagator denominators, applying a Fourier decomposition to separate the two particle-hole blocks, and finally performing a change of integration variables that brings the geometric structure into a tractable form. The reduction to a one-variable integral kernel is possible if and only if the screened interaction belongs to a one-pole class characterized by a single momentum-independent frequency scale~$μ$, which we call the reduction-compatible single-pole (RC-SP) model. The RC-SP model does not approximate plasmon dispersions in real materials, but provides an exactly reducible reference model for analyzing dynamically screened exchange, and gives natural basis elements for approximating more general one-pole screening. We analyze the $μ$-dependence of the SOSEX energy asymptotically at both small and large~$μ$ and establish the leading behaviors at the theorem level. Under a power-law mapping from~$μ$ to the density parameter~$r_s$, this asymptotic structure constrains the analytic form of the screened-exchange correction in $r_s$-space, providing a diagrammatically justified basis for beyond-RPA functional construction. Direct numerical integration of the reduced representation confirms the asymptotic behaviors quantitatively.

Dissipative coupling is known to induce synchronization. Conversely it may be hypothesized that oscillators driven to synchronize may reduce power dissipation in their coupling. The latter scenario is realized in the human cardiorespiratory system where cardiac and respiratory rhythms are controlled by the central nervous system while interacting viscoelastically through the pulmonary vasculature. Here we examine the functional significance of this coupling which is observed in respiratory sinus arrhythmia (RSA). By modelling electrical and viscoelastic interactions within the cardiorespiratory system, we identify the conditions leading to synchronization. We demonstrate that, when present, synchronization reduces cardiac power losses by 10% in humans and up to 55% in other species. The predicted gain in cardiac output is compared to the gain observed in-vivo by pacing the heart with a device restoring RSA. It is therefore surmised that RSA may improve cardiac pumping efficiency by reducing dynamic stress and power dissipation in the pulmonary vasculature.

Integrated opto-electronic devices have the potential to revolutionize information processing, with substantial increase in compute speed, seamless information transfer and reduction of energy consumption. A key missing unit for the successful implementation of compact functional devices are nanometer scale modular and tunable light sources. Monotonically grown semiconducting nanowire lasers (NWLs) fills this gap. However, NWLs operation improvement and optimization require the characterization of their near-field and its dynamics at the nanometer scale, which is hindered due to the light diffraction limit. Here we show how synchronous electron near-field and photon far-field time-resolved spectroscopies surpass this limitation and map a NWLs near-field with nanometer and sub-picoseconds temporal resolution. We quantitatively measured the evolution of the absolute number of stimulated photons $N_0(t)$ in the NWL cavity, measuring that up to 4x10$^5$ are present simultaneously in the cavity. We mapped the lasing cavity mode's near-field, showing that both whispering gallery and Fabry-Perot modes can participate in the lasing. Our results demonstrate how the near-field of a NWL under operation evolves in the sub-picoseconds and the nanometer scales. We anticipate that a direct observation of the near-field will help to elucidate the influence of materials heterogeneities (defects, chemical changes, contaminants, interface roughness, strain) in NWL operation.

Using smoothed particle hydrodynamics (SPH) simulations, we investigate the coefficient of restitution (COR) in wet collisions and identify a scaling law governing its behavior. The simulations employ an updated-Lagrangian, mesh-free framework that is validated against experimental measurements. We neglect surface tension effects since the impact conditions correspond to a moderate-to-high Weber number regime. The COR is found to depend on the Stokes number and a dimensionless film thickness defined as the ratio of the liquid film thickness to the diameter of the impacting solid bead. Two distinct regimes are observed, each characterized by different power-law exponents.

Chemical imaging enabled by Raman processes is crucial to investigating biological and chemical samples in a label-free manner. Stimulated Raman spectroscopy (SRS) overcomes the key limitation associated with low signal levels in spontaneous Raman spectroscopy, however, at the expense of probing only narrow Raman bands. Time-domain implementation of coherent anti-Stokes Raman spectroscopy (CARS) by dual frequency combs can achieve broad Raman bandwidths; nevertheless, its execution is demanding due to strenuous temporal-synchronization of two independent ultrashort laser sources. Here, we introduce time-domain coherent Raman spectroscopy utilizing two frequency combs generated by the Doppler effect from a single ultra-broadband laser source. In contrast to CARS, in our approach, the interference of impulsively launched vibrations by two broadband frequency combs (τ ~ 6 fs) periodically modulates the Kerr nonlinear response of the medium, leading to cross-phase modulation (XPM) experienced by both the combs. This phase modulation leads to spectral broadening and periodic modulation in the anti-Stokes region of the combs. Down-conversion by a factor of ~ 10-8 in the frequency of the vibrations enabled by the dual-comb approach empowered us to use photon-counting methodology in the anti-Stokes region. This makes our technique extremely versatile, background-free, sensitive and fast (millisecond acquisition times), in probing a range of samples from wide bandgap dielectrics and liquids to individual micro-particles with nondestructive pulse energies (~ 100 pJ) incident on the sample. Owing to the higher-order nonlinearity involved in the XPM process, we achieved ~ 2.5 times improvement in diffraction-limited spatial resolution (~ 280 nm) in ultra-broadband chemical imaging of a ~ 8 μm bead of poly-methyl-methacrylate.

Structured quantum light is crucial for high-dimensional quantum information processing, yet its direct generation from quantum emitters remains challenging due to their intrinsic locality and omnidirectional radiation. Metasurfaces have been adopted for quantum-light wavefront shaping, typically in cascaded or stacked configurations that suffer from low efficiency and limited resolution. Here, we demonstrate a semiconductor metasource that directly embodies single quantum dots in a nonlocal GaAs metasurface. Spontaneous emission from quantum dot is efficiently funneled into an extended quasi-bound-state-in-the-continuum mode while sustaining strong mode-emitter overlap. A lateral core-barrier heterostructure tunes mode volume and spatial distribution to balance Purcell enhancement and holographic resolution. Using spatially modulated geometric phase, our compact metasource enables deterministic generation of diverse single-photon radiation patterns, including orbital-angular-momentum beams and holographic images. Our work brings versatile single-photon wavefront control into the nanoscale cavity quantum electrodynamics regime, offering a scalable route toward integrated sources of structured quantum light.

High-entropy alloys (HEAs) and their two-dimensional counterparts (2D-HEAs) have recently attracted attention due to their tunable properties and catalytic potential, yet their chemical complexity makes direct density functional theory (DFT) calculations computationally prohibitive. The complexity also makes training of machine-learning interatomic potentials (MLIPs) challenging, but this could possibly be overcome by employing universal MLIPs as starting point. In this work, we investigate the applicability of universal MLIP models for 2D transition metal sulfide HEAs and develop effective fine-tuning strategies. Training structures are systematically generated and selected, and the performance of universal and fine-tuned models are benchmarked against DFT. We find that all universal MLIPs employed in this work yield unsatisfactory mixing energies without fine-tuning. Applied to the experimentally synthesized (Mo,Ta,Nb,W,V)S$_2$ system, fine-tuned models based on enumerated structures can achieve near-DFT accuracy in predicting mixing energies while enabling Monte-Carlo simulations and random structure sampling at scales inaccessible to DFT.

The Weak Equivalence Principle (WEP) is a central pillar of general relativity. Its precise test with quantum systems in space offers a unique window onto new physics. Here we report the first in-orbit quantum test of the WEP. A dual-species (85Rb/87Rb) atom interferometer is realized aboard the China Space Station. Methods of platform motion suppression, fluorescence detection switching, and two-photon detuning switching are developed to eliminate phase noise and improve measurement accuracy. A test uncertainty of 2.8*10-8 is obtained from 280 days of WEP test data, and a test result of (-3.1+/-4.6)*10-7 is achieved after error estimation. This improves prior atom-interferometric WEP tests in microgravity by three orders of magnitude. This work paves the way for space-borne quantum inertial sensors and their application to future fundamental physics in space.

The evolution of cooperation is a central enigma in evolutionary game theory. Traditionally, the combination of pairwise networks and repeated Public Goods Games with a single state fails to adequately describe realistic group interaction scenarios. On the one hand, pairwise networks lack clear group definitions. On the other hand, a participant's decision affects not only competitors' fitness but also the state of the surrounding environment. To address this problem, we propose a Public Goods Game with game transition mechanisms based on Uniform Random Hypergraphs. In our model, game groups formed by hyperedges transition between two types of games, one with abundant public resources and the other with scarce public resources. The transition probability is closely related to the strategies of players within the hyperedges. By developing a Monte Carlo simulation framework that incorporates payoff accumulation, strategy imitation, and game state transitions, we aim to reveal the coevolutionary patterns of strategies and game states in group interactions. Our study highlights a nonlinear relationship between defection sensitivity and cooperation frequency under game transitions, as well as the asymmetric effects of the two sensitivities in state-dependent transitions. These observations open new directions for how to approach social dilemmas.

Active frequency comb generation and waveform control are central challenges in the terahertz (THz) domain. In THz quantum cascade lasers (QCLs), these functions have typically been achieved through active bias modulation, which alters the operating point of the device and imposes severe limitations on its flexibility. To address these challenges, we propose an approach based on the direct modulation of the QCL bandstructure using GHz-frequency standing bulk acoustic waves (BAWs), promising direct and localized control of the optical gain and chromatic dispersion. To this end, we fabricated a bulk acoustic transducer on top of a THz QCL in order to excite GHz standing BAWs within its active region. We demonstrate that radio-frequency driving of the transducer leads to the tunable generation of standing BAWs in 5-12 GHz frequency range with wavelengths commensurate to the QCL period length. The effect of the BAW on the QCL bandstructure is revealed by measuring photoluminescence (PL) of the active region, where the BAW strain leads to a considerable modulation of the PL energy up to a few meV around its non-modulated value. We also develop a model and perform bandstructure simulations to predict the effect of the BAW on the QCL subband structure and gain. These results mark the first demonstration of dynamic bandstructure modulation in a THz QCL using GHz acoustic strain, introducing a fundamentally new paradigm that establishes a powerful synergy between QCLs and BAWs towards coherent control and frequency comb engineering in the THz domain.

The altermagnetism with antiparallel spin alignment exhibits anisotropic spin splitting and may possess an insulating state with a high Neel temperature, while the charge-order-induced ferroelectricity has ultrafast electric polarization switching. Considering that altermagnetism requires breaking space inversion,, the physical foundation for exploring ultrafast electrically controlled magnetism in altermagnetic ferroelectric materials is thus established. In this Letter, based on symmetry analysis and first-principles electronic structure calculations, we predict that LiV$_2$F$_6$ is a material that simultaneously hosts altermagnetism and charge-order-induced ferroelectricity. Since both the altermagnetism and ferroelectricity originate from charge order, LiV$_2$F$_6$ should exhibit strong magnetoelectric coupling. Our calculations indeed demonstrate that electric polarization reversal can induce band spin-polarization switching in LiV$_2$F$_6$. Moreover, time-dependent density functional theory calculations show that the electric polarization reversal in LiV$_2$F$_6$ occurs in 15 femtoseconds. Consequently, ultrafast electrically controlled magnetism can be realized in LiV$_2$F$_6$. Given that LiV$_2$F$_6$ has already been experimentally synthesized, our work provides a promising material platform for achieving ultrafast electrically controlled magnetism, which might have significant implications for the design of future electronic devices.

Two-dimensional altermagnets exhibit exceptional potential for low-power spintronics via nonrelativistic spin splitting and zero net magnetization. Here, we systematically investigate the influence of interlayer interactions on the electronic, magnetic and quantum transport properties of bilayer vanadium oxysulfide (V2S2O), a prototypical layered altermagnet, using DFT and NEGF calculations. Our results reveal that interlayer interactions predominantly modulate the p-orbital derived top valence bands, inducing a profound competitive valence band maximum position between Gamma-point pz and X/Y-point pxy orbitals, with an energy difference as small as 9 meV. Furthermore, interlayer interactions suppress the piezomagnetic effect and impose additional requirements on the type of strain for the bilayer system, compared to its monolayer counterpart. Out-of-plane external electric fields effectively weaken interlayer coupling by enlarging the energy difference of Gamma/X-Y top valence bands to 170 meV. Quantum transport simulations on a bilayer Au/V2S2O/Au two-probe device demonstrate the presence of pronounced spin current. Interlayer interactions reduce the transmission spin polarization from nearly 100% (monolayer) to 60% (bilayer) for energies above the Fermi level. Notably, gate-voltage modulation exhibits significant asymmetry in controlling charge-to-spin current conversion efficiency, originating from the out-of-plane symmetry breaking induced by the electrode geometry. Specifically, a positive gate voltage markedly enhances the contribution of the bottom layer to the overall spin polarization, while a negative gate voltage induces a marginal reduction of transmission spin polarization, attributed to the inherently weak polarization contribution of the bottom layer. These findings provide essential insights for the design and optimization of multilayer altermagnetic spintronics.

Wafer-to-wafer (WxW) bonding is a key enabler for three-dimensional integration, including hybrid bonding for fine-pitch Cu-Cu interconnects. During bonding, wafer deformation and the air entrapped between the wafers interact through a strongly coupled, time-dependent fluid-structure interaction (FSI) that can produce non-intuitive bonding dynamics and process sensitivities. This paper develops a mathematically consistent reduced-order model for WxW bonding by deriving a Kirchhoff-Love plate equation for wafer bending from three-dimensional linear elasticity and coupling it to a Reynolds lubrication equation for the inter-wafer air film. The resulting nonlinear plate-Reynolds system is discretized and solved monolithically in the high-performance FEniCSx framework using a $C^0$ interior-penalty formulation for the fourth-order plate operator, standard continuous Galerkin discretization for the pressure field, implicit time integration, and a Newton solver with automatic differentiation. Simulations reproduce experimentally reported probe-displacement histories for multiple initial gaps and verify force equilibrium at the bond front, where the Reynolds pressure acts as an effective contact reaction. Parametric studies reveal nonlinear, and in some cases non-monotonic, sensitivities of bonding-front kinetics to the initial gap, air viscosity, and interfacial energy, providing actionable trends for process optimization.

Ferroelectric materials are an ideal platform for high-speed reconfigurable photonic integrated circuits (PICs) for classical and quantum photonic computations, communications, and sensing. Most reconfigurable PIC devices achieve their functionalities via interference and are therefore highly sensitive to phase errors. Under static bias, carrier drift in ferroelectric waveguides induces continuous phase drift, creating a severe bottleneck for both PIC functionality and scalability. Here we propose achieving bias-stable and scalable ferroelectric PICs by exploiting the intrinsically low carrier drift of lithium tantalate (LT). Taking one of the PIC devices that is most sensitive to phase drift, the optical phased array (OPA), as an example, we designed and fabricated an integrated LT OPA that can keep the far-field main lobe 8 dB higher than side lobes for over 4 hours, representing at least a two-order-of-magnitude improvement over the state of the art. We demonstrated our device's capability in generating arbitrary spatiotemporal waveforms with a modulation frequency as low as 0.1 Hz, leading to practical applications in optical tweezers, trapped-ion quantum computers, adaptive optics for astronomy, AR, 3D printers, LiDAR, and free-space optical communications. Beyond OPA, our work establishes LT as a bias-stable, scalable, and high-speed PIC platform for large-scale classical and quantum photonic systems.

In complex engineering systems such as electro-thermal-fluid coupling, rapid and accurate prediction of multi-physics fields is essential for advanced applications like digital twins and real-time condition monitoring. Traditional numerical methods often suffer from high computational latency, whereas standard Physics-Informed Neural Networks (PINNs) frequently fail to capture critical local features, such as irregular interfaces, localized high-gradient regions, and multi-peak transport structures. To address these limitations and provide high-fidelity intelligent predictions for engineering decision-making, this paper proposes a Residual-Attention Physics-Informed Neural Network (RA-PINN) as a powerful surrogate modeling engine. The proposed method incorporates residual learning and attention enhancement into the network backbone to improve the representation of oblique transition structures, narrow charge layers, and distributed hotspots while strictly preserving global field consistency. To evaluate its effectiveness as an intelligent prediction framework, three representative benchmark cases are constructed, including an oblique asymmetric interface, a bipolar high-gradient charge layer, and a multi-peak Gaussian charge migration field. Under unified training settings, the proposed RA-PINN is systematically compared with a standard pure PINN and an LSTM-PINN in terms of average error, local maximum error, structural similarity, and convergence behavior. The results show that RA-PINN consistently achieves the best overall performance across all benchmark cases, demonstrating its tremendous potential as a highly reliable core inference engine for the condition monitoring and digital twin modeling of complex multi-physics engineering systems.

A concept of an innovative optical trap based on the retro-reflected standing-wave photon nanojet (SWOT) is presented. An open resonance cavity is formed between two coaxial microparticles of different geometries (sphere, cylinder, ring, truncated cone) with one particle docked to a plain mirror. Numerical simulations have shown the achievement of a record-high optical field intensity in the SWOT workspace, almost seven times higher than that of a conventional photonic nanojet trap due to a triple-focused optical beam, which contributes to improved optical capture. The proposed design of the optical trap allows for multi-position particle confinement in the trap area. The advantages of the proposed solution are the simple technical implementation and the possibility of integration with microfluidic technologies for optical manipulation of nanoobjects (Chip-on-flex optical sorting and ordering of nanoobjects, particle beaming).

We report a novel mechanism of boundary-sensitive PT symmetry breaking in one-dimensional Floquet systems. By designing a time-periodic driving protocol, we realize a Floquet Hamiltonian that is Hermitian under periodic boundary conditions yet acquires non-Hermitian boundary terms under open boundary conditions due to the non-commutativity of driving Hamiltonians. We establish that a PT symmetry breaking transition occurs when the quasienergy bandwidth expands to cover the entire frequency Brillouin zone. This condition highlights a crucial difference from static non-Hermitian systems, where such transitions typically require band touching. Furthermore, we demonstrate that in the PT-broken phase, the eigenstates exhibit scale-free localization, a phenomenon arising from the specific system-size scaling of non-Hermitian terms. Finally, we provide a general framework for constructing multi-band models that exhibit this boundary-induced phase transition.

Thermal interface materials (TIMs) critically regulate heat dissipation from electronic chips to heat spreaders, yet their thermal conductivity (k), volumetric heat capacity (C), and interfacial thermal resistance (ITR) evolve with mechanical pressure and cannot be determined simultaneously using existing steady-state or transient techniques. As a result, the coupled roles of bulk compaction and interfacial contact in governing heat transport in TIM assemblies remain poorly resolved. Here, we present a square-pulsed source (SPS) thermoreflectance method that enables simultaneous determination of k, C, and ITR in TIM stacks under controlled mechanical loading. By spanning square-wave modulation frequencies from 1 Hz to 10 MHz, SPS probes a broad range of thermal penetration depths, enabling distinction between heat diffusion in the TIM bulk and interfacial heat transfer at the Al/TIM contact. Measurements on a thermally conductive gel, a thermal pad, and a high-vacuum grease during compression-unloading cycles reveal distinct pressure-dependent thermal transport mechanisms. The gel and pad exhibit increases in k and C, reduced ITR, and pronounced hysteresis, indicating coupled bulk densification and persistent interfacial conformity during loading cycles. In contrast, the grease shows nearly pressure-independent bulk properties but a strong pressure dependence of ITR, consistent with an interface-dominated response. These results resolve the long-standing challenge of simultaneously quantifying bulk and interfacial thermal transport in mechanically loaded TIM assemblies, enabling experimentally constrained thermal management and reliability analysis in electronic packaging.

Solid-state spins in diamond are promising building blocks for quantum computing and quantum sensing, both of which require precise nanoscale addressing of individual spins. To explore the resolution limit of this approach, we demonstrate Fourier magnetic imaging of nitrogen-vacancy centers in diamond under state-of-the-art conditions. We constructed a highly compact experimental platform featuring thermal drift compensation under ambient conditions and generated a pulsed magnetic field gradient of up to 13.5 G/$μ$m. By implementing the Fourier magnetic imaging protocol, we achieved localization of a single nitrogen-vacancy center with a spatial resolution of 0.28 $\pm$ 0.10 nm and a magnetic field measurement deviation of 9 nT. This technique holds potential for applications such as localizing spins within proteins and cells.

Photonic systems provide a highly tunable platform for emulating quantum Hall physics. This tunability enables probing of the interplay between strong disorder and robust topological transport that remains difficult to access in solid-state systems. Here we realize a photonic version of the Harper-Hofstadter and Aubry-André models using a one-dimensional multilayer photonic crystal (Bragg stack) with a synthetic dimension encoded in its geometry. By modulating the layer thicknesses, we observe the Hofstadter butterfly and its chiral edge states from a family of one-dimensional multilayer structures, consistent with the Thouless pump picture. Exploiting the quasiperiodicity in this model, we show that increasing quasiperiodic modulation induces a wavelength-selective localization transition: specific Chern bands become fully localized along one dimension, while chiral edge states persist and continue to wind across the gap. We confirm this behavior through numerical simulations and experiments, and eigenmode analysis reveals that edge transport in this regime proceeds via a sequence of Landau-Zener transitions between localized states. These results demonstrate a crossover from adiabatic Thouless pumping under weak quasiperiodic modulation to a Landau-Zener-mediated topological pump at strong modulation, realized in a a compact and easily tunable photonic system.

Spin-wave-based physical reservoir computing (RC) is a promising candidate for energy-efficient physical implementations of artificial intelligence because of its potential for nanoscale integration with low power consumption. Most of the previous studies on spin-wave RC have utilized spin waves excited in a single-layer ferromagnet. In this study, we focused on spin waves in a synthetic antiferromagnet (SAF), consisting of two ferromagnetic layers coupled antiferromagnetically, and investigated additional memory properties of spin-wave RC. We theoretically and numerically demonstrate the emergence of two distinct memory properties in the SAF device due to the distinct spin-wave characteristics of the acoustic and optical modes inherent in SAFs.

Gausslets are one of the few examples of basis sets for electronic structure which allow for two-index/diagonal electron-electron interaction terms. A weakness of gausslets is that, because of their 1D origin, they have been tied to Cartesian coordinates. Here we generalize the gausslet construction for the radial coordinate in three dimensions for atomic basis sets. These radial gausslets make a very compact radial basis with a relatively modest number of functions, with diagonal interaction terms. We illustrate the accuracy of this construction with Hartree--Fock and exact diagonalization on atomic systems.

Magnetic resonance imaging (MRI) can see deep inside the body without ionizing radiation, but image quality depends strongly on how well the radio-frequency field is controlled. Passive resonant pads and metasurfaces can help, yet they often lose their tuning when they are placed next to water-rich tissue or tissue-like materials. Here we show a simple way to bring such a device back into tune. We built a bilayer metasurface made of two aluminum wire arrays. One layer can rotate relative to the other, and the gap between the two layers can also be adjusted. Bench measurements show that adding a controlled water load shifts the resonance to lower frequency by about \SIrange{4.2}{11.4}{\mega\hertz}. Rotating the layers shifts it back by about \SIrange{13.2}{14.9}{\mega\hertz}, which is much stronger than changing the gap alone. One loaded setting lands essentially at the proton frequency used in \SI{3}{\tesla} MRI. In a proof-of-concept scan on a clinical \SI{3}{\tesla} system, the metasurface made internal features in a structured pineapple phantom easier to see than in a substrate-only control. These results show that a passive MRI metasurface can be tuned after fabrication and retuned under load using geometry alone, opening a practical route to simple adjustable RF accessories for MRI.

Trapped ions are one of the most advanced platforms for quantum technologies, with applications ranging from quantum computing to precision timekeeping. A crucial step towards more compact and scalable systems involves integrating photonic integrated circuits (PICs) into surface ion traps to enable on-chip light delivery and optical addressing of individual ions. Currently, most implementations rely solely on transverse-electric (TE) mode grating couplers, where the emitted light is polarized in the plane of the chip. In this work, we design, fabricate and characterize silicon nitride (Si\(_3\)N\(_4\)) PIC components, including incoupling structures, splitters, and grating couplers that support both TE and transverse-magnetic (TM) modes with comparable optical losses. We benchmark the PIC at 760\,nm, which is a typical wavelength for Yb$^{+}$-applications. The fabricated grating couplers enable the outcoupling of collimated free-space beams for both polarizations, exhibiting distinct emission angles. This dual-polarization capability gives more flexibility in polarization control and expands the accessible optical design space for trapped-ion quantum technologies.

We introduce Biphoton Entanglement Light Spectroscopy (BELS), a quantum spectroscopic technique that employs polarization entangled Bell pairs and two photon interference to probe material properties. In BELS, the measured signal arises not from single photon intensities but from changes in the joint polarization and path correlations of biphoton Bell pairs transmitted through or scattered by a sample and analyzed via cross channel coincidences. A key concept of BELS is the explicit mapping between Jones matrix operations and transformations within the Bell state manifold. Optical elements that are equivalent under classical polarization optics can produce qualitatively distinct signatures in the coincidence landscape when interrogated with entangled photons. We demonstrate that linear birefringence and Faraday rotation generate orthogonal admixtures of Bell states, yielding experimentally distinguishable coincidence channels within a single measurement. We measure birefringence in an anisotropic dielectric and Faraday rotation in $\text{Tb}_3\text{Ga}_5\text{O}_{12}$. By mapping the changes to the photonic entanglement, BELS establishes a new framework for future entanglement enhanced spectroscopy, a potentially powerful approach in characterizing quantum materials, nanophotonic devices, and light matter interactions perhaps eventually at a fundamentally quantum level.

Absolute binding free energy (ABFE) calculations offer a theoretically rigorous approach for predicting protein--ligand binding affinities without the scaffold constraints of relative binding free energy (RBFE) perturbations. However, broad adoption of ABFE in high-throughput hit discovery campaigns has been hindered by high computational costs and a lack of large-scale validation. Here, we present Felis, an open-source, automated, and scalable toolkit designed for high-throughput ABFE calculations. Paired with ByteFF, a previously developed data-driven molecular mechanics force field for drug-like molecules, Felis achieves ranking performance comparable to state-of-the-art RBFE methods on a diverse dataset comprising 43 protein targets and 859 ligands. Furthermore, we demonstrate robust convergence and ranking performance of Felis on a more challenging KRAS(G12D) dataset, where some ligands and the cofactor are highly charged. Crucially, all Felis predictions in this study were generated in a strict zero-shot manner, eschewing custom force-field modifications and alchemical schedule fine-tuning. This demonstrates the viability of Felis as an effective, ready-to-use tool for computational structure-based drug design.

Plastic pollution has become a critical global challenge, with microplastics pervading ecosystems and entering human food chains. Effectively monitoring this widespread contamination demands rapid, reliable, and portable material identification techniques that often elude conventional Raman and FTIR spectroscopy. Undetected photon spectroscopy within a nonlinear interferometer (NLI) offers a solution, allowing the retrieval of mid-infrared absorption spectra by detecting only near-infrared signal photons using standard silicon-based technology. Here, we demonstrate a highly compact, micro-integrated, thermally-stabilised NLI with a Michelson-like geometry designed for the rapid spectroscopy of plastics. We benchmarked its room-temperature performance, demonstrating a signal-to-noise ratio of 34 with a measurement rate of 100 Hz and a spectral resolution of 6 cm$^{-1}$. We show that we can accurately and rapidly retrieve the characteristic vibrational absorption spectra of common polymers such as polypropylene, polyethene, and polystyrene, without using mid-infrared technology. These results establish our compact module as a promising field-deployable platform for robust, real-time environmental monitoring systems and other mid-infrared spectroscopy applications.

Many natural and social science systems are described using probability distributions over elements that are related to each other: for instance, occupations with shared skills or species with similar traits. Standard information theory quantities such as entropies and $f$-divergences treat elements interchangeably and are blind to the similarity structure. We introduce a family of divergences that are sensitive to the geometry of the underlying domain. By virtue of being the Bregman divergences of structure-aware entropies, they provide a framework that retains several advantages of Kullback-Leibler divergence and Shannon entropy. Structure-aware divergences recover planted patterns in a synthetic clustering task that conventional divergences miss and are orders of magnitude faster than optimal transport distances. We demonstrate their applicability in economic geography and ecology, where structure plays an important role. Modelling different notions of occupation relatedness yields qualitatively different regionalisations of their geographic distribution. Our methods also reproduce established insights into functional $β$-diversity in ecology obtained with optimal transport methods.

In this work, we theoretically propose a broadband asymmetric transmission (AT) device based on periodic Si nanoring arrays embedded in a SiO2 substrate. Results indicate that the device achieves a remarkable broadband AT effect in the near-infrared region (1750-2400 nm), with forward transmissivity exceeding 0.8 (maximum of 0.98), backward transmissivity less than 0.15 (minimum of 0.015) and an isolation ratio (IR) reaching a maximum of 17.8 dB at 2280 nm. Furthermore, the transmissivity spectrum exhibits excellent scalability and tunability through uniform scaling of the structure, allowing the operational band to be tailored across a wide spectral range, from 890 to 3300 nm. This Si-based nanostructure offers a robust and flexible platform for applications in optical isolation, multi-channel sensing, and integrated photonic circuits.

This study presents the first quantifications of ultrasound attenuation in oral soft tissues using validated standard techniques and serves as foundational step in advancing quantitative ultrasound (QUS) imaging in dentistry. Current standards of care in clinics for diagnosing periodontal diseases such as inflammation are limited by subjectivity, qualitive assessment, and late-stage indication. As a result, the application of ultrasonography is emerging as a surrogate for non-invasive and quantitative assessments and a relatively new research area with significant potential biomarkers to be explored. Many QUS analyses rely on quantifying ultrasound attenuation coefficient (UAC), as a confounding factor. Here, in a swine cohort (N=10), we characterized the high-frequency (24 MHz) UAC of healthy periodontal tissues (gingiva) in vivo. UAC were estimated using spectral difference method. Five interproximal oral sites were imaged from four oral quadrants: Premolar 3-Mesial, Premolar3-Distal, Premolar4-Distal, Molar1-Distal, and Molar2-Distal. A total of 162 oral sites were analyzed. The respective medians (1st-quartile|3rd-quartile) UACs for these oral sites were 1.66 (1.25|1.99), 1.37 (1.06|1.64), 0.99 (0.8|1.25), 1.08 (0.89|1.47), and 1.28 (0.94|1.24) dB/MHz.cm. The gingival attenuation mean at Premolar3-Mesial was significantly higher than any other oral sites while the rest of them showed non-significance difference in their means. Across all non-significant oral sites, the average UAC was 1.17 dB/MHz.cm with a standard deviation of 0.49 dB/MHz.cm. This work not only characterized an important acoustic property of oral tissues for the first time but also contributes to future development of a number of QUS biomarkers for periodontal/dental healthcare that rely on accurate attenuation knowledge.

Marathons are now common ways of seeing cities, yet little is known about how representative their routes are. Using 311 marathon routes across five continents, we compare landmarks and amenities along the course with those elsewhere in the same city, finding that museums are 15.7 times denser near the route and that the median city has about 8.5 times more luxury brands near the route than elsewhere in the city. These patterns persist under perturbed routes with the same start and finish lines: monuments and landmarks, in particular, are more prevalent on the race course than on similar alternative routes, suggesting that marathons function as intentionally selective urban portraits.

Optical skyrmions are members of the emerging topological branch of solid-state physics and photonics, allowing for control over topological light textures through light-matter interactions. However, in nanophotonics their practical application has been severely limited by high inherent losses in plasmonic materials, resulting in the lack of tunability between different topological properties. Here, we utilize the strong dispersion of silicon carbide thin films to realize highly confined surface phonon-polariton skyrmion lattices, which we image via near-field microscopy. We experimentally demonstrate topological tuning between bubble- and Néel-type skyrmions, a unique advantage that polar dielectrics offer over most existing approaches. Changing the excitation wavelength by only 10% switches the skyrmion type, revealed by examination of the skyrmion number density contrast. Analysis of domain wall size and steepness in analogy to magnetic materials also confirms this transition. Our results are a starting point to investigate other topological features in phononic systems such as merons, skyrmion bags, and other complex structured light fields. Furthermore, strong light-matter hybridization and nonlinear effects owing to anharmonicity of the phonons may be observed in the future, possibly leading towards the discovery of polaritonic skyrmion-skyrmion interactions and hence applications in topology-based information processing.

Gender disparities in academia manifest and persist in various aspects of the scientific enterprise, yet their influence on the interplay between research productivity and journal prestige remains underexplored. Here we analyze the academic trajectories of over 6,000 elite Brazilian researchers by jointly tracking their annual productivity and the average prestige of the journals in which they publish. By projecting individual career years onto a standardized productivity-prestige plane and applying Bayesian hierarchical modeling, we find that male researchers are more likely to follow productivity-oriented trajectories and are markedly overrepresented in the hyperprolific region of this plane. Female peers, in contrast, more often occupy regions that prioritize journal prestige over publication quantity. Although male researchers publish more throughout their careers, their female counterparts achieve comparable or higher average journal prestige, particularly in later career stages and among outlier individuals. Male researchers also exhibit greater temporal persistence in their productivity and impact levels and are especially averse to simultaneously changing both metrics compared to their female peers. Among non-outliers, productivity and career age have a negative overall impact on the average journal prestige of researchers of both genders, with slightly stronger effects observed among female researchers; however, these patterns vary across disciplines, highlighting the complexity and heterogeneity of academic careers.

The basic and effective reproduction numbers are widely used metrics for characterizing the dynamics of infectious disease epidemics. However, the interpretation of these numbers is based on the assumption of homogeneous mixing and may not hold in real-world populations where the contact patterns deviate from that assumption. In this paper, we present a network-based framework to compare reproduction numbers in populations with and without spatial structure, while other parameters of the disease remain fixed. Using this framework, we show that in homogeneously mixed populations, in the absence of external interventions, the effective reproduction number decreases exponentially as the susceptible population declines. In contrast, in spatially structured populations, the basic reproduction number is smaller, and the effective reproduction number initially decreases faster but eventually converges to unity. We show that the reproduction number is determined by the level of competition between infectious nodes, which is governed by the network structure. Our results suggest that without knowledge of the network structure, reproduction numbers may not be informative for parameterizing the contagiousness of the disease or predicting the behavior of epidemic spreading.

Here we present a density matrix based KS inversion method formulated entirely within a Gaussian basis representation to optimize a KS potential matrix that reproduces a target electron density. Inverse Kohn-Sham (KS) density functional theory (DFT) aims to determine the effective local KS potential that reproduces a target electron density, and is important both for electronic structure analysis and for the development of orbital based correction methods. In finite Gaussian basis implementations, however, conventional inverse KS-DFT approaches such as the Zhao-Morrison-Parr (ZMP) method often become poorly constrained and inefficient, because the real space penalty potential is projected onto a limited number of Gaussian basis matrix elements, which can strongly coarse-grain its spatial variation. In the present method, the density matrix mismatch is defined in a Lowdin orthogonalized basis, which yields a penalty energy invariant under unitary rotations in that basis. The corresponding penalty potential contribution to the KS Hamiltonian is derived analytically in the original nonorthogonal Gaussian basis. Across a wide range of penalty strengths, the self consistent field (SCF) optimization remains robust and efficient for various open shell systems, while progressively tightening the penalty drives the electron density into accurate agreement with the target. Benchmarks on molecules and condensed phase systems show that the method achieves substantially smaller attainable density deviations than the conventional ZMP method. The method provides a fast and accurate route to KS inversion in finite Gaussian basis sets and may also be useful for future orbital based correction schemes.

We introduce a low-cost every-$K$-step correction for one-dimensional Euler computations. The correction uses differentiated Riemann variables (DRVs) -- characteristic derivatives that isolate the left acoustic wave, the contact, and the right acoustic wave -- to locate the current wave packet, sample the surrounding constant states, perform a short Newton update for the intermediate pressure and contact speed, and conservatively remap a sharpened profile back onto the grid. The ingredients are elementary -- filtered centered differences, local state sampling, a single Newton step, and conservative cell averaging -- yet the effect on accuracy is disproportionate. On a long-time severe-expansion benchmark ($N=900$, $t=0.4$), intermittent correction drives the intermediate-state errors from $O(10^{-2})$ to $O(10^{-13})$, i.e. to machine precision. On a long-time LeBlanc benchmark ($N=800$, $t=1$), the method crosses a qualitative threshold: one-shot final-time reconstruction fails entirely (shock location error $2.7\times 10^{-1}$), whereas correction every three steps recovers an almost exact sharp solution with contact and shock positions accurate to machine precision. The same detector-and-Newton mechanism handles two-shock and two-rarefaction packets without case-specific logic, with plateau improvements of four to sixteen orders of magnitude. In an unoptimized Python prototype the wall-clock overhead is below a factor of two even on the most aggressively corrected benchmark. To our knowledge, no comparably lightweight fixed-grid add-on has been shown to recover this level of coarse-grid accuracy on the long-time LeBlanc and related near-vacuum problems.

Nanoscale control of optical dispersion is essential for applications ranging from miniaturized spectrometers to color printing, all of which demand broadband spectral tunability. However, the Kramers-Kronig relations impose a fundamental trade-off between dispersion and loss, strictly limiting the design ability of single-material devices across the deep ultraviolet (DUV) to near-infrared (NIR) regimes. Consequently, the fabrication of miniaturized dispersion devices heavily relies on costly nanofabrication or heterogeneous integration. Here we overcome these limitations by shifting the light-matter interaction from solid structure into air-filled voids. We introduce a fabrication strategy termed "Mie-lithography", in which laser printed seed nanocavities excite Mie resonances in air and the resulting localized field enhancement drives the self-assembly of three-dimensionally tunable void-type optical resonators. Because the resonant modes are primarily confined within air voids, this architecture effectively circumvents material-imposed dispersion-loss constraints, allowing on-demand customization of the broadband spectral response. This approach enables single-step, high-throughput (>= 10^6 pixels/s) printing of dispersion units with a resolution of 63,500 DPI. As a proof of concept, we demonstrate a DUV-NIR nano spectrometer integrated in a single material covering an unprecedented range from 200 nm to 800 nm. Our approach can be extended into a platform for ultra-broadband nano devices fabrication and design, opening avenues for high-pixel-density displays and miniaturized spectrometers.

Photon superbunching, distinguished by second-order coherence values far exceeding the Gaussian thermal limit, represents a highly desirable resource for quantum optics and correlation-based imaging technologies. However, existing approaches typically rely on fragile experimental platforms, inefficient nonlinear conversion processes, or mechanically complex optical architectures. Here, we demonstrate a fully fiber-integrated superbunched random fiber laser (SRFL) in which intrinsic Rayleigh scattering cooperatively interacts with cascaded stimulated Brillouin scattering and quasi-phase-matched four-wave mixing to tailor extreme photon statistics. The SRFL generates a multi-wavelength comb, in which individual spectral components exhibit widely tunable photon bunching, with the second-order coherence g(2)(0) continuously controlled from ~1 to ~26 by tuning the pump power, spectral order and diffusion length. Moreover, we establish a direct correlation between photonic phase transitions (quantified by the Parisi overlap order parameter) and the emergence of superbunching, thereby bridging macroscopic disorder physics and microscopic photon statistics. Finally, we employ the superbunched emission for temporal ghost imaging, realizing high-fidelity temporal object reconstruction with a substantial reduction in required ensemble averaging. These findings validate random fiber lasers as a robust, scalable, and integrated platform for generating extreme photon statistics and unlock new avenues for correlation-enhanced photonic sensing and quantum optics investigations in complex photonic systems.

Femtosecond laser micromachining (FLM) fabricated waveguides inherently form elliptical cores due to differences in focal spot size and the Rayleigh range of the microscope objective. Consequently, it is essential to study their propagation characteristics, which differ from those of conventional circular-core waveguides. In this work, we present the results of a parametric optimization of these waveguides to identify fabrication parameters that lead to minimal loss. A propagation loss characterization study revealed that, for a laser wavelength of 1030 nm, a pulse width of $\sim$300 fs, a pulse energy of 600 nJ, a scan speed of 2 mm/s, and a repetition rate of 100 kHz, a transparent and micro-bubble-free waveguide with a propagation loss of $\sim$0.4 dB/cm was formed. The modal analysis further demonstrated that the V-number depends on the core aspect ratio. The waveguide modes were compared with computationally generated modes, revealing a correlation that aligns well with existing literature.

Topological phase transitions provide a unique window into the interplay between structure, magnetism, and Weyl physics in magnetic Weyl semimetals. However, realizing an intrinsic Weyl phase transition between two distinct Weyl states near room temperature remains challenging. Here, we demonstrate that a magnetostructural transition effectively induces such a transition in the kagome magnet Mn$_3$Ga. High-resolution neutron diffraction, magnetization characterizations and first-principles calculations reveal that Mn$_3$Ga undergoes a chiral antiferromagnetic transition below 485 K, followed by a magnetostructural transition to a monoclinic structure with highly canted antiferromagnetic order near room temperature. These cooperative changes in lattice and magnetic symmetries reorganize Weyl nodes, driving a transition from a primary type-II Weyl state to a distinct Weyl state, accompanied by dramatic variations in the anomalous Hall effect and appearance of topological Hall effect. Our findings open a new pathway for discovering novel topological Weyl states and potential spintronic applications.

Copper fire-through metallization is a cost-effective alternative to Ag counterpart for industrial high efficiency solar cells. The fire through dielectric metallization relies on belt speed, which dictates the ramp up and ramp down rates for effective contact formation. In this paper three belt speeds (325oC, 360oC, 390oC) at constant peak firing temperature, were used to process PERC (homogeneous emitter) cells. After the contact firing the electrical parameters were dependent on belt speed, but after LECO treatment, they were identical. The SEM/EDS cross sectional analyses showed increased elemental Cu with belt speed, and the series resistance was lowest for the middle belt speed before LECO. However, after the LECO treatment, the series resistance dropped, respectively, to 0.503 ohm-cm-2, 0.428 ohm-cm-2 and 0.500 ohm-cm-2 leading to efficiency of 20.8% on homogeneous PERC emitter.

This proof-of-concept study introduces a novel multimodal framework combining synchronized EEG-fNIRS modalities with neuronal avalanche analysis to identify early network dysfunction in Alzheimer's disease. The approach leverages complementary neural signals to examine motor network dynamics during execution and imagery tasks within an interactive task environment. Preliminary analysis of a small pilot cohort (N=4 subjects, including one with Mild Cognitive Impairment) validated the technical feasibility of the multimodal framework and revealed observable condition-dependent patterns in network organization. Two primary observations emerged: a reduced neural contrast between motor execution and imagery states, and increased trial-to-trial variability in network organization in the MCI participant. These initial results successfully validate the technical pipeline and provide hypothesis-generating observations for future statistically powered studies. The convergence of findings across modalities suggests that multimodal assessment of network flexibility may help detect functional changes in early Alzheimer's continuum, supporting the future development of this BCI-inspired framework into an engaging diagnostic tool.

The disjunction effect in human decision making is often taken to show that the classical law of total probability is violated, motivating quantum-like models. We re-examine this claim for the Prisoner's Dilemma disjunction effect. Under the mental-event reading of the opponent-choice events, the conventional classical decision-process model implicitly builds in a certainty-only premise: its standard partition assumptions leave no room for ambiguity, forcing every participant to be certain that the opponent will defect or will cooperate. We relax this by introducing a new classical model in which each participant carries a continuous expectation parameter representing the anticipated likelihood of opponent defection, and the participant pool is partitioned by expectation level; the resulting ambiguity set is precisely the union of the interior expectation bins. In contrast, under the quantum-like event semantics, ambiguous pure states are generic (dense and of full unitarily invariant measure on the unit sphere), so "certainty states" are mathematically exceptional. We prove that an instance of our classical model can realize any empirically observed triple of defection rates across the three information conditions, including strong disjunction-effect patterns, while strictly obeying the classical law of total probability. We further prove that for any such triple produced by a standard quantum-like model of the same experiment, there exists a classical instance reproducing it exactly. In this sense, classical and quantum-like approaches have the same observable-rate expressiveness; their substantive difference lies in how ambiguity is represented and in their respective event semantics, not in a breakdown of classical probability.

Alzheimer's disease (AD) develops over a prolonged preclinical phase, during which neuropathological changes accumulate long before cognitive symptoms appear. Identifying cognitive functions affected at early stages is critical for the preclinical detection of asymptomatic individuals at-risk of AD. Early risk identification could enable timely interventions aimed at mitigating the development of significant future cognitive impairment. While episodic memory decline typically appears after substantial medial temporal lobe damage, spatial navigation has emerged as a particularly sensitive cognitive function in preclinical AD. In this review, we provide an overview of spatial navigation computations and the tasks used to assess them, highlighting how spatial navigation relies on neural circuits corresponding to the earliest sites of AD pathology. We synthesize evidence from cognitively unimpaired individuals with AD biomarkers, i.e. individuals at-risk of AD, and discuss future research directions. Overall, performance on spatial navigation tasks, particularly path integration and wayfinding, correlates with plasma and CSF biomarkers of AD pathology, notably p-tau. Spatial navigation assessment can represent a sensitive and scalable approach for early detection of individuals at-risk of AD in preclinical stages, and will inform future interventions to mitigate the progression toward clinically significant cognitive impairment.

The incorporation of neuroimaging techniques such as electroenchephalography (EEG) and functional near-infrared spectroscopy (fNIRS) has provided new opportunities for the analysis of dynamic brain processes involved in cognitive and motor functions. Despite the great contribution of the open-access neuroimaging datasets to neuroscience studies, they have mainly remained on a single modality and isolated task paradigms performed in a controlled environments. These limitations restrict the analysis of multi-task effects in real-world applications, thus creating a gap in the understanding of how cognitive and motor processes interact in daily life activities. To address these limitations, we present a multi-modal dataset containing neurophysiological (EEG, fNIRS), physiological (ECG), behavioral, and subjective measures collected from 30 healthy participants over three sessions. This dataset includes a hierarchical series of seven tasks ranging from single cognitive and motor activities, such as N-back, motor, passive motor, mental arithmetic and motor imagery, to combined cognitive-motor interactions simulating real life scenarios. This raw dataset provides a resource for developing advanced preprocessing methods and analysis pipelines, with potential applications in brain-computer interfaces, neurorehabilitation, and other fields requiring an understanding of multi-tasks brain dynamics. https://doi.org/10.18112/openneuro.ds007554.v1.0.0

Background: Autism spectrum disorder (ASD) is characterized by significant clinical and biological heterogeneity. Conventional group-mean analyses of eye movements often mask individual atypicalities, potentially overlooking critical pathological signatures. This study aimed to identify idiosyncratic oculomotor patterns in ASD using an "outlier analysis" of smooth pursuit eye movement (SPEM). Methods: We recorded SPEM during a slow Lissajous pursuit task in 18 adults with ASD and 39 typically developed (TD) individuals. To quantify individual deviations, we derived an "outlier score" based on the Mahalanobis distance. This score was calculated from a feature vector, optimized via Principal Component Analysis (PCA), comprising the temporal lag ($Δ$t) and the spatial deviation ($Δ$s). An outlier was statistically defined as a score exceeding $\sqrt{10}$ (approximately 3.16$σ$) relative to the TD normative distribution. Results: While the TD group exhibited a low outlier rate of 5.1\%, the ASD group demonstrated a significantly higher prevalence of 38.9\% (7/18) (binomial P = 0.0034). Furthermore, the mean outlier score was significantly elevated in the ASD group (3.00 $\pm$ 2.62) compared to the TD group (1.52 $\pm$ 0.80; P = 0.002). Notably, these extreme deviations were captured even when conventional mean-based comparisons showed limited sensitivity. Conclusions: Our outlier analysis successfully visualized the high degree of idiosyncratic atypicality in ASD oculomotor control. By shifting the focus from group averages to individual deviations, this approach provides a sensitive metric for capturing the inherent heterogeneity of ASD, offering a potential baseline for identifying clinical subtypes.

Musculoskeletal injuries (MSI) in Thoroughbred racehorses are a leading cause of death and premature retirement in racehorses and are heavily influenced by training practices. Greater distances of high-speed galloping accumulated during racing campaigns are associated with MSI. Bone injury is the most common MSI, and understanding how training practices influence bone damage accumulation is critical for improving both horse welfare and racing outcomes. This study builds on an existing mathematical model of bone adaptation and damage to investigate the impact of different training programs on bone injury risk. Several training programs (three progressive, four race-fit, six rest programs and two with rest replaced by low-intensity training) were constructed to reflect representative practices undertaken by professional trainers in Victoria, Australia. Training programs varied in training volume, rest frequency and program duration. Lower volume training programs that included high-speed training, achieved sufficient bone adaptation with less accumulation of bone damage, and subsequently lower risk of bone failure. In addition, incorporating more frequent rests (at least 2 per year) and/or longer rest periods (at least 6 weeks) reduced bone damage due to the extended opportunity to remove and repair bone damage. These results provide an in-silico mathematical model of the bone response to training, demonstrating the effects of training programs on bone adaptation, damage formation and repair. The findings can guide the design of training programs that balance both bone adaptation and bone health throughout horses racing career.

We consider the SEIRS epidemiology model with such features of the COVID-19 outbreak as: abundance of unidentified infected individuals, limited time of immunity and a possibility of vaccination. The control of the pandemic dynamics is possible by restricting the transmission rate, increasing identification and isolation rate of infected individuals, and via vaccination. For the compartmental version of this model, we found stable disease-free and endemic stationary states. The basic reproductive number is analysed with respect to balancing quarantine and vaccination measures. The positions and heights of the first peak of outbreak are obtained numerically and fitted to simple in usage algebraic forms. Lattice-based realization of this model is studied by means of the asynchronous cellular automaton algorithm. This permitted to study the effect of social distancing by varying the neighbourhood size of the model. The attempt is made to match the quarantine and vaccination effects.

Spatial transcriptomics data analysis integrates cellular transcriptional activity with spatial coordinates to identify spatial domains, infer cell-type dynamics, and characterize gene expression patterns within tissues. Despite recent advances, significant challenges remain, including the treatment of batch effects, the handling of mixed cell-type signals, and the imputation of poorly measured or missing gene expression. This work addresses these challenges by introducing a novel Subspace Tensor Orthogonal Rotation Model (STORM) that aligns multiple slices which vary in their spatial dimensions and geometry by considering them at the level of physical patterns or microenvironments. To this end, STORM presents an irregular tensor factorization technique for decomposing a collection of gene expression matrices and integrating them into a shared latent space for downstream analysis. In contrast to black-box deep learning approaches, the proposed model is inherently interpretable. Numerical experiments demonstrate state-of-the-art performance in vertical and horizontal batch integration, cell-type deconvolution, and unmeasured gene imputation for spatial transcriptomics data.

Recent biological evidence suggests the presence of a two-phase ageing process in several species. We introduce a system of two age-structured partial differential equations (PDE) representing two phases of ageing of a wild population. The model includes a coupling of both equations through birth and transition between phases and non-linearities due to competition. We show the existence, positivity and uniqueness of weak solutions in a general setting. For a simplified system of ordinary differential equations (ODE), we show existence and uniqueness of a strictly positive steady state attracting all trajectories. We study another simplification, a coupled PDE-ODE model, for which we prove existence, uniqueness and local asymptotic stability of a strictly positive steady state. Under further assumptions, but without assuming weak non-linearities, we show the global asymptotic stability of that steady state. The uniqueness of steady states and absence of oscillations in these systems show that the proportion of individuals in each phase at equilibrium is a unique feature of the model. This paves the way to ecological applications as the experimental measure of such a proportion could help gain some insight on the health of a wild population.

Traditional Direction of Arrival (DOA) estimation methods struggle to simultaneously address three physical constraints in Ultra-Wideband (UWB) electromagnetic sensing: spatial undersampling, asynchronous array phase, and beam squint. Existing solutions treat these issues in isolation, leading to limited performance in complex scenarios. This paper proposes a novel dynamic manifold perspective, which models UWB signal observations as a continuous manifold curve in a high-dimensional space driven by temporal evolution and array topology. We theoretically demonstrate that the DOA can be uniquely determined solely by the geometric shape of the manifold, rather than the absolute arrival phase. Based on this perspective, we construct a geometric parameter system comprising extrinsic and intrinsic parameters, along with a corresponding DOA estimation framework. Extrinsic vector parameters serve as a dynamic extension of traditional array processing, effectively expanding the degrees of freedom to suppress grating lobes. Intrinsic scalar invariants provide a new geometric perspective independent of traditional phase models, offering intrinsic robustness against array channel phase errors. Simulation results show that the derived analytical expressions for geometric parameters are highly consistent with numerical truths. The proposed framework not only completely eliminates spatial ambiguity in sparse arrays but also achieves high-precision direction finding under conditions with calibration-free phase errors.

The growing demand for higher data rates necessitates continuous innovations in wireless communication systems, particularly with the emergence of 6G. Channel coding plays a crucial role in this evolution. In 5G systems, rate-adaptive raptor-like quasi-cyclic irregular low-density parity-check codes are used for the data link, while polar codes with successive cancellation list decoding handle short messages on the synchronization channel. However, to meet the stringent requirements of future 6G systems, a versatile and unified coding scheme should be developed - one that offers competitive error-correcting performance alongside low complexity encoding and decoding schemes that enable energy-efficient hardware implementations. This white paper outlines the vision for such a unified coding scheme. We explore various 6G communication scenarios that pose new challenges to channel coding and provide a first analysis of potential solutions.

Near-field integrated sensing and communication (ISAC) enables object-level sensing from distance-dependent array responses, yet most existing near-field methods still rely on point-target models and realistic extended targets remain largely unexplored. In this paper, joint target classification and range-azimuth localization are studied from channel responses of realistic extended targets. A dual-branch inference framework is proposed. Semantic and geometric branches are used for classification and localization, respectively. Cross-task attention is introduced after task-specific encoding so that complementary cues can be exchanged without forcing full feature sharing from the input stage. To improve localization on the same backbone, uncertainty-aware regression and a physics-guided structured objective are adopted, including planar consistency, peak-response regularization, and geometry-coupling constraints. Training and evaluation data are generated from full-wave electromagnetic scattering simulations of voxelized vehicle targets with randomized heading angles, material contrasts, and placements. The compared variants show that cross-task attention mainly benefits classification, while uncertainty-aware and structured supervision are needed to recover strong localization performance on the same backbone. Under the adopted shared-OFDM benchmark, the proposed framework reaches the best joint operating point with fewer sensing tones for the same target performance region.

We derive fundamental accuracy limits for distributed localization when a fusion center has access only to independently rate-distortion (RD)-optimally compressed versions of multi-sensor observations, under a line-of-sight propagation model with a Gaussian wideband waveform. Using the Gaussian RD test-channel model together with a Whittle spectral Fisher-information characterization, we obtain an explicit frequency-domain Cramér-Rao lower bound. A two-band, two-level specialization yields closed-form expressions and reveals a rate-induced regime change: RD-optimal compression under a squared-error distortion measure can eliminate localization-informative spectral content. A simple band-selective scheme can outperform RD compression by orders of magnitude at the same rate, motivating localization-aware compression for networked sensing and integrated sensing and communication systems.

Nonlinear extensions of the Kalman filter (KF), such as the extended Kalman filter (EKF) and the unscented Kalman filter (UKF), are indispensable for state estimation in complex dynamical systems, yet the conditions for a nonlinear KF to provide robust and accurate estimations remain poorly understood. This work proposes a theoretical framework that identifies the causes of failure and success in certain nonlinear KFs and establishes guidelines for their improvement. Central to our framework is the concept of covariance compensation: the deviation between the covariance predicted by a nonlinear KF and that of the EKF. With this definition and detailed theoretical analysis, we derive three design guidelines for nonlinear KFs: (i) invariance under orthogonal transformations, (ii) sufficient covariance compensation beyond the EKF baseline, and (iii) selection of compensation magnitude that favors underconfidence. Both theoretical analysis and empirical validation confirm that adherence to these principles significantly improves estimation accuracy, whereas fixed parameter choices commonly adopted in the literature are often suboptimal. The codes and the proofs for all the theorems in this paper are available at https://github.com/Shida-Jiang/Guidelines-for-Nonlinear-Kalman-Filters.

Diffusion models (DMs) have achieved remarkable success across various domains owing to their strong generative and denoising capabilities. Meanwhile, semantic communication based on neural joint source-channel coding (JSCC) has emerged as a promising paradigm for robust and efficient image transmission. However, severe channel noise can still distort the transmitted semantic symbols, resulting in significant performance degradation. Applying DMs to digital semantic symbols, particularly in vector quantization (VQ)-based systems, is fundamentally challenging because the Markov assumption does not hold for the symbol transition dynamics. To address this issue, we introduce SSCDM, a semantic symbol correcting diffusion model whose discrete-time transition dynamics are constructed using solutions from continuous-time Markov chain theory. Furthermore, to promote synergy between DMs and JSCC, our DM structure embeds discrete symbols into a latent feature space using a learned VQ codebook, and a self-organizing map-based loss is incorporated during codebook learning to enhance the geometric vicinity between neighboring digital symbols, thereby promoting topology-preserving semantic representations. Experimental results show that the proposed method significantly improves image reconstruction quality and outperforms previous symbol-level denoising techniques under low signal-to-noise ratio scenarios and different datasets.

Wireless systems are expanding their purposes, from merely connecting humans and things to connecting intelligence and opportunistically sensing of the environment through radio-frequency signals. In this paper, we introduce the concept of triple-functional networks in which the same infrastructure and resources are shared for integrated sensing, communications, and (edge) Artificial Intelligence (AI) inference. This concept opens up several opportunities, such as devising non-orthogonal resource deployment and power consumption to concurrently update multiple services, but also challenges related to resource management and signaling cross-talk, among others. The core idea of this work is that computation-related aspects, including computing resources and AI models availability, should be explicitly considered when taking resource allocation decisions, to address the conflicting goals of the services coexistence. After showing the natural coupling between theoretical performance bounds of the three services, we formulate a service coexistence optimization problem that is solved optimally, and showcase the advantages against a disjoint allocation strategy.

First-order Ambisonics (FOA) is a standard spatial audio format based on spherical harmonic decomposition. Its zeroth- and first-order components capture the sound pressure and particle velocity, respectively. Recently, physics-informed neural networks have been applied to the spatial interpolation of FOA signals, regularizing the network outputs based on soft penalty terms derived from physical principles, e.g., the linearized momentum equation. In this paper, we reformulate the task so that the predicted FOA signal automatically satisfies the linearized momentum equation. Our network approximates a scalar function called velocity potential, rather than the FOA signal itself. Then, the FOA signal can be readily recovered through the partial derivatives of the velocity potential with respect to the network inputs (i.e., time and microphone position) according to physics of sound propagation. By deriving the four channels of FOA from the single-channel velocity potential, the reconstructed signal follows the physical principle at any time and position by construction. Experimental results on room impulse response reconstruction confirm the effectiveness of the proposed framework.

Future wireless networks are expected to support increasingly high data rates and user densities, motivating advanced multi-antenna architectures capable of adapting to dynamic propagation environments. Movable antenna (MA) arrays have recently emerged as an extension of massive MIMO, enabling physical repositioning of antenna elements to better exploit spatial diversity and mitigate inter-user interference. While prior studies report promising gains under idealized assumptions, their performance under realistic wideband multi-user operation remains insufficiently understood. This paper presents a comprehensive evaluation of MA-enabled systems in practical uplink and downlink scenarios. A wideband OFDM system model is developed, and novel closed-form sum-rate expressions are derived for both uplink and downlink under linear and nonlinear processing. Hardware impairments are incorporated via an EVM-based model, from which a distortion-aware UL/DL duality is established and the resulting high-SNR sum rate ceiling is analytically characterized. In addition, the interactions between antenna position optimization, receiver processing, and user loading are examined, and performance is evaluated under both time-division duplexing (TDD) and frequency-division duplexing (FDD). The results show that movable antennas can provide noticeable gains in low-impairment regimes with strong multi-user interference, but these benefits are highly scenario-dependent and diminish under hardware-impairment-limited conditions or in rich-scattering environments. These findings highlight the importance of carefully assessing deployment conditions when considering antenna mobility as an alternative to conventional fixed array configurations.

We present a compressive beamforming method for coherent plane-wave compounding (CPWC) ultrasound imaging based on a far-field decomposition of the received radiofrequency (RF) data into virtual plane waves. This decomposition recasts the imaging operation entirely in the spatial frequency domain ($k$-space), allowing direct and flexible control over $k$-space sampling distributions based on the principle of coarrays. We present vernier-type sampling strategies designed to optimize the tradeoff between image contrast and resolution with minimum redundancy, including strategies that favor dense low-frequency sampling for high contrast, shifted schemes that extend the frequency support for improved resolution, and confocal or hybrid compounding schemes that approximate the spatial-frequency transfer function of conventional DAS beamforming. Our method, called KK beamforming, is validated with a calibration phantom and in-vivo human tissue data, demonstrating compression factors of an order of magnitude while maintaining image qualities comparable to conventional DAS. We further demonstrate that KK beamforming yields improvements in computational speed owing to its reduced memory footprint and more efficient cache utilization of the compressed data and associated look-up tables.

We develop a pragmatic multi-user (MU) massive multiple-input multiple-output (MIMO) channel model tailored to the THz band, encompassing factors such as molecular absorption, reflection losses and multipath diffused ray components. Next, we propose a novel semi-blind based channel state information (CSI) acquisition technique i.e. MU whitening decorrelation semi-blind (MU-WD-SB) that exploits the second order statistics corresponding to the unknown data symbols along with pilot vectors. A constrained Cramer-Rao Lower Bound (C-CRLB) is derived to bound the normalized mean square error (NMSE) performance of the proposed semi-blind learning technique. Our proposed scheme efficiently reduces the training overheads while enhancing the overall accuracy of the channel learning process. Furthermore, a novel hybrid receiver combiner framework is devised for MU THz massive MIMO systems, leveraging multiple measurement vector based sparse Bayesian learning (MMV-SBL) that relies on the estimated CSI acquired through our proposed semi-blind technique relying on low resolution analog-to-digital converters (ADCs). Finally, we propose an optimal hybrid combiner based on MMV-SBL, which directly reduces the MU interference. Extensive simulations are conducted to evaluate the performance gain of the proposed MU-WD-SB scheme over conventional training-based and other semi-blind learning techniques for a practical THz channel obtained from the high-resolution transmission (HITRAN) database. The metrics considered for quantifying the improvements include the NMSE, bit error rate (BER) and spectral-efficiency (SE).

Digital self-interference cancellation (D-SIC) plays a crucial role in in-band full-duplex radios. Unfortunately, its fundamental limit remains unclear. In this paper, we aim to address this problem by exploring the performance limit of the parallel Hammerstein (PH) canceller for D-SIC, which is most commonly used in practice. First, a comprehensive analysis of the power of the residual self-interference (RSI) after the PH canceller with the least squares (LS) estimator is provided, which takes into account the truncation error, reconstruction error and transmitter noise. Specifically, the analysis is greatly simplified by equivalently expanding the PH canceller via generalized Laguerre polynomials (GLP), which enjoys the desirable property of mutual orthogonality among the basis functions. As a by-product of this orthogonal expansion, we establish that the LS estimator for the weights of the GLP canceller is asymptotically \textit{unbiased}, if the pilot sequence is Gaussian distributed. Second, in order to minimize the reconstruction error of the PH canceller, we propose a succinct criterion for optimizing the pilot sequence, which essentially seeks for small eigenvalue spread and large minimum eigenvalue of the Gram matrix corresponding to the pilot sequence. Specifically, the criterion is to minimize the product of the Shannon rank, an effective rank of a positive semidefinite matrix and the minimum eigenvalue of the Gram matrix. Simulation results demonstrate that with the optimized pilot sequence of a single OFDM symbol, over 10 dB gain can be achieved compared to the conventional pilot sequence (HE-LTF) for the PH canceller, and the corresponding RSI can be as low as -87.6 dBm.

Due to the directive property of each antenna element, the received signal power can be severely attenuated when the emitter deviates from the array boresight, which will lead to a severe degradation in sensing performance along the corresponding direction. Although existing rotatable array sensing methods such as recursive rotation (RR-Root-MUSIC) can mitigate this issue by iteratively rotating and sensing, several mechanical rotations and repeated eigendecomposition operations are required to yield a high computational complexity and low time-efficiency. To address this problem, a pre-rotation initialization with recieve power as a rule is proposed to signifcantly reduce the computational complexity and improve the time-efficiency. Using this idea, a low-complexity enhanced direction-sensing framework with pre-rotation initialization and iterative greedy spatial-spectrum search (PRI-IGSS) is develped with three stages: (1) the normal vector of array is rotated to a set of candidates to find the opimal direction with the maximum sensing energy with the corresponding DOA value computed by the Root-MUSIC algorithm; (2) the array is mechanically rotated to the initial estimated direction and kept fixed; (3) an iterative greedy spatial-spectrum search or recieving beamforming method, moviated by reinforcement learning, is designed with a reduced search range and making a summation of all previous sampling variance matrices and the current one is adopted to provide an increasiong performance gain as the iteration process continues. To assess the performance of the proposed method, the corresponding CRLB is derived with a simplified rotation model. Simulation results demonstrate that the proposed PRI-IGSS method performs much better than RR-Root-MUSIC and achieves the CRLB in term of mean squared error due to the fact there is no sample accumulation for the latter.

We show that to every small and decaying solution of the linearized constraint equations about Minkowski spacetime, one can add a quadratically small correction to obtain a solution of the full constraint equations. Near spacelike infinity, the correction is given by Kerr black hole initial data, up to a term that decays faster than the linearized solution, and that has Schwartz decay if the linearized solution has Schwartz decay. Using a recent result, we obtain that the solutions of the Einstein equations with these initial data admit a regular conformal compactification along null and timelike infinity. The construction is based on a right inverse (up to necessary integrability conditions) for the linearized constraint operator about Minkowski initial data obtained previously, that has optimal mapping properties relative to weighted b-Sobolev spaces, where the weights measure decay towards infinity. On an algebraic level, we show that the constraint equations can be derived using the homotopy transfer theorem, rather than using the geometric Gauss and Codazzi equations.

We propose the Locally Pumped Dark Energy (LPDE) mechanism in which cosmic acceleration is triggered by the emergence of non-linear dark matter structure. In an effective-field-theory description, coarse-graining over the density contrast profile, whose short-wavelength modes grow during halo formation, induces a shift in the local equilibrium point of a second, sufficiently heavy scalar field $χ$. At early times, the pump mechanism is negligible and $χ$ remains fixed at the origin, contributing no DE. As structures form, the equilibrium value of $χ$ is locally displaced within halos, generating a vacuum energy whose global contribution, in a mean-field picture, is controlled by the halo volume filling factor. If the $χ$ field is sufficiently heavy, with a Compton wavelength limited by halo scales, its response is localised, and spatial gradients are exponentially suppressed on large scales. After volume-averaging over the halo population, the resulting contribution on large scales behaves as a homogeneous DE component. Using the halo mass function of a fiducial $Λ$CDM cosmology, we show that vacuum-energy domination generically emerges at $z\sim\mathcal{O}(1)$, naturally correlating cosmic acceleration with structure formation. For reference, we present an explicit realisation of such a mechanism and show that, by naturally featuring a transient acceleration epoch, it can be in excellent agreement with the most recent cosmological data, including the Dark Energy Spectroscopic Instrument (DESI).

We establish an information-theoretic scaling law for generic autoregressive neural quantum states, determined by the middle-cut mutual information of the wavefunction amplitude. By formalizing the virtual bond as an effective information channel across a sequence bipartition, we rigorously prove that exact autoregressive representation of a quantum state requires the virtual-bond dimension to scale with the amplitude mutual information. For stabilizer-state families, we show that this law yields an explicit, analytical rank formula. Applying this framework across quantum-state tomography, ground-state and finite-temperature learning, our numerical experiments expose precise exponent matching, architecture-dependent scaling differences between recurrent and Transformer neural quantum state, and the critical role of autoregressive basis ordering. These results establish a rigorous physical link between the intrinsic structure of a quantum many-body state and the corresponding neural-network capacity required for its faithful representation.

In this note we consider no-boundary instantons and wine-glass geometries which are of interest in the context of quantum cosmology. While the former appears as a dominant saddle in the path-integral, the later sub-dominant saddle has been argued to have a longer inflationary phase of the Universe. Kontsevich-Segal-Witten (KSW)-allowability criterion which classifies geometries on the basis of the requirement of having a meaningful QFT on it, pushes one to analyse the allowability of the saddles. In this note we do a simple study to seek answer to the allowabilty of no-boundary instantons and wine-glass geometries which are of interest in quantum cosmology. Our simple analysis which make use of a milder version of KSW allowability criterion shows that no-boundary instanton is KSW allowed while wine-glass geometries are KSW disallowed.

Today's most accurate clocks are based on laser spectroscopy of electronic transitions in single trapped ions and feature fractional frequency uncertainties below $1\times10^{-18}$. Scaling these systems to multiple, simultaneously interrogated ions reduces measurement times, driving recent advances in multi-ion clocks. However, maintaining state-of-the-art systematic uncertainties while increasing the number of ions remains a central challenge. Here, we report on a multi-ion optical atomic clock with a fractional frequency uncertainty of $5.3\times10^{-19}$ and up to 10 \Sr ions. Ion-resolved state detection enables minimization of position-dependent shifts, with residual effects suppressed below the $10^{-20}$-level. Clock operation with eight to ten ions reduces the measurement time by a factor of 4.8 compared to single-ion operation. A comparison with an established \Yb single-ion clock yields an unperturbed frequency ratio of $0.6926711632159660405(20)$, with a statistical uncertainty of $0.9\times10^{-18}$ and a combined uncertainty of $2.9\times 10^{-18}$. These results demonstrate robust multi-ion clock operation with reduced averaging time and state-of-the-art accuracy.

We introduce the Adaptive Derivative-Assembled Pseudo-Trotter ansatz Variational Majorana Propagation Eigensolver (ADAPT-VMPE), a quantum-inspired classical algorithm that exploits Majorana Propagation (MP) to produce circuits for approximating the ground state of molecular Hamiltonians. Equipped with the theoretical guarantees of MP, which provide controllable bounds on the approximation error, ADAPT-VMPE offers an efficient and scalable approach for iterative ansatz construction. A theoretical analysis of the computational complexity demonstrates that it is polynomial in both the number of qubits and the number of iterations. We present an in-depth analysis of circuit construction strategies, analyzing their impact on convergence and provide practical guidance for efficient ansatz generation. Using ADAPT-VMPE, we construct up to 100-qubit ansätze for a strongly correlated photosensitizer currently undergoing human clinical trials for cancer treatment. Our results demonstrate that constant overlap with the ground state across system sizes can be reached in polynomial time with polynomially sized circuits.

In this paper, we prove energy and Morawetz estimates for solutions to Teukolsky equations in spacetimes with metrics that are perturbations, compatible with nonlinear applications, of Kerr metrics in the full subextremal range. The Teukolsky equations are written in tensorial form using the non-integrable formalism in \cite{GKS22}, and we follow the approach in \cite{Ma} of relying on a Teukolsky wave/transport system. The estimates are proved by extending the ideas from our earlier result \cite{MaSz24} on the corresponding problem for the scalar wave, notably the use of $r$-foliation-adapted microlocal multipliers for the wave part, and by incorporating techniques from \cite{Ma} to control the linear coupling terms between the components of the Teukolsky wave/transport system. Additionally, in order to adapt the methodology of \cite{MaSz24} to tensorial waves, we introduce a well-suited regular scalarization procedure which is of independent interest. This result, alongside our companion paper \cite{MaSz24}, is an essential step towards extending the current proof of Kerr stability in \cite{GCM1} \cite{GCM2} \cite{KS:Kerr} \cite{GKS22} \cite{Shen}, valid in the slowly rotating case, to a complete resolution of the Kerr stability conjecture, i.e., the statement that the Kerr family of spacetimes is nonlinearly stable for all subextremal angular momenta.

Dynamics of open quantum systems with structured reservoirs can often be simulated efficiently with tensor network influence functionals. The standard variants of the time-evolving matrix product operator (TEMPO) method are applicable when the systems is coupled to Gaussian bosonic baths via hermitian coupling operators that mutually commute. In this work we introduce a generalization to cases where the system is coupled to a single reservoir through multiple non-commuting operators, representing the most general form of linear system-bath coupling. We construct a Gaussian influence functional that properly handles Trotter errors arising from a finite evolution time step, thus ensuring convergence for long evolution times. Based on this result, the uniform TEMPO scheme can be employed to obtain a matrix product operator form of the influence functional, enabling efficient simulations of the real-time dynamics of the open system. As a demonstration, we simulate the time evolution of driven two-level emitters coupled to a bosonic lattice at different lattice sites.

Exciton transport in molecular aggregates is a fundamental process governing the performance of organic optoelectronics and light-harvesting systems. While most theoretical studies have emphasized long-time transport behavior, recent advances in ultrafast spectroscopy have brought into focus the short-time regime, in which exciton motion remains ballistic on femtosecond-to-picosecond timescales. In this work, we develop an analytical framework for short-time exciton dynamics in a one-dimensional lattice subject to both on-site energetic (diagonal) disorder and intermolecular coupling (off-diagonal) fluctuations. Utilizing the reciprocal-space analysis, we derive closed-form expressions for the first and second spatial moments considering both localized excitation and moving Gaussian initial conditions. Our analytical and numerical results show that, while the long-time dynamics are influenced by diagonal disorder, the short-time ballistic expansion is governed primarily by off-diagonal disorder. Crucially, we reveal a synergistic interplay between the average intermolecular coupling and the off-diagonal coupling disorder strength, demonstrating that they contribute equivalently to short-time exciton transport. Moreover, we integrate this generic disorder model with a realistic molecular system within the framework of macroscopic quantum electrodynamics, thereby providing a theoretical foundation for characterizing and optimizing ultrafast energy flow of disordered molecular aggregates in complex dielectric media.

Motzkin spin chain is a well-known mathematical model with connections to symmetry-protected topological phases, such as the Haldane phase, as well as to concepts in the AdS/CFT correspondence. They exhibit highly entangled ground states that violate the area law and are exceptionally difficult to simulate with conventional numerical methods. Numerical simulations of the Motzkin ground state become further challenging at large system sizes due to their high-dimensional spin structure, rendering it a natural test bed for quantum simulation with ultra-cold systems. Here, we propose a Rydberg-atom based quantum simulation scheme that effectively realizes Motzkin spins using an experimentally accessible set of parameters. We show that the resulting effective Motzkin ground state reproduces the characteristic entanglement scaling and the block-structure properties of the reduced density matrix associated with the ideal Motzkin state. Our results establish a pathway toward a concrete experimental realization of Motzkin spins beyond purely mathematical constructions, opening avenues for exploring other similar exotic non-area-law entangled phases in programmable Rydberg simulators.

The one-way distillable entanglement is a central operational measure of bipartite entanglement, quantifying the optimal rate at which maximally entangled pairs can be extracted by one-way LOCC. Despite its importance, it is notoriously hard to compute, since it is defined by a regularized optimization over many copies and adaptive one-way protocols. At present, single-letter formulas are only known for (conjugate) degradable and PPT states. More generally, it has remained unclear when one-way distillable entanglement can still be additive beyond degradability and PPT settings, and how such additivity relates to additivity questions of quantum capacity of channels. In this paper, we address this gap by identifying three explicit families of non-degradable and non-PPT states whose one-way distillable entanglement is nevertheless single-letter. First, we introduce two weakened degradability-type conditions--regularized less-noisy and informationally degradable--and prove that each guarantees additivity and hence a single-letter formula. Second, we show a stability result for orthogonally flagged mixtures: when one component has orthogonal support on Alice's system and zero one-way distillable entanglement, the mixture remains single-letter, even though degradability is typically lost under such mixing. Finally, we propose a generalized spin-alignment principle for entropy minimization in tensor-product settings, which we establish in several key cases, including a complete Rényi-2 result. As an application, we obtain additivity results for generalized direct-sum channels and their corresponding Choi states.

Generative quantum machine learning models are trained to deduce the probability distribution underlying a given dataset, and to produce new, synthetic samples from it. The majority of such models proposed in the literature, like the Quantum Circuit Born Machine (QCBM), fundamentally work on a binary level. Real-world data, however, is often numeric, requiring the models to translate between binary and continuous representations. We analyze how this transition influences the performance of quantum models and show that it requires the models to learn correlations that are solely an artifact of the way the data is encoded, and not related to the data itself. At the same time, structure of the original data can be obscured in the binary representation, hindering generalization. To mitigate these effects, we propose a strategy based on Gray-codes that can be implemented with essentially no overhead, conserves structures in the data, and avoids artificial correlations in situations in which the standard approach creates them. Considering datasets drawn from various one-dimensional probability distributions, we verify that, in most cases, QCBMs using the reflected Gray code learn faster and more accurately than those with standard binary code.

Family-Vicsek (FV) scaling provides an understanding for the growth and finite-size saturation of fluctuations in classical systems. Here, we extend the FV roughness to transferred segment magnetization after quantum quenches in a dissipative XXZ spin chain with homogeneous gain and loss, starting from a nonequilibrium steady state with finite magnetization. In the non-interacting limit, we derive a closed-form expression for the roughness in the presence of dissipation. It displays two-parameter FV scaling and smoothly interpolates between the clean ballistic behavior and the dissipation dominated scalings. For interacting chains, tensor-network simulations show that the non-dissipative ballistic growth at finite magnetization is robust, whereas the full Lindblad evolution is generically controlled by the dissipative relaxation time and exhibits a dissipation-dominated collapse.

We introduce the concept of quantum actuators as mediators for globally controlled quantum computation. Auxiliary quantum systems act as controllable elements that transiently store and release interaction energy, enabling the selective activation of multi-qubit gates within globally driven architectures. During compilation they remain passive and require no fine-grained local control, while during operation they allow for controlled activation of interactions and directional flow of quantum information. We provide a framework for embedding quantum actuators in globally controlled processors, showing how they enhance connectivity, enable long-range entangling operations, and bridge distant regions without increasing local control overhead. We discuss physical implementations and architectural strategies illustrating how these elements extend the capabilities of global-control schemes. A complementary interpretation in terms of quantum batteries naturally emerges, connecting global-control architectures with concepts from quantum thermodynamics while highlighting the distinct operational role of quantum actuators.

Ultralight bosonic dark matter with masses in the meV range, corresponding to terahertz (THz) Compton frequencies, remains largely unexplored due to the difficulty of achieving both efficient signal conversion and single-photon-sensitive detection at THz frequencies. We propose a hybrid detection architecture that integrates a dielectric haloscope, Rydberg-atom transducer, and superconducting nanowire single-photon detection within a unified cryogenic platform operating at $\lesssim 1\,\text{K}$. The dielectric haloscope converts dark matter into THz photons via phase-matched resonant enhancement, achieving form factors $C \sim 0.4$ and loaded quality factors $Q_L \sim 10^4$. A cold $^{87}$Rb ensemble then coherently up-converts the THz signal to the optical domain through six-wave mixing among Rydberg states. The intrinsic directionality and narrow bandwidth ($Δν_{\mathrm{atomic}} \sim 1\,\text{MHz}$) of this process provide extra suppression of isotropic thermal backgrounds. With 10 days of integration at $0.3\,\text{K}$, we project sensitivity to the axion-photon coupling $g_{aγγ} \sim 10^{-13}\,\mathrm{GeV}^{-1}$ at $m_a \sim 0.4\,\text{meV}$, reaching the QCD axion band and opening the THz window for searches of both axion and dark photon dark matter.

Topological pumping is conventionally governed by single-particle band topology. Here we show that promoting tunneling to a dynamical, occupation-conditioned variable fundamentally reshapes this paradigm, leading to occupation-selective topological pumping. In a periodically driven one-dimensional superlattice with density-dependent hopping, two-body bound states (doublons) acquire Chern numbers distinct from those of single particles and exhibit quantized transport even when the single-particle pump is trivial, including counter-propagating responses. We identify a dynamical-gauge-field mechanism that induces topological phase transitions in the bound-state sector absent from the single-particle spectrum. Furthermore, the gauge field concentrates Berry curvature into sharply localized resonant regions without compromising adiabatic quantization. A Floquet realization with ultracold atoms is proposed to realize such occupation-selective pumping. Our results reveal a mechanism for occupation-selective topological responses that can persist across higher-occupancy bound states.

The Traveling Salesman Problem is a fundamental combinatorial optimization problem widely studied in operations research. Despite its simple formulation, it remains computationally challenging due to the exponential growth of the search space and the large number of constraints required to eliminate subtours. This paper introduces a preprocessing strategy that significantly reduces the size of the optimization model by restricting the set of candidate arcs and retaining only the lowest-cost neighbors for each vertex. Computational experiments on TSPLIB benchmark instances demonstrate that the proposed approach substantially reduces the number of decision variables. The method is evaluated using both classical and quantum optimization techniques, showing improvements in computational time and reductions in optimality gaps. Overall, the results indicate that the proposed preprocessing enhances the scalability of the formulations and makes them more suitable for both classical solvers and emerging quantum optimization frameworks.

Vacuum energy, a prediction of quantum field theory, manifests itself as a cosmological constant in general relativity. In this Letter, we propose a novel inflationary scenario driven by a bare cosmological constant $Λ$, which terminates naturally through a self-tuning mechanism. Within Fab-Four gravity, self-tuning destabilizes the de Sitter state and drives the system toward a stiff-fluid attractor, thereby yielding a graceful exit. We construct two explicit models in which the slow-roll parameter evolves exponentially or as a power law. We show that the latter model, derived from center-manifold dynamics, significantly relaxes the required tuning of initial conditions. Our results establish, for the first time, that bare-vacuum-energy inflation with natural termination constitutes a viable dynamical possibility.

To approximate solutions of complex nonlinear partial differential equations remains a computational challenge, especially for sets of equations relevant in industry, such as Euler or Navier-Stokes equations. Even the most sophisticated computational fluid dynamic algorithms coupled with powerful supercomputers can not find approximate solutions for several design challenges in both adequate time and scale-resolving accuracy. One difficulty arises from solving high dimensional, strongly nonlinear partial differential equations, such as the Navier-Stokes equations, which capture the underlying physics. For nearly all classical algorithms, methods closely related to Newton's method are used to approximate a solution to the problem. Approximately solving the large-scale linear systems of equations occurring in this iterative scheme is generally a main contributor to the total computational complexity. In this paper a new quantum linear system solver supporting Newton's classical method to solve nonlinear partial differential equations is introduced. We present a new variant of the HHL algorithm, requiring less apriori information regarding the eigenvalues of the corresponding matrix. We apply this quantum linear system solver in a hybrid quantum-classical fashion to solve nonlinear partial differential equations. Moreover, a resource estimation for advanced use-cases of practical relevance is provided. Our results demonstrate how quantum computation may improve existing classical methodologies for solving nonlinear partial differential equations. This approach provides another promising application of quantum computers and presents a possible way forward for handling nonlinearities on inherently linear quantum systems.

Coherent Ising machines (CIMs) have emerged as specialized quantum hardware for large-scale combinatorial optimization. However, for large instances that remain challenging for classical methods, some platforms support only finite-precision inputs, and the required scaling and quantization can degrade solution quality. Dynamic portfolio optimization (DPO) can be formulated as a quadratic unconstrained binary optimization (QUBO) problem, but large instances are especially vulnerable to precision loss under global scaling. We propose a block coordinate descent method that decomposes the DPO model along the time dimension and iteratively solves compact time-block subproblems on the device. Experiments on finite-precision CIM hardware show that the method enables these instances to be solved under hardware precision limits, yields portfolios competitive with classical benchmark solvers, and reduces runtime through fast CIM solution of the resulting subproblems. These results demonstrate the promise of finite-precision CIMs as a practical and scalable approach to structured large-scale combinatorial optimization.

We develop a framework for identifying nonclassical speedups in systems with polarization, likewise spin degrees of freedom. By confining the dynamics to the manifold of angular momentum coherent states, which act as the classical reference in this case, we compute the speed limit that bounds the rate of change of the state achievable without generating quantum coherence. A comparison with the unrestricted quantum speed limit enables the quantitative identification of speedups arising from polarization nonclassicality. We apply this framework to the cross-Kerr interaction, demonstrating a persistent speedup scaling as $\mathcal{O}(\sqrt{N})$ with the photon number $N$. The results establish polarization nonclassicality as a genuine dynamical resource, linking quantum coherence to quantum-enhanced evolution speeds in nonlinear photonic systems.

We investigate the propagation of optical vector vortices of slow light in a coherently prepared four-level tripod atomic system. The vector vortex consists of superposed pulse pairs with opposite circular polarizations and orbital angular momentum (OAM) charges $\pm l$, weakly interacting with an atomic medium initially prepared in a coherent superposition of two ground states. A third unoccupied state is coupled to a stronger control laser without OAM, creating a phase-dependent configuration. In the linear regime, the vortex OAM is mapped onto the medium, producing symmetrical azimuthally structured absorption patterns, with losses significantly reduced by the control field. For small detunings, complementary spatially dependent amplification and absorption occur for the two circular polarization components. This OAM-structured coherence induces a dynamical anisotropy, affecting both the intensity and polarization of the slow-light vortex. Polarization states evolve periodically between left-circular, linear, and right-circular polarizations during propagation. Once the beam reaches a stationary regime, the ring-shaped intensity transforms into a petal-like structure, and the final polarization states stabilize according to the initial superposition. The rate of polarization transitions is tunable via the control field strength, demonstrating flexible control over slow-light vector vortex dynamics.

We present a complete and rigorous model of an EPR--Bell-type experiment within the framework of the de Broglie--Bohm theory. The purpose of this work is to show explicitly how a deterministic hidden-variable theory can reproduce all quantum-mechanical predictions, including the violation of Bell inequalities. Combining analytical arguments with numerical simulations, our approach offers a unified and transparent illustration of the central ingredients of de Broglie--Bohm theory, including particle trajectories, spin dynamics, and quantum entanglement. The model is designed to be pedagogical and self-contained, making it suitable for readers seeking a concrete understanding of how a nonlocal hidden-variable theory can describe the EPR--Bell experiment and illustrate Bell's theorem.

In the context of a Bianchi type-VI space-time characterized by Lyras geometry, we investigate the influence of a nonlinear spinor field on the evolution of the Universe. Our previous research has examined the nonlinear spinor field within Bianchi diagonal models, revealing that the spinor field exhibits non-trivial non-diagonal components of the energy-momentum tensor. These components impose various constraints on both the space-time geometry and the spinor field itself. The incorporation of Lyras geometry into the framework does not alleviate these constraints; however, it alters the preservation of the energy-momentum tensor. Furthermore, this integration complicates the relationship between the invariants of the spinor field and the space-time, ultimately affecting the outcomes of our analysis. Moreover, in this case energy-momentum tensor does not preserve.

We study zero-uncertainty states with quantum memory from an operator-algebraic perspective, which naturally accommodates degenerate projective-valued measurements. In the equal-dimension setting, we prove a rigidity theorem for purity and maximal entanglement. We then analyze two mechanisms by which this rigidity can fail: one arising from proper observable subalgebras, and the other from allowing larger memory dimensions. In these cases, we give corresponding algebraic decomposition and representation-theoretic descriptions, and compare their mathematical structure with their physical interpretation. Finally, we present an example from quantum steering to illustrate how our framework helps resolve a concrete physical question in a specific setting.

We investigate the potential to detect low-frequency gravitational waves (GWs) through their imprints on the proper motions of distant quasars observed by the Gaia mission. Using astrometric data from Gaia DR3, we simulate the effect of GWs on the proper motions of quasars, incorporating their actual sky positions and measurement uncertainties. We investigate two data analysis techniques for the extraction and characterization of GW signals from quasar proper motions: Vector Spherical Harmonics (VSH) and angular correlation functions, commonly referred to as Hellings-Downs curves (HDC). Using realistic simulated data, we forecast their sensitivity and accuracy to GWs, and evaluate the impact of systematic errors. From these simulations, we derive an upper limit on the amplitude of a stochastic GW background, constrained by the observational timespan, astrometric precision, and the sky distribution of quasars. Compared to HDC, VSH appears more statistically robust, less prone to selection effects, and with a significantly smaller computational cost, scaling as N. The HDC method is more sensitive for detecting gravitational waves, but its complexity scales as N^2. We find that, with Gaia DR3 proper motion errors, the lower limit for a detectable GW strain is of 10^{-11}, with possible improvements to about 3 x 10^{-12} for the next Gaia Data Release 4 (for the same number of quasars). This limit holds for a stochastic GW spectrum integrated over all frequencies less than half the inverse of the 34-month observational timespan of Gaia DR3, corresponding to approximately 5.6 nHz. We also investigate how different data-restriction and weighting schemes influence the final estimate of the gravitational wave strain.

Bose-Einstein condensates initially arranged in a long chain freely expand and interfere. If the initial phases of the condensates are identical, the initial density distribution is restored periodically during the expansion, giving rise to the Talbot effect. Even a slight disorder in the initial phases leads to a transformation of the interference pattern. In response to the phase disorder, the spectrum of the spatial density distribution acquires peaks that are absent in the case of identical phases. We derive an analytical expression for the spectrum of the spatial density distribution for an arbitrary phase disorder. We show that the new peaks emerging due to the phase disorder originate from pairwise interferences of the condensates. The positions of these peaks coincide with the wave vectors of the density modulations (wavelets) generated by such pairwise interferences. The absence of these peaks, when the initial phases are identical, is explained by the mutual destruction of the overlapping wavelets during their summation.

Primordial black holes (PBHs) can form in the early Universe, for instance during radiation domination, from the collapse of large-amplitude density perturbations shortly after horizon re-entry. This mechanism establishes an approximate one-to-one correspondence between the PBH mass and the scale of the peak in the primordial curvature perturbations. Consequently, the constraints on PBH abundances can be translated into upper limits on the amplitude of the primordial curvature power spectrum, thereby providing an indirect probe of the last e-folds of inflation corresponding to these smaller scales. We derive constraints on the amplitude of primordial curvature power spectra with both narrow and broad peaks using the most up-to-date bounds on PBH abundances. Given the theoretical uncertainties in PBH formation, we systematically compare the constraints obtained using the Press-Schechter (PS) formalism and peak theory, accounting for the nonlinear relation between curvature perturbations and density contrast. We quantify the impact of spherical versus non-spherical collapse criteria and show that including non-sphericity significantly increases the inferred amplitude of the primordial power spectrum, reflecting the larger threshold density contrast required for PBH formation. We also find that whereas the constraints obtained using the PS formalism and peak theory remain largely similar for the monochromatic case, they differ significantly toward smaller scales in the case of a broad primordial power spectrum. This discrepancy underscores that current constraints remain sensitive to the choice of statistical formalism. Our consistent treatment of monochromatic and extended mass functions provides a systematic mapping based on existing methodologies, while highlighting that reducing these theoretical uncertainties is a crucial step toward probing the early Universe through PBHs.

We investigate the relation between non-Hermitian Hamiltonian and Lindblad dynamics in nonequilibrium open quantum systems. Non-Hermitian models can extend phase diagrams and enable sensing advantages, but such effects often rely on postselection, raising questions about their relevance for unconditional dynamics. Using a minimal two-qubit setup mediating a heat current, we compare local and global Markovian master equations with their non-Hermitian counterparts. We observe that exceptional points emerge only in the local master equation and in the corresponding non-Hermitian Hamiltonian at sufficiently strong nonequilibrium. We further consider hybrid configurations, where one bath is treated with a Lindblad description and the other with a non-Hermitian approach, interpolating between the two extremes. Our results contribute understanding the role of quantum jumps and exceptional points in nonequilibrium open quantum systems and identify a simple, experimentally accessible architecture, realizable, for instance, in circuit-QED platforms, for their exploration.

The realization of scalable quantum battery architectures requires concern not only with how much energy can be stored, but also with how energy is transported, distributed, and converted into extractable work across connected battery nodes. While previous studies mainly focused on collective charging in multi-cell quantum batteries, the topology-dependent transport law and the corresponding work-oriented performance of quantum battery networks remain largely unexplored. In this work, we investigate quantum battery networks with engineered reciprocal and nonreciprocal couplings and compare different connection topologies, including cascaded and parallel architectures, within a unified transport framework. In the nonreciprocal regime, the optimal coupling follows distinct scaling laws for the two connection topologies, namely $J_{\rm op}^{c}\propto N$ for cascaded transport and $J_{\rm op}^{p}\propto N^{-1/2}$ for parallel charging in the large-$N$ limit. In reciprocal cascaded networks, a parity-dependent spectral response produces an odd-even transport effect that is absent in the nonreciprocal and parallel configurations. We further analyze the role of thermal and squeezed reservoirs and show that thermal noise mainly increases passive energy, whereas squeezing enhances ergotropy and thus the useful fraction of stored energy. These results shift the emphasis from charging enhancement to transport engineering and provide architecture-level design principles for quantum battery networks.

We investigate the dynamical evolution of quantum discord, entanglement and purity in an open quantum system of two coupled asymmetric harmonic oscillators interacting with a thermal environment. Using the Kossakowski-Lindblad master equation we analyze the time evolution starting with a squeezed vacuum state. In contrast to our previous study on entanglement evolution in asymmetric oscillators, the present work introduces XY-type position-position coupling together with a systematic joint analysis of quantum discord and purity alongside entanglement. We examine the combined effects of the squeezing parameter, asymmetry parameter, coupling constant, dissipation rate and temperature. We find that quantum discord and entanglement exhibit, in general, a non-monotonic decrease over time. Increasing temperature consistently accelerates the degradation of both quantum correlations and purity, whereas increasing dissipation accelerates the degradation of quantum correlations but leads to higher steady-state purity. Increasing the squeezing parameter provides a protective effect by enhancing initial correlations and prolonging entanglement survival time, while increasing the coupling constant leads to higher quantum correlations. The asymmetry parameter exhibits only a weak influence on the correlation evolution. Our analysis reveals that quantum discord demonstrates stronger resilience than entanglement, which can present more complex behaviour including entanglement sudden death and possible temporary revivals and re-suppressions. These findings provide valuable insights for developing robust quantum information protocols and strategies for preserving quantum correlations in realistic open quantum systems, with potential extensions to non-Markovian regimes and multi-mode architectures.

Quantum steering, as a manifestation of nonlocal quantum correlations, plays a crucial role in enabling various quantum information processing tasks. However, practical implementations are often hindered by significant challenges arising from imperfect or untrusted measurement devices. This study investigates the impact of measurement inaccuracies on quantum steering, with a particular focus on errors in the untrusted party's measurement devices. We first analyze how such errors affect the evaluation of steering inequalities, and then derive bipartite steering inequalities based on correlation matrices under imperfect measurements. Our findings show that even small measurement errors can significantly compromise the certification of quantum steerability, an effect that becomes particularly pronounced as the system dimension increases. Furthermore, by extending the proposed steering inequality to a modified tripartite scenario via correlation matrices, we demonstrate that the influence of measurement imperfections is far more severe in multipartite quantum steering than in the bipartite case. Our results underscore the critical need to account for measurement imperfections in experimental quantum steering and provide a theoretical framework for characterizing and mitigating these effects in high-dimensional quantum systems.

The non-Hermitian skin effect (NHSE), which drives bulk states toward system boundaries, modifies bulk-boundary correspondence and complicates the identification of topological edge modes. Although breaking translational symmetry is known to influence this behavior, a systematic comparison of different structural classes remains limited. Here we investigate periodic, random, and quasiperiodic (Fibonacci) systems using a one-dimensional non-Hermitian quantum walk model. By matching the local scattering parameters in a topologically nontrivial regime, we isolate the role of spatial structure in the presence of the NHSE. We find that periodic systems exhibit pronounced boundary accumulation of bulk states. Random systems suppress this accumulation through Anderson localization, but the topological gap becomes partially filled with localized in-gap states. In contrast, the Fibonacci quasiperiodic system suppresses large-scale boundary accumulation while maintaining a well-defined topological gap. Analysis of the wave functions suggests that the hierarchical quasiperiodic structure fragments bulk states across multiple length scales, thereby mitigating the NHSE. These results identify deterministic quasiperiodicity as a distinct structural regime for controlling non-Hermitian skin dynamics and isolating topological boundary modes.

We present a study of virtual absorption modes (VAMs) in Schwarzschild-de Sitter (SdS) spacetime under semi-open boundary conditions, where the VAMs correspond to total transmission modes (TTMs) with the reflection amplitude being vanished. Our numerical analysis reveals that as the reflectivity $|\mathcal{K}|$ decreases, the VAM spectra migrate systematically toward regions of less negative imaginary parts, with each overtone exhibiting a critical reflectivity at which $\text{Im}(ω_{\text{VAM}})=0$. Using simulations based on spectral collocation methods, it is demonstrated that excitation precisely at a VAM spectrum leads to coherent perfect absorption (CPA). These results establish VAMs as the spectrum signatures of CPA for exotic compact objects (ECOs).

We introduce STAR-magic mutation, an efficient protocol for implementing logical rotation gates on early fault-tolerant quantum computers. This protocol judiciously combines two of the latest state preparation protocols: transversal multi-rotation protocol and magic state cultivation. It achieves a logical rotation gate with a favorable error scaling of $\mathcal{O}(θ_L^{2(1-Θ(1/d))}p_{\text{ph}})$, while requiring only the ancillary space of a single surface code patch. Here, $θ_L$ is the logical rotation angle, $p_{\text{ph}}$ is the physical error rate, and $d$ is the code distance. This scaling marks a significant improvement over the previous state-of-the-art, $\mathcal{O}(θ_L p_{\text{ph}})$, making our protocol particularly powerful for implementing a sequence of small-angle rotation gates, like Trotter-based circuits. Notably, for $θ_L \lesssim 10^{-5}$, our protocol achieves a two-order-of-magnitude reduction in both the execution time and the error rate of analog rotation gates compared to the standard $T$-gate synthesis using cultivated magic states. Building upon this protocol, we also propose a novel quantum computing architecture designed for early fault-tolerant quantum computers, dubbed ``STAR ver.~3". It employs a refined circuit compilation strategy based on Clifford+$T$+$φ$ gate set, rather than the conventional Clifford+$T$ or Clifford+$φ$ gate sets. We establish a theoretical bound on the feasible circuit size on this architecture and illustrate its capabilities by analyzing the spacetime costs for simulating the dynamics of quantum many-body systems. Specifically, we demonstrate that our architecture can simulate biologically-relevant molecules or lattice models at scales beyond the reach of exact classical simulation, with only a few hundred thousand physical qubits, even assuming a realistic error rate of $p_{\text{ph}}=10^{-3}$.

The Hierarchical equations of motion (HEOM) method is an important non-perturbative technique, allowing numerically exact treatment of open quantum systems with strong coupling and non-Markovian memory. However, its encoding of bath memory into auxiliary density operators often limits direct access to detailed bath information. In contrast, the reaction-coordinate (RC) mapping allows direct and transparent access to the dominant collective bath mode, but its perturbative and often Markovian treatment of the residual bath restricts its reliability. In this work, we introduce RC-HEOM, a hybrid method that unifies the strengths of both approaches by combining RC mapping with a fully non-perturbative HEOM description of the residual bath. RC-HEOM simultaneously retains exact non-Markovian memory and access to the RC mode, which enables analysis of system-RC information. Applying this method to the Anderson impurity models, we directly track the emergence of the Kondo singlet from the growth of the Kondo resonance and uncover a nontrivial RC-mediated coherence revival. These results demonstrate that RC-HEOM is a promising method for characterizing open quantum systems in regimes that are difficult to access with conventional master-equation methods.

Entangled quantum probes can achieve Heisenberg-limited measurement precision, but this advantage is typically destroyed by noise. We address this issue by introducing a framework that we call encoded quantum signal processing, which unifies quantum error detection and quantum signal processing into an effective single-qubit framework, and provides a paradigm for constructing logical sensors that are robust to noise while remaining sensitive to the signal of interest. We show that encoding sensor qubits into a repetition code and using syndrome measurements as a signal-processing primitive restores Heisenberg scaling under realistic noise, without applying recovery operations. We prove that product-state sensing with syndrome post-processing is fundamentally limited to standard quantum limit (SQL) scaling, and develop four protocols that overcome this barrier through entanglement or sequential signal amplification, achieving Heisenberg-limited precision with exponential error suppression in code distance. For spatially inhomogeneous fields, Bayesian marginalization preserves Heisenberg scaling provided noise decreases sufficiently with system size. The underlying mechanism, which we formalize as encoded quantum signal processing, reduces multi-qubit metrology to an effective single-qubit problem where syndrome measurement implements nonlinear signal transformations. Numerical simulations validate the theoretical predictions: syndrome-based inference achieves near-Heisenberg scaling at noise levels where bare probes approach the SQL, and a concatenated protocol maintains this scaling under joint transverse noise and longitudinal inhomogeneities.

Numbers-on-Forehead (NOF) communication model is a central model in communication complexity. As a restricted variant, one-way NOF model is of particular interest. Establishing strong one-way NOF lower bounds would imply circuit lower bounds, resolve well-known problems in additive combinatorics, and yield wide-ranging applications in areas such as cryptography and distributed computing. However, proving strong lower bounds in one-way NOF communication remains highly challenging; many fundamental questions in one-way NOF communication remain wide open. One of the fundamental questions, proposed by Gavinsky and Pudlák (CCC 2008), is to establish an explicit exponential separation between quantum and classical one-way NOF communication. In this paper, we resolve this open problem by establishing the first exponential separation between quantum and randomized communication complexity in one-way NOF model. Specifically, we define a lifted variant of the Hidden Matching problem of Bar-Yossef, Jayram, and Kerenidis (STOC 2004) and show that it admits an ($O(\log n)$)-cost quantum protocol in the one-way NOF setting. By contrast, we prove that any $k$-party one-way randomized protocol for this problem requires communication $Ω(\frac{n^{1/3}}{2^{k/3}})$. Notably, our separation applies even to a generalization of $k$-player one-way communication, where the first player speaks once, and all other $k-1$ players can communicate freely.

Dark energy stars (DESs), described by the modified Chaplygin gas (MCG), can be dynamically stable and fall within different observational measurements. In this work, we employ diverse macroscopic properties, such as compactness $C$, moment of inertia $I$, tidal deformability $Λ$, gravitational binding energy $E_g$ and $f$-mode nonradial pulsation frequency, to explore whether they are correlated by universal relations (URs). Remarkably, our stellar configurations always obey the causality condition and are compatible with several observational mass-radius constraints. Via the $C-I-\text{Love}-f$ URs, our results reveal that we cannot distinguish quark stars (QSs) from DESs in the sense that DESs satisfy several URs very similar to those of QSs. However, when we involve $E_g$, DESs and QSs can be strongly distinguished through the $I-E_g^{-2}$, $Λ-E_g^{-5}$ and $f-E_g^{-2}$ URs. We also make use of these findings and the tidal deformability constraint from the GW170817 event to forecast the canonical properties of a $1.4\, M_\odot$ compact star. Furthermore, we present a set of fine empirical correlations involving the tidal deformability, obtained from an extensive scan of the parameter space of our DE stellar models.

Quantum simulation of molecular electronic structure is one of the most promising applications of quantum computing. However, achieving chemically accurate predictions for strongly correlated systems requires quantum phase estimation (QPE) on fault-tolerant quantum computing (FTQC) devices. Existing resource estimates for typical FTQC architectures suggest that such calculations demand millions of physical qubits, thereby placing them beyond the reach of near-term devices. Here, we investigate the feasibility of performing QPE for chemically relevant molecular systems in an early-FTQC regime, characterized by partial fault tolerance, constrained qubit budgets, and limited circuit depth. Our framework is based on single-ancilla, Trotter-based QPE implementations combined with partially randomized time evolution. Within this framework, we develop a novel Hamiltonian optimization strategy, termed unitary weight concentration, that reduces algorithmic cost by reshaping linear-combination-of-unitaries representations. Applying this framework to active-space models of iron-sulfur clusters, cytochrome P450 active sites, and CO$_2$-utilization catalysts, we perform end-to-end resource estimation using the latest version of the space-time efficient analog rotation (STAR) architecture. Our results indicate that ground-state energy estimation for active spaces of approximately 20 to 50 spatial orbitals, well beyond the reach of classical full configuration interaction, is achievable using $\sim 10^5$ physical qubits, with runtimes on the order of days to weeks. These findings demonstrate that while full-fledged fault-tolerant quantum computers will be required for even larger molecular simulations, chemically meaningful quantum chemistry problems are already within reach in an experimentally relevant, early-FTQC regime.

Enhancement of quantum battery performance is a popular subject in quantum thermodynamics. An interesting phenomenon is the quick charging effect [Phys. Rev. Res. 6, 023136 (2024)], which has been explored by utilizing a quantum interferometric technique known as superposition of trajectories. A similar technique used to boost quantum battery performance is indefinite causal order. Here, we propose a new charging protocol that utilizes cyclic indefinite causal order, whereby $N$ charging sequences are superposed when utilizing $N$ chargers. We observe charging efficiency bursts when implementing our cyclic indefinite charging protocol. The duration of these bursts increase with $N$. Additionally, we present a circuit model to implement our charging protocol for the two-charger scenario and perform proof-of-concept demonstrations on IonQ, Quantinuum and IBMQ quantum processors. The results validate the existence of charging efficiency bursts as shown by our theoretical analysis and numerical simulations.

This work investigates future cosmological singularities in a flat FLRW universe filled with a single barotropic fluid, ($p = (γ- 1)ρ$), within the framework of unimodular gravity. In this setting, the non-conservation of the energy-momentum tensor is encoded through an energy diffusion function $Q$. While a constant diffusion term leads to an effective cosmological constant and preserves adiabatic evolution, a time-dependent $Q(t)$ induces non-adiabatic dynamics. We consider a power-law Ansatz for $Q$ as a function of the redshift and impose the condition of positive entropy production. This requirement leads to non-trivial constraints on the model parameters, with direct implications for the admissible singularity structure. In particular, within the thermodynamically allowed sector, we show that Big Rip singularities are excluded for non-phantom fluids when the cosmological constant is positive. For phantom fluids, the model reproduces the expected Big Rip behavior, as well as Big Crunch solutions for negative cosmological constant. More importantly, we show that diffusion can induce an effective phantom regime even when the fundamental fluid is non-phantom. In particular, for a negative cosmological constant, we present an explicit realization of a Big Rip singularity in unimodular gravity driven by diffusion, while consistently preserving a non-phantom equation of state and positive entropy production. These results reveal a novel mechanism for the emergence of future singularities, with no direct analogue in standard General Relativity.

Interference provides a fundamental mechanism for generating and manipulating entanglement in many-body quantum systems. Here, we develop an interference framework in which the nonlinear dynamics of collective spin-$\tfrac{1}{2}$ ensembles are mapped onto phase accumulation and self-interference in phase space, providing a direct and physically transparent description of entanglement formation. Within this framework, one-axis twisting produces Greenberger-Horne-Zeilinger (GHZ) states, while two-axis twisting generates multi-component GHZ superpositions relevant for multiparameter quantum metrology. Building on this interference-based description, we analyze metrological performance under realistic Markovian noise, including local and collective emission, pumping, and dephasing, and examine the role of Hamiltonian control based on linear and nonlinear interactions. We show that while control can enhance single-parameter sensitivity in a noise-dependent regime, the achievable precision in multiparameter estimation is fundamentally constrained. These results establish interference as a unifying principle linking nonlinear dynamics, entanglement generation, and metrological performance, and reveal intrinsic limitations of multiparameter quantum sensing. Our framework provides broadly applicable insight into the design of robust quantum-enhanced measurement protocols in noisy many-body systems.

In waveguide quantum electrodynamics (QED) systems, a giant cavity can be engineered to interact with quantum fields by multiple distant coupling points so that its non-Markovian dynamics are quite different from traditional quantum optical cavity systems. Towards feedback control this system, this paper designs an optimal filter for the giant cavity systems to estimate its state evolution under continuous quantum measurements. Firstly, the Langevin equation in the Heisenberg picture are derived, which is a linear continuous-time system with both states and inputs delays resulting from the unconventional distant couplings. Compared to existing modeling approaches, this formulation effectively preserves the nonlocal coupling and multiple delay dynamic characteristics inherent in the original system. In particular, the presence of coupling and propagation delays leads to noncommutativity among the system operators at different times, which prevents the direct application of existing quantum filtering methods. To address this issue, an optimal filter is designed, in which the delayed-state covariance matrices are computed. By iteratively evaluating the delayed-state covariance over successive time intervals, the resulting optimal filter can be implemented in an interval-wise backward recursion algorithm. Finally, numerical simulations are conducted to evaluate the tracking performance of the proposed optimal filter for the giant cavity. By comparing between the evolutions of Wigner functions of coherent and cat states and the filter, the effectiveness of the optimal filter is validated.

In this work, we derive a deterministic master equation to model a general, possibly non-Markovian, feedback. The master equation describes a system with a general evolution and measurement operation, with feedback being applied in terms of signal processing. The feedback signal has an arbitrary structure with dimensionality that indicates the degree of non-Markovianity of the information processing. We present examples to illustrate how such a master equation can be used to model systems with memory feedback and non-trivial frequency dependence.

Using matching properties of the gravitational field at timelike and spatial infinity, together with universal formulas for gravitational wave tails, we obtain exact expressions for the peeling-violating components of the Weyl tensor at future and past null infinity. The coefficients depend solely on incoming scattering data and reduce, in the Newtonian limit, to those found by Damour long ago. The computation is facilitated by the observation that the tail and peeling-violating coefficients organize into a "celestial diamond" introduced in the context of celestial holography. The analysis suggests a new matching condition at spatial infinity which, together with previously found matching at timelike infinity, capture both tail and peeling-violation formulas. We conclude by pointing out that stronger peeling violations occur in perturbative quantum gravity, in agreement with a recently reported amplitude-based result.

We investigate the role of causality in the generalized EPRL spinfoam model introduced by Kaminski, Kisielowski, and Lewandowski (EPRL-KKL). We first propose a definition of causal structure on an arbitrary 2-complex and analyze how causal orientations can be assigned to its 1-skeleton and 2-skeleton. We identify a criterion that determines when an orientation of the 2-skeleton induces a consistent causal structure on the 1-skeleton. This characterization naturally involves tools from graph theory and linear algebra over the Galois field $\mathbb{F}_2$. Using these results, we introduce a causal vertex amplitude that generalizes previous proposals by Bianchi--Dussaud and Bianchi--Chen--Gamonal. We study its asymptotic behavior and revisit the interpretation of the two critical points appearing in the semiclassical analysis of the EPRL and EPRL-KKL spinfoam models. Finally, we discuss several directions in which the causal amplitude introduced here may provide new insights for future developments, like the cosine problem and the presence of irregular light cone structures.

Quantum information measures are used to study the quantum phase diagrams of the two-level Dicke model including the atomic dipole-dipole interaction, for a finite number of particles, with and without the rotating-wave approximation, which yields the conservation of the total number of excitations in the first case and its parity in the general case. We show that the quantum phase transitions can be observed in the fluctuation of the atomic populations and that of the number of photons, and also that the conditional probability distribution of the population of the excited state with zero photons carries the information of the quantum phase transitions when the matter-field interaction is weak.

Levitated sensors have emerged as a new frontier to detect ultra-light dark matter such as axion-like particles and dark photons. In this work we study how a magnetically levitated superconductor responds to a spin-2 dark matter field, the dark graviton, in the dHz to kHz frequency range. To do so, we compute the forces that the dark graviton exerts on the superconductor, separately for matter and light couplings. The matter coupling produces a strain-like tidal acceleration between the superconductor and the readout pick-up loop in a way that is akin to a slow, continuous, massive gravitational wave. The light coupling instead induces an effective current that sources an oscillating magnetic field, thus driving the superdiamagnetic response of the superconductor. We find that, even with significant experimental improvements, the sensitivity reach for the matter coupling is not competitive with existing interferometers or fifth-force experiments. On the other hand, magnetically levitated superconductors could be among the most sensitive laboratory probes of the dark-graviton coupling to electromagnetism, especially at low frequencies, provided technical and readout noise can be kept under control.

The distance of a classical or quantum code is a key figure of merit which reflects its capacity to detect errors. Quantum LDPC code families have considerable promise in reducing the overhead required for fault-tolerant quantum computation, but calculating their distance is challenging with existing methods. We generally assess the performance of a quantum code under circuit level error models, and for such scenarios the circuit distance is an important consideration. Calculating circuit distance is in general more difficult than finding the distance of the corresponding code as the detector error matrix of the circuit is usually much larger than the code's check matrix. In this work, we benchmark a wide range of distance-finding methods for various classical and quantum code families, as well as syndrome-extraction circuits. We consider both exact methods (such as Brouwer-Zimmermann, connected cluster, SAT and mixed integer programming) and heuristic methods which have lower run-time but can only give a bound on distance (examples include random information set, syndrome decoder algorithms, and Stim undetectable error methods). We further develop the QDistEvol algorithm and show that it performs well for the quantum LDPC codes in our benchmark. The algorithms and test data have been made available to the community in the codeDistance Python package.

We study neutron star configurations in a teleparallel gravity model featuring a scalar field coupled to both matter and torsion. In the Einstein frame, the theory includes a derivative coupling between the scalar field and the torsion vector, together with a conformal matter coupling \(A(φ)=\exp(βφ^{2}/2)\). Static and slowly rotating neutron-star solutions are constructed for realistic equations of state, focusing on the APR and MS1 equations of state. Scalarized solutions appear only within a finite range of central densities and correspond to localized deviations from the general-relativistic mass--radius and mass--central-density relations. The onset and extent of scalarization depend on the equation of state and on the strength of the derivative torsional interaction, which can either enhance or suppress scalarization relative to the general-relativistic scalarized branch. At high central densities, scalarization is quenched and the solutions approach the general-relativistic limit, remaining bounded even for strong torsional couplings. No scalarized solutions are found in the absence of matter coupling (\(β=0\)). The normalized scalar charge follows trends consistent with the global mass relations, indicating an intermediate scalarized regime suppressed at high compactness. For slowly rotating stars, the moment of inertia depends systematically on the torsional coupling and the equation of state, with stiffer equations yielding larger values. These results highlight the potential of neutron-star radius and rotational measurements to test teleparallel scalarization scenarios.

We consider the scalar-tensor theory witn non-minimal coupling in the Jordan frame. The action of the model contains a potential term $U(\varphi)$, a coupling function $f(\varphi)$. We explore a reconstruction procedure for a generic static spherically symmetric metric written in the Buchdahl parametrization: $ds^2 = \left(A(u)\right)^{-1}du^2 - A(u)dt^2 + C(u)dΩ^2$, with given $A(u) > 0$ and $C(u) > 0$. The procedure gives the relations for $U(\varphi(u))$, $f(\varphi(u))$ and $d\varphi/du$, which lead to exact solutions to equations of motion with a given metric. A key role in this approach is played by the solutions to a first order linear differential equation for the function $f(\varphi(u))$. The formalism is illustrated by two examples when: a) the Reissner-Nordström-(Anti-)de Sitter metric and b) the Bocharova-Bronnikov-Melnikov-Bekenstein-(Anti)de-Sitter metric are chosen as a starting point.

We calculate the present-day and cosmological volumetric rate of primordial black hole (PBH) quantum tunneling events to white holes, incorporating the competition between Hawking evaporation and tunneling, cosmological depletion, realistic mass-dependent abundance constraints, extended mass functions, and the alternative memory-burden scenario. The burst rate is maximized along a narrow ridge in the mass--tunneling-parameter plane where the effective PBH lifetime is comparable to the age of the Universe. Within the canonical Planck-star range of tunneling timescales, FRB-level rates arise only in two highly restricted regions: a low-mass window near the evaporation boundary, and a narrow sequential window where evaporation precedes tunneling; broadening the PBH mass function does not qualitatively alter this conclusion. We further assess observational constraints from FRB repetition statistics, radio spectral properties, prompt and diffuse gamma-ray limits, host-galaxy demographics, and gravitational-wave signatures. All current observations are consistent with a subdominant white-hole contribution to the FRB population but strongly disfavor a dominant origin. The rate calculation does not generically support FRB-level event densities, and any viable FRB interpretation requires narrow, fine-tuned, and strongly assumption-dependent corners of parameter space.

As Hermitian operators, many-body Bell operators can naturally be identified as many-body Hamiltonians. An important subclass of such Hamiltonians is the stoquastic class, characterized by having nonpositive off-diagonal matrix elements in a given basis. Interestingly, this property is shared by the permutationally invariant (PI) Bell operators underlying the largest Bell-correlation experiments to date. In this work, we explore the connection between many-body PI Bell operators and stoquasticity. We introduce the stoquasticity cone, which allows for a full characterization of the stoquastic parameter regimes for any PI Bell operator. We use this to show that PI Bell operators of the binary-input binary-output scenario consisting of at most three-body correlators can always be rendered stoquastic for any set of measurement parameters. Additionally, we also provide examples that use the stoquasticity cone to optimize for the quantum-classical gap. Numerical evidence suggests that the Bell operator used in the largest experiments to date is optimal with respect to stoquasticity. To the best of our knowledge, this work establishes the first connection between PI Bell operators and stoquasticity.

The spin-weighted spheroidal functions are the eigenfunctions of the angular Teukolsky equation. They are a generalization of the widely used spin-weighted spherical functions, and are extremely important in the area of black-hole perturbation theory. Like other special functions, they have an inherent phase ambiguity and need to be phase fixed to be uniquely defined. Clearly, such a phase choice does not have a direct physical impact. But, a poorly constructed phase choice could hinder the extraction, from mode information, of physical information about a system. To date, possible phase choices for the spin-weighted spheroidal functions have received little attention. Here, we clearly define and extensively explore two useful phase fixing schemes, and we propose that the spherical-limit phase-fixing scheme be adopted as the default phase-fixing scheme for the spin-weighted spheroidal functions.

Using state-of-the-art scattering results in post-Minkowskian (PM) gravity, we improve the tidal sector of four different flavors of the effective-one-body (EOB) formalism. We notably explore both adiabatic and post-adiabatic gravitoelectric and gravitomagnetic quadrupolar tidal effects at the next-to-next-to-leading PM-order. When comparing the predictions of the so-constructed Lagrange-PM-tidal version of EOB to recent numerical-relativity data on the scattering of neutron stars, we find improved agreement with respect to existing EOB models and PM expansions. Our work lays the foundation for the development of an accurate tidal sector of the PM EOB models, and points out the need to explore improved resummation schemes in PN EOB for bound and circularized orbits.

We revisit the population properties of neutron star-black hole (NSBH) mergers using low-mass compact binary coalescences reported through GWTC-4. Employing pyEFPE, an inspiral-only waveform model that captures both orbital eccentricity and spin-induced precession, we reanalyse all binary neutron star (BNS) and NSBH events observed via gravitational waves. The BNS systems GW170817 and GW190425 are fully consistent with quasi-circular inspirals, while GW200105 stands out among the NSBH binaries as the only system exhibiting significant residual eccentricity at 20 Hz, strengthening evidence for dynamically driven formation pathways. The remaining NSBH events show no measurable eccentricity and appear broadly compatible with low-spin binaries formed through isolated stellar evolution. Using hierarchical Bayesian inference, we obtain the first joint constraints on the mass, spin, and eccentricity distributions of NSBH binaries. Our results also yield the first simultaneous constraints on spin precession and orbital eccentricity in NSBH mergers, while the inferred merger rates remain fully consistent with previous LVK measurements. Treating all NSBH systems as a single population yields results compatible with formation in hierarchical triples, whereas the quasi-circular population remains broadly consistent with isolated evolution. Our results highlight the emerging role of eccentricity as a key discriminator between formation channels. As the number of NSBH detections grows, joint constraints on masses, spins, and orbital eccentricity will enable increasingly sharp tests of dynamical versus isolated binary evolution, establishing NSBH systems as powerful probes of compact-object astrophysics.

In phase-randomized weak coherent pulse (WCP) implementations of Quantum Key Distribution (QKD) BB84 protocol, the decoy method is often used to compensate BB84's vulnerability against photon number splitting (PNS) attacks. However, this typically introduces extra complexities and requirements on experimental devices. In this paper, we are therefore interested in phase-randomized WCP implementations without the decoy method. We examine the performance of three QKD protocols with different classical announcement structures, namely BB84, SARG04, and No Public Announcement of Basis (NPAB) BB84, using numerical security proof techniques. We compare secure key rates of the three protocols in asymptotic and finite-size regimes for lossy and noisy channels. The three protocols show different relative advantages depending on the channel behaviour. Canonical BB84 shows robustness against errors and depolarization, SARG04 demonstrates resilience against high loss channels, and NPAB BB84 shows potential advantages against physical misalignment between QKD devices.

In the current world, the use of artificial intelligence is penetrating every aspect of human life. The basic element of any artificial intelligence is a digital neuron, called a perceptron, while its quantum analogue is called a quantum perceptron. Here, we introduce a model of perceptron called the informationally-restricted measurement-based perceptron (IMP), where each input is composed of two bits, while at the node, depending on a free input variable, the perceptron decides which bit to evaluate. Additionally, the states transmitted from the input to the node are restricted to a bit (qubit). We establish that under this restriction, the quantum IMP predicts better than a classical IMP. This means that under dimensional restriction of the transmitted states, when both the classical and quantum perceptrons learn the same, the quantum perceptron predicts better than the classical perceptron. For our purpose, we find specific learned values of the perceptron that can display the advantage of a quantum perceptron over its classical counterpart. Restricting to discrete binary inputs, we establish that the observed quantum advantage is universal, that is, for any non-trivial function implementable by both the quantum and classical IMP, one can always find a quantum implementation that outperforms the predictive capability of every classical one. This points to the fact that, given identical learning and resources, a quantum perceptron would predict better than any classical one.

We present an extension of the Evolving density matrices on Qubits (E$ρ$OQ) framework that enables efficient fault-tolerant preparation of fermionic quantum states. The original method circumvents state preparation by stochastic sampling, but faces a sign problem in fermionic systems leading to a large number of circuits necessary. We resolve this by combining classical stochastic sampling with a linear combination of unitaries method that avoids the exponential circuit scaling that plagued naïve implementations. The resulting algorithm requires $\mathcal{O}(M^2)$ $R_Z$ rotations for circuit preparation, where $M$ is the number of retained basis states. We validate the method for ground and excited states in the Thirring model, including by computing two-point correlation functions relevant to scattering. In this model for fixed accuracy $\varepsilon$, $M$ is found to scale empirically as $M \propto \frac{1}{mg}\log(1/g)\log(1/m)$.

The coalescence of degenerate helium cores during red giant collisions - a process we term erythrohenosis - introduces a novel class of transient astrophysical sources of high-energy neutrinos. Using stellar models generated with MESA and SPH simulations of the final inspiral phase, we develop a semi-analytical model to estimate the amount of hydrogen mixed into the cores, the energy release ($\approx 4.28 \times 10^{49}$ erg) that heats the remnant to $T_f \approx 5.3 \times 10^8$ K, the magnetic field amplification ($B \approx 1.77 \times 10^{10}$ G), and the resulting neutrino flux. We find that the predicted TeV--PeV neutrino signal can account for the diffuse neutrino flux observed by IceCube and demonstrate that a single merger event within $\sim 2$ Mpc would be detectable in this energy regime. Furthermore, we discuss the probability of a magnetized helium flash and assess the subsequent activation of the CNO cycle in the remnant core due to hydrogen mixing. In particular, neutrinos from the decay of $^{18}$F offer a direct observational test of the detonation. The simultaneous emission of high-energy hadronic neutrinos, gravitational waves, and -- if the optical depth permits -- an electromagnetic signal would constitute a unique multimessenger signature of red giant core collisions, positioning erythrohenosis events as exotic yet potentially observable phenomena in dense stellar systems.

We study magnetic materials whose low energy physics can be effectively described by a Dicke model, which we term Dicke materials. We show how a Dicke model emerges in such materials due to a coexistence of fast-dispersing and slow-dispersing spins, which are strongly coupled. Analogous to the paradigmatic Dicke model describing light-matter interactions, these materials also exhibit signatures of a superradiant phase transition. The ground state near the superradiant phase transition is expected to be squeezed, making Dicke materials a resource for quantum metrology and witnessing entanglement in solid-state systems. However, as an entanglement measure, squeezing can be sensitive to perturbations that are otherwise irrelevant for usual correlation functions and order parameters. Motivated by the prospect of observing squeezing in such Dicke materials, we study the robustness of ground state squeezing under ubiquitous imperfections such as finite temperature, disorder, and local interactions. Using analytical and numerical techniques, we show that the squeezing obtained is perturbatively stable against these imperfections and quantitatively evaluate regimes promising for experimental observation.

Casimir force searches provide among the most sensitive laboratory probes of new short range interactions. Existing constraints rely almost exclusively on a single geometry. We show that Casimir geometry constitutes an independent observable, as Yukawa-type interactions and Casimir background exhibit different geometric scaling for bulk forces and surface quantum effects. We derive the first constraints from sphere-sphere and plate-plate geometries, thereby completing the canonical set of Casimir geometries, obtaining the most stringent Casimir-based bounds for $λ\lesssim 10^{-8}~\mathrm{m}$. Our results establish geometry as a new handle for systematic searches for short range forces.

Recent analyses combining data from the cosmic microwave background (CMB), baryon acoustic oscillations (BAO), and Type Ia supernovae (SN) have revealed a tentative observational preference for phantom crossing in the dark energy equation of state $w$. We argue that this preference is a natural consequence of the $Ω_m$ tensions that arise when these datasets are individually fit to $Λ$CDM, specifically because of the ordering $Ω_m^\mathrm{BAO} < Ω_m^\mathrm{CMB} < Ω_m^\mathrm{SN}$. We show both theoretically and empirically that models with phantom crossing can shift all of these inferred $Ω_m$ values toward mutual alignment. In contrast, quintessence theories restricted to $w \geq -1$ can alleviate the tensions with SN data but only at the cost of exacerbating the BAO-CMB discrepancy. We therefore conclude that it is the BAO and CMB measurements - not the SN data - that drive the preference for phantom crossing over quintessence in joint analyses. Moreover, we point out that SN data exhibit greater tensions with the other datasets when fit to phantom-crossing models than when fit to quintessence, causing the preference for phantom crossing to be weaker in joint CMB+BAO+SN analyses than in analyses of CMB+BAO data alone.

Dynamical dark energy theories generically introduce a time-dependent field that causes the accelerated expansion of the Universe on large scales. When embedding black hole solutions in such a cosmological space-time, this time dependence naturally gives rise to cosmological hair, i.e. the local black hole physics is no longer controlled by just the mass and spin of the black hole, but also impacted by the dark energy field. However, known such solutions are unstable. Focusing on the cubic Galileon as a concrete and illustrative example, we discuss the restrictions imposed on physical solutions by their regularity and stability in detail. We explicitly derive regular and stable solutions, that both recover the desired cosmological long-range behaviour and give rise to well-behaved short-range dynamics around black holes. We show how the nature of the scalar hair around these local black hole solutions encodes cosmological information, highlighting novel and tantalising prospects of directly probing cosmological dynamics with black hole observations.

Exceptional points (EPs) in quasinormal mode (QNM) spectra are non-Hermitian degeneracies at which both the eigenvalues and eigenfunctions coalesce. In this paper, we identify an EP in the scalar QNM spectrum of hairy black holes in the Einstein-Maxwell-scalar theory by scanning the parameter space. We then investigate its implications for ringdown signals by extracting QNMs from time-domain waveforms. Our results show that an EP ansatz, which includes a resonant contribution containing a term linear in time, provides a more robust description of ringdown at the EP than the standard ansatz based on a superposition of independent damped modes. In particular, it captures the resonant contribution associated with spectral coalescence more naturally and enables a more reliable extraction from the ringdown signal.

For strings of letters from a small alphabet, such as DNA sequences, we present a quantum encoding that empirically provides a strong correlation between the Levenshtein edit distance and the fidelity between quantum states defined by the encodings. It is based on the principles of Rotary Position Embeddings (RoPE), employed in modern large language models. Classically, this encoding yields RotorMap - a GPU-accelerated DNA mapping algorithm that achieves speedups of 50-700x over single-thread Minimap2 in proof-of-concept tests on human and maize genomes. For use on quantum devices, we introduce the Angular encoding, which is built from RoPE and directly outputs state preparation circuits. To verify its properties and utility on NISQ devices, we report results of experiments conducted on quantum computers from Quantinuum: the 56-qubit H2-1, H2-2 and the latest 98-qubit Helios-1. As a potential application, we consider a quantum DNA authentication problem and conjecture that a quantum advantage in one-way communication complexity could be achieved over any comparable classical solution.

We derive a dressed-state master equation in Lindblad form for two strongly coupled two-level atoms. The resulting decay dynamics are governed by Lindblad operators that couple different dressed states. We show that the eigenvalues and eigenvectors of the Liouvillian can be obtained in a compact form, since each off-diagonal element in the dressed-state basis constitutes an eigenvector. Depending on the interatomic distance and the atomic transition frequency, two distinct time scales emerge. On a short time scale, the system relaxes toward two states, one of which corresponds to a transient, maximally entangled configuration. On a longer time scale, this entangled state gradually decays to the steady state.

This thesis develops a general theoretical and numerical framework for achieving high-contrast atom interferometry based on double Bragg diffraction (DBD). While DBD offers intrinsic symmetry, reduced sensitivity to internal-state systematics, and suitability for microgravity experiments, its performance has long been limited by imperfect diffraction and contrast loss. This work overcomes these limitations by constructing an analytic Hamiltonian description of DBD -- including Doppler effects and polarization imperfections -- and by deriving reduced two- and five-level models via a truncated Magnus-expansion approach. These models clarify the origin of AC-Stark shifts, polarization-induced errors, and Doppler selectivity, and they provide accurate predictions for realistic input momentum distributions. Building on this theoretical foundation, the thesis introduces a tri-frequency laser scheme with dynamically tunable detuning and evaluates different detuning-control strategies using a five-level S-matrix formalism. Linear detuning sweeps and optimal-control pulses are shown to provide near-ideal beam-splitter and mirror performance, respectively, ensuring robust contrast across a wide range of experimental imperfections. Complementary full three-dimensional simulations using the GPU-accelerated Universal Atom Interferometer Simulator (UATIS) incorporate interacting Bose-Einstein condensates and realistic optical potentials, revealing transverse effects and polarization-induced distortions that extend the predictions of the one-dimensional non-interacting models. Taken together, this thesis establishes a coherent theoretical and numerical framework demonstrating that, with appropriate detuning control, double-Bragg atom interferometers can achieve the robustness required for precision inertial sensing and future space-based quantum tests of fundamental physics.

Ansatz selection is a key factor in the performance of variational quantum algorithms (VQAs). While much of the state-of-the-art still relies on heuristic choices, an inadequate circuit structure can compromise both the expressive power and the trainability of the resulting model. Recent results have also established theoretical connections between expressibility and the onset of barren plateaus, highlighting the need for systematic criteria for ansatz selection. In this work, the ansatz is treated as a design feature to be optimized rather than a fixed block, and a design space exploration (DSE) is performed over a diverse set of parametrized quantum circuits (PQCs). Three complementary metrics -- expressibility, trainability, and resource cost -- are evaluated and used to analyze the trade-offs that emerge across different PQCs. Beyond identifying Pareto-optimal candidates, this multi-objective perspective helps clarify the interplay between these metrics and contributes quantitative evidence toward understanding the expressibility--trainability tension in variational circuits.

Quantum simulations before fault tolerance suffer from the intrinsic noise present in quantum computers. In this regime, extracting meaningful results greatly benefits from stability against that noise. This stability, defined as an error in observables that is independent of the system's size, is expected in local systems under local noise. In fermionic systems, the encoding of the fermionic degrees of freedom into qubits can introduce non-locality, making stability more delicate. Here, we investigate the stability to noise of fermion-to-qubit mappings. We consider noisy quantum circuits in $D$ dimensions modeled by alternating layers of local unitaries and general, single-qubit Pauli noise. We show that, when using local fermionic encodings, expectation values of quadratic fermionic observables are stable to noise in states with spatially decaying correlations: a power-law decay with exponent $μ>D$ is sufficient for stability. By contrast, we show that this stability cannot be achieved by non-local encodings such as Jordan-Wigner in $2D$, or quasi-local ones such as the Bravyi-Kitaev transform. Our findings formalize the intuition that decaying correlations of the physical systems under study provide protection against noise for local fermionic encodings, and help inform design principles in near-term quantum simulations.

In this work we present a comprehensive analysis of a post-selective attack on quantum key distribution protocols employing phase-encoded linearly independent coherent states (or similar alternatives). The attack relies on multimode projection onto a Fock subspace and enables probabilistic extraction of information by an eavesdropper. We derive analytical expressions for the information accessible to the adversary and show that it depends only on three protocol parameters: the mean photon number of the signal states, the phase separation in the information basis, and the expected optical loss of the quantum channel. Several optical realizations of phase-encoded quantum key distribution protocols are analyzed to illustrate the applicability of the results. Possible countermeasures against the proposed attack are also discussed.

We present a standalone frequency-offset locking system for controlling narrow-linewidth lasers using off-the-shelf electronic components. We lock two frequency-doubled 1560 nm lasers to a stable primary laser operating at 780 nm via their optical beat note. This radio-frequency beat note is fed through a broadband variable divider, a frequency-to-voltage converter, and a proportional-integrator controller to lock each follower laser to a tunable offset frequency relative to the primary. This architecture provides a large capture range ($> 1$ GHz), fast response times ($< 1$ ms), and high linearity. We achieve a frequency resolution of 1.9 kHz and a short-term fractional frequency instability $10^{-11}/\sqrt{τ\rm (s)}$ at 780 nm without the need for a dedicated, precise clock reference. We perform high-resolution spectroscopy of cold $^{87}$Rb atoms to demonstrate the tunability and precision of our locking system. We designed the system to be modular and extensible, making it applicable to a wide variety of atomic physics experiments, including laser cooling, spectroscopy, and quantum sensing with atoms, ions, and molecules.

Large optical coherent-state superpositions are essential to advance quantum sensing, quantum repeaters and error-correction codes. We propose a deterministic feedforward protocol employing qubit-mode dispersive coupling, currently available in cavity quantum electrodynamics (QED). We show this single-mode protocol to outperform the advanced three-mode Gaussian-photon-number-resolving detector scheme both in terms of average fidelity and quantum non-Gaussian phase-space properties, and propose sensitivity to weak displacements of interference fringes as a feasible and conclusive witness of quantum interference. This approach combining QED with electro-optical feedforward is extendable to tailored states for applications and other platforms.

The decoding problem is a ubiquitous algorithmic task in fault-tolerant quantum computing, and solving it efficiently is essential for scalable quantum computing. Here, we prove that minimum-weight decoding is NP-hard in three quintessential settings: (i) the color code with Pauli $Z$ errors, (ii) the surface code with Pauli $X$, $Y$ and $Z$ errors, and (iii) the surface code with a transversal CNOT gate, Pauli $Z$ and measurement bit-flip errors. Our results show that computational intractability already arises in basic and practically relevant decoding problems central to both quantum memories and logical circuit implementations, highlighting a sharp computational complexity separation between minimum-weight decoding and its approximate realizations.

There are several inequivalent proposals in the literature for how to compute the probability distribution of the time that a detector registers for the arrival of a quantum particle. For two of these proposals, based on absorbing boundary conditions and imaginary potentials, we compute the predicted distribution for an experimental setup involving a single non-relativistic quantum particle with spin 0 or 1/2 in a wave guide along the $z$ axis with the detector waiting downstream. We find that the distribution shows signs of partial reflection of the wave function off of the detector; for a spin-1/2 wave function, it is independent of the initial spin orientation but does depend, for boundary conditions coupling to the spin, on the width of the wave guide. We also compare our predictions with the competing ones of Das and Dürr [arXiv:1802.07141].

The Drinfeld center fusion category $\mathcal{Z}(\mathrm{Vec}_G)$ famously models anyons in certain lattice models. Here we demonstrate how its fusion rules may also describe topological order in fractional topological insulator materials, in the vicinity of point defects in the Brillouin zone. Concretely, we prove that $\mathcal{Z}(\mathrm{Vec}_G)$ reflects, locally over a punctured disk in the Brillouin zone, the monodromy (topological order) of gapped quantum states over the parameter space of Bloch Hamiltonians whose classifying space has fundamental group $G$.

We present a state-of-the-art theoretical approach for computing bound-state energies in muonic atoms, incorporating improved quantum electrodynamics effects and nuclear polarization corrections with a systematic assessment of theoretical uncertainties. Our approach is based on a combination of the $Zα$-expansion and the all-order formalism (Furry picture) optimized for the medium-mass range $(3 \leq Z \lesssim 30)$ and guided by the accuracy requirements of modern muonic spectroscopy experiments. These calculations are directly relevant to ongoing and forthcoming measurements aimed at extracting nuclear structure parameters, particularly nuclear charge radii, with unprecedented precision.

We demonstrate violation of objective realism in quantum world using unconstrained weak measurements. Instead of limited Leggett-Garg approach with artificial bounds on the observed values, we assume two identical and indepenent weak detectors and final conditioning. The experimental verification has been performed on public quantum computers, IBM and IonQ. Thanks to sufficiently large statistics, the violation is observed at the level of 10 standard deviations. The tests confirmed also high quality of parametric two-qubit gates offered by main quantum hardware providers.

This document sets out the conventions used for data simulations, waveforms, and analysis pipelines within the Distributed Data Processing Centre (DDPC) of the Laser Interferometer Space Antenna (LISA). It can also be considered a best practice guide for all publications related to the LISA mission. Topics covered include time-to-frequency transformations, gravitational-wave source parametrization, the instrumental response to gravitational waves, time-delay interferometry, and reference frame definitions.

Fixed-frequency transmon qubits, characterized by simple architectures and long coherence times, are promising platforms for large-scale quantum computing. However, the rapidly increasing frequency collisions, which directly reduce the fabrication yield, hinder scaling, especially in cross-resonance (CR) gate-based architectures, wherein the restricted drive frequency severely limits the available design space. We investigate the Stark-induced ZZ by level excursions (siZZle) gate, which relaxes this limitation by allowing arbitrary drive-frequency choices. Extensive numerical analyses across a broad parameter range -- including the far-detuned regime that has received negligible prior attention -- reveal wide operating windows that support controlled-Z (CZ) fidelities >99.6%. Leveraging these windows, we design lattice architectures containing >1000 qubits, showing that even under 0.25% fabrication-induced frequency dispersion, the zero-collision yields in square and heavy-hexagonal lattices reach 80% and 100%, respectively. Thus, the siZZle-CZ gate is a scalable and collision-robust alternative to the CR gate, offering a viable route toward high-yield fixed-frequency transmon quantum processors.

We develop a novel analytical dynamical analysis to derive precise energy density ratio evolutions for the $φ$CDM and $polyΛ$CDM models, comparing them to the standard $Λ$CDM model and validating against numerical solutions. Analytical solutions for the quintessence, i.e. $φ$CDM, show sub percent agreement with $Λ$CDM with greater reliability than numerical integration of stiff systems. The $polyΛ$CDM model, a phenomenological modified gravity framework, captures radiation, matter, dark energy, and exotic epochs, offering a streamlined yet comprehensive alternative to existing studies. Its dynamics reveal a global transition from a dark energy-dark matter exchange reflector, through saddle points of matter, radiation, curvature, and modified gravity, to an SVT modified gravity attractor-saddle, and finally to a cosmological constant attractor in the far future, with saddle transitions between modified gravity components. The $polyΛ$CDM model integrates modified gravity models, using dynamical analysis to distinguish observationally viable critical points and differentiate gravity epochs. All three models align with observed cosmic evolution, but $polyΛ$CDM richer phenomenology provides deeper insights into modified gravity dynamics. Code available at GitHub.

While recent kinematically irrotational warp-drive spacetimes remove the Hawking--Ellis Type IV algebraic pathologies of earlier models, the source irrotational background still exhibits residual null-energy-condition violation and hence residual weak, dominant, and strong energy-condition violation. Motivated by the physical precedent of vacuum birefringence, we ask whether these exotic-stress deficits can be mitigated by replacing the metric vacuum with a weakly birefringent area-metric deformation. Within the Schneider--Schüller--Stritzelberger--Wolz (SSSW) closure framework, this becomes a perturbative admissibility question: can the residual mixed-sector transport of the irrotational Hawking--Ellis Type I background be absorbed while remaining inside the locally hyperbolic weakly birefringent regime? A minimal screening container in the restricted ansatz class studied here is a symmetric $3\times 3$ coefficient block on the meridional bivector-index sector, whose sparse image in the 17-variable SSSW parametrization occupies 6 weakly birefringent variables. This yields a derived reduced working model coupled to the exact background transport system. Exploratory integration, using the irrotational profile and a shear-based lift envelope, is reported at the benchmark angles $θ=0,π/4,π/2$. The mixed-sector tilt vanishes on the $θ=0$ symmetry axis. In the representative coupled case $θ=π/4$, it is approximately cubic only for mild velocities $v\lesssim 1$; at higher velocity it departs strongly from that trend, and $φ_{17}$ changes sign between $v=2$ and $v=3$. At $θ=π/2$, by contrast, the benchmark values follow the exact decoupled reduced-model cubic law. The resulting evidence disfavors fast walls within the reduced perturbative model, while indicating that smaller subluminal velocities are less strained by the reduced admissibility conditions.

We present a geometric formulation of quantum mechanics based on the symplectic structure of the projective Hilbert space. Building upon the standard Kähler framework, we introduce an extension in which the symplectic structure is allowed to couple to a metric-affine background geometry, leading to a deformation of the Hamiltonian flow on the state space. We show that, under suitable conditions, the deformed structure remains symplectic and defines a well-posed Hamiltonian system. The formulation reduces to standard Schrödinger dynamics in the limit where the geometric deformation vanishes. Explicit analytical examples are constructed to illustrate the effect of the deformation. In particular, curvature-dependent deformations lead to a rescaling of Hamiltonian flows, while torsion-induced contributions produce direction-dependent corrections. In addition, geometric phases acquire corrections determined by the deformed symplectic structure. These results provide a mathematically consistent framework for exploring geometric modifications of quantum evolution induced by background curvature and affine structure.

Contextuality is widely regarded as a hallmark of quantum information, yet its structural origin is often obscured by probabilistic or operational formulations. In this work, we show that non-distributive orthomodular structure need not be postulated, but arises canonically as a left adjoint from classical Boolean contexts. We introduce a gluing functor that takes pairs of Boolean algebras and identifies only their minimal and maximal elements via a categorical pushout. The resulting lattice is orthomodular but generically non-distributive. We prove that this construction is left adjoint to a forgetful functor selecting Boolean subalgebras, thereby providing a free but constrained generation of quantum-logical structure from classical contexts. Furthermore, we demonstrate that the failure of this pushout to remain Boolean is equivalent to the absence of global sections in the sheaf-theoretic framework of Abramsky and Brandenburger. This establishes a precise correspondence between contextuality as a sheaf obstruction and non-distributivity as a colimit failure. Our results offer a categorical and lattice-theoretic reconstruction of contextuality that precedes probabilistic notions and clarifies the structural necessity of quantum logic in information-theoretic settings.

Characterizing quantum magic -- the resource enabling computational advantage beyond stabilizer circuits -- is subtle in qubit systems because established measures can give conflicting information about the same state. We introduce C(rho), the l1 distance from a state's discrete Wigner function to the convex hull of stabilizer Wigner functions, and study its relationship to the stabilizer extent Gamma(rho) via the tightness ratio kappa(rho) := (Gamma(rho)-1)/C(rho). For three two-qubit families in the repetition-code subspace span{|00>,|11>}, we prove kappa takes exact integer values constant over each family: kappa=1 for the R_y and Bell+R_z families, kappa=2 for the R_x family. The factor-of-2 gap arises because imaginary coherence concentrates Wigner negativity at 2 of 16 phase-space points rather than 4, leaving Gamma unchanged. The optimal dual witnesses are logical Pauli operators of the repetition code, revealing that C is a fault-tolerant observable invariant under correctable errors -- an unexpected connection between phase-space geometry and quantum error correction. We prove a sharp bound Gamma >= 1 + C/M_n, establish a hemispheric dichotomy in tensor-product behavior where superadditivity of C fails for northern-hemisphere states with deficit approximately 0.335 C(rho), and show C is not a magic monotone under the full Clifford group, so asymptotic distillation rates require Gamma.

Open quantum batteries (QBs) operate under unavoidable system-environment interactions, where both dissipation and coherent renormalization influence their performance. While most previous studies focus on dissipative effects, the role of environment-induced frequency renormalization, such as the Lamb shift, remains insufficiently explored.In this work, we investigate an externally driven QB composed of two coherently coupled quantum harmonic oscillators, representing the charger and the battery. By incorporating both dissipation and Lamb-shift corrections within a Lindblad master equation, we show that the Lamb shift effectively renormalizes the system eigenfrequencies and thereby modifies the resonance condition with the external drive. We demonstrate that tuning the driving frequency relative to the renormalized eigenmodes leads to a mode-selective energy transfer process, resulting in a controllable redistribution of energy between the charger and the battery. This behavior manifests as a switching of the dominant energy storage channel and can be quantitatively understood through a supermode decomposition of the coupled system. Our results clarify the dynamical role of environment-induced frequency shifts in open quantum batteries and provide a physically transparent framework for optimizing work extraction under realistic operating conditions.

We establish an exact noise-model-derived characterization of quantum error correction under diagonal local phase noise. Under uniform locality, the maximal logical dimension under t-local phase errors equals Aq(n,2t+1), the classical q-ary packing function. Because no affine or stabilizer structure is imposed, nonlinear spectral supports achieve this bound and strictly exceed all affine constructions whenever Aq(n,2t+1)>Bq(n,2t+1). This follows from a harmonic translation principle: diagonal phase operators act as rigid translations in the Fourier domain, reducing the Knill-Laflamme conditions exactly to an additive non-collision constraint (S-S) cap Et={0}. For structured phase noise, exact correction is equivalent to independence in an additive Cayley graph, connecting biased quantum capacity to classical zero-error theory and the Lovasz theta function. Under mixed Pauli noise, simultaneous protection in conjugate domains incurs an intrinsic rate penalty R <= 1-(gamma_X+gamma_Z)/2, exposing a discrete harmonic uncertainty principle. In contrast with stabilizer- or graph-based frameworks, this classical correspondence is derived directly from the phase-noise model itself rather than from an auxiliary algebraic construction.

We study the Hall viscosity of the quark gluon plasma (QGP) created in non-central heavy-ion collisions. In the presence of a strong magnetic field or vorticity, rotational symmetry is broken from O(3) to O(2), allowing for two independent Hall viscosities associated with shear deformations transverse and parallel to the symmetry-breaking direction. We find the corresponding constitutive relations by extending the kinetic-theory mechanism to three spatial dimensions and provide parametric estimates of the Hall viscosities under realistic QGP conditions. Both kinetic-theory and holographic estimates indicate that Hall viscosities are comparable in magnitude to the shear viscosity at zero magnetic field. We further show that Hall viscous stresses at hydrodynamic initialization can be as large as standard viscous corrections and identify observable consequences in flow and event-plane correlations.

A search for new particles decaying to top quark-antiquark pairs is performed using proton-proton collision data at a centre-of-mass energy of 13 TeV. The data set recorded with the CMS detector between 2016 and 2018 is used, corresponding to an integrated luminosity of 138 fb$^{-1}$. Final states with 0, 1, and 2 leptons are analyzed, covering all decay modes of the top quark-antiquark pairs. Heavy Z' bosons with relative widths of 1, 10, and 30% are excluded for masses in the ranges 0.4$-$4.8, 0.4$-$6.2, and 0.4$-$7.4 TeV, respectively. A Kaluza$-$Klein gluon in the Randall$-$Sundrum model and a dark-matter mediator are excluded for masses between 0.5$-$5.5 and 1.0$-$4.2 TeV, respectively. These results set the most stringent limits to date for the considered models in the $\mathrm{t\bar{t}}$ final state. In addition, in the two-Higgs-doublet models, upper limits are set on the coupling strength modifier for scalar and pseudoscalar Higgs bosons with relative widths of 2.5, 10, and 25% in the mass range of 0.5$-$1.0 TeV.

I develop a mixed-space formulation of high-energy deep-inelastic scattering in the shock-wave formalism at sub-eikonal order. Starting from the quark propagator in the background field, I derive the corresponding mixed-space Feynman rules from the LSZ reduction formula in the presence of a shock wave, and compute the sub-eikonal corrections to unpolarized and polarized structure functions. I show that the first sub-eikonal correction to the small-$x_B$ formalism already reproduces the standard quark and helicity light-ray operator structure of DIS and, in the inclusive limit, recovers the corresponding distributions at nonzero Bjorken $x_B$. I then calculate the cross section in dipole form and obtain the sub-eikonal corrections to $F_L$, $F_T$, and $g_1$, identifying the operator combinations relevant in the flavor singlet and non-singlet sectors. The resulting operator basis differs from that used in previous approaches and is naturally organized in terms of dipole operators which vanish in the zero-size limit, making the unitarity property and the small-dipole behavior manifest. I analyze the divergence structure of the sub-eikonal contributions and show that the logarithmic singularities of the transverse and helicity-dependent structure functions are precisely those generated by the one-loop evolution of the corresponding sub-eikonal operators. In the double-logarithmic approximation, I solve the evolution equations, recover the known fixed-coupling ladder exponent, and clarify its relation to previous small-$x_B$ helicity resummations. This provides an explicit operator-level connection between the sub-eikonal Wilson-line description of high-energy DIS and the light-ray operator structure underlying quark and helicity distributions.

We analysis the $\sqrt{s}$-dependent line shapes of the $e^+e^-\to D_s^{+}D_{s1}^{*-}(2536)$, $D_s^{+}D_{s2}^{*-}(2573)$, $φχ_{c1,2}$, $K^+K^-J/ψ$, $K_S^0 K_S^0 J/ψ$, and $K^+K^-ψ(2S)$ cross sections measured by the BESIII experiment using the four resonant structures $ψ(4230)$, $ψ(4500)$, $ψ(4660)$, and $ψ(4710)$, by performing a simultaneous $χ^2$-minimized fit. Their masses and widths are obtained. We find that the processes $e^+e^-\to D_s^{+}D_{s1}^{*-}(2536)$, $e^+e^-\to D_s^{+}D_{s2}^{*-}(2573)$, and $e^+e^-\to φχ_{c1,2}$ are all dominantly produced via the $ψ(4660)$ and $ψ(4710)$ decays.

Explaining the observed charged fermion mass hierarchies points to flavour symmetries inducing a suppression of the lighter species' masses. When the symmetries are global, it is expected that such symmetries are broken by gravity via Planck scale suppressed effective operators. The potential of the spontaneous symmetry-breaking "flavon" field, if the symmetry is discrete, then possesses several minima, with the vacuum-degeneracy lifted by the gravity effects. In such scenarios, domain walls might be generated in the process of symmetry breaking. Due to the bias, however, they potentially annihilate sufficiently before Big Bang nucleosynthesis, avoiding conflict with observations and generating a characteristic contribution to the stochastic gravitational wave background. We discuss whether and how minimalistic supersymmetric and non-supersymmetric realisations of such theories can give rise to observable gravitational waves.

We analyse the gauge-dependence of the effective action in an interacting quantum theory of gravitational and matter fields. An explicit off-shell result is obtained in a general background gauge at one-loop order, which reduces in a particular gauge to the effective action found by 't Hooft-Veltman. We confirm the validity of DeWitt-Kallosh theorem, which implies that the on-shell effective action should be independent of the gauge-fixing parameter. We employ this theorem to expose the non-renormalizability of the theory in a general background gauge.

The discovery of hidden-charm pentaquarks and open-flavor tetraquarks motivates the search for even more exotic hadron configurations. In this work, we investigate genuinely exotic molecular pentaquark candidates comprising five different flavors, focusing on the $Ξ_b^{(\prime,\,*)} \bar D^{(*)}$ and $Ξ_c^{(\prime,\,*)} B^{(*)}$ systems. Employing the one-boson-exchange model with the $S$-$D$ wave mixing and coupled-channel dynamics, we identify the most promising molecular pentaquark candidates comprising five different flavors. These include the $Ξ_b \bar D$, $Ξ_b^{\prime} \bar D$, $Ξ_c B$, and $Ξ_c^{\prime} B$ states with $I(J^P)=0(1/2^-)$, the $Ξ_b \bar D^{*}$, $Ξ_b^{\prime} \bar D^{*}$, $Ξ_c B^{*}$, and $Ξ_c^{\prime} B^{*}$ states with $I(J^P)=0(1/2^-,\,3/2^-)$, the $Ξ_b^{*} \bar D$ and $Ξ_c^{*} B$ states with $I(J^P)=0(3/2^-)$, as well as the $Ξ_b^{*} \bar D^{*}$ and $Ξ_c^{*} B^{*}$ states with $I(J^P)=0(1/2^-,\,3/2^-,\,5/2^-)$. Importantly, these loosely bound states exhibit pronounced spin splittings across different total angular momentum configurations after incorporating the spin-dependent interactions or the channel couplings. In addition, we identify several possible isovector molecular pentaquark candidates within the $Ξ_b^{(\prime,\,*)} \bar D^{(*)}$ and $Ξ_c^{(\prime,\,*)} B^{(*)}$ systems. Our predictions provide clear targets for experimental searches at facilities such as LHCb and Belle II, where the unique five-flavor quark configuration offers a distinctive experimental signature.

The soft-breaking parameter, $m_{12}^2$, frequently appearing in the 2HDM scalar potential is much more remarkable than being just a nonstandard parameter that helps make the BSM scalars super heavy. In fact, as we show through explicit calculations, it should be treated as the direct but concise embodiment of new non-electroweak spontaneous symmetry breaking effects at very high energy scales, wherein lies its quiddities. Consequently, it is argued that $m_{12}^2$ and the electroweak VEV serve as two distinct sources for the nonstandard scalar masses, which are completely unrelated to each other. Such distinctions allow us to define parameters that conveniently capture the fraction of the nonstandard scalar masses derived from the electroweak VEV. Finally, we demonstrate that constraints can already be placed on such fractions from the current measurements of the diphoton signal strength and from direct searches of new nonstandard scalar resonances in the diphoton channel.

We study the Dirac operator on a finite warped cylinder coupled to a background $U(1)$ gauge field. We identify the intrinsic endpoint operators defining the Atiyah-Patodi-Singer (APS) boundary condition and derive a determinant characterization of the modewise APS spectrum. In the constant-gauge, invertible setting, the endpoint reduced $η$ contributions cancel, so the APS index vanishes. For smooth gauge families, the APS projector becomes discontinuous when a boundary mode crosses zero. We therefore introduce a regularized APS-type family of self-adjoint endpoint conditions that remains continuous across such crossings. This regularized family admits a real-symplectic boundary formulation within the standard spectral-flow/Maslov framework: for nondegenerate regularization, the zero-mode set coincides with the boundary-zero set, and transverse boundary zeros give isolated regular crossings.

Sterile neutrinos in the keV mass range are a well-motivated extension of the Standard Model and viable dark matter candidates. Their existence can be probed in laboratory experiments, as the admixture of a sterile state would induce a characteristic kink-like distortion in the $β$-decay electron energy spectrum. The KATRIN experiment is designed to measure the effective electron neutrino mass with sub-eV sensitivity by analyzing the endpoint region of the tritium $β$-decay spectrum. Following the completion of its neutrino mass program, KATRIN will extend its physics reach to the search for keV-scale sterile neutrinos. This effort will be enabled by the TRISTAN detector, a newly developed silicon drift detector array optimized for differential measurements at high rates and energies well below the endpoint. In this article, we present the projected sensitivity of KATRIN to keV-scale sterile neutrinos using a dedicated simulation framework. With four months of detector livetime, KATRIN has the statistical power to probe mixing amplitudes at the level of $|U_{e4}|^2 \sim 10^{-6}$ for sterile neutrino masses in the (4$-$13) keV range, significantly extending the reach of previous laboratory searches. The major experimental systematic uncertainties investigated in this work reduces the sensitivity by a factor of 10$-$50 over the same mass range.

The type-I seesaw model with two-zero textures of the Majorana mass matrix for the right-handed neutrinos $M^{}_{\rm R}$ provides a highly predictive framework for neutrino mass generation and baryon asymmetry of the Universe via leptogenesis. In this work, we systematically investigate the compatibility of viable two-zero textures of $M^{}_{\rm R}$ with leptogenesis within the type-I seesaw scenario. We consider both a general diagonal Dirac neutrino mass matrix $M^{}_{\rm D}$ and two theoretically motivated special cases, namely the SO(10) GUT-inspired $M^{}_{\rm D} \sim \mathrm{diag}(m_u,m_c,m_t)$ and the flavor symmetry-induced $M^{}_{\rm D} \propto I$. Our study shows that two-zero textures of $M^{}_{\rm R}$ yield strong correlations between neutrino parameters and leptogenesis, offering distinctive phenomenological implications for neutrino flavor physics and baryon asymmetry of the Universe.

A generic aspect of low-energy effective field theories (EFTs) coming from string compactifications is the appearance of moduli fields. Among these moduli, the axion and dilaton are present as (pseudo-) Goldstone bosons from the spontaneous breaking of an exact (or approximate) global symmetry. These moduli have a different microscopic coupling to matter but appear kinetically coupled in such a way that their interaction can compete with gravity at low energies and have an important effect in strong gravity environments. In this talk, we will discuss some of the astrophysical implications of a stringy-inspired multi-scalar-tensor theory. In particular, we show the numerical solution of the Tolman-Oppenheimer-Volkov (TOV) system of equations, necessary to probe the existence of a screening mechanism that reduces the Brans-Dicke dilaton coupling to macroscopic matter sources such as a neutron star.

We present the first calculation of the Efremov-Teryaev-Qiu-Sterman functions and associated twist-3 quark-gluon correlation functions for both the proton and pion. These functions are determined using the light-front wave functions obtained by diagonalizing a light-front effective Hamiltonian within a Fock space truncated to include a dynamical gluon. We compute the twist-3 correlations in the hard-pole region and extrapolate them to the soft-gluon pole limit. After the scale evolutions, our predictions demonstrate quantitative consistency with recent experimental extractions, providing a unified description of the Sivers and Boer-Mulders asymmetries from a light-front Hamiltonian approach.

We propose a heuristic for the brick wall in AdS/CFT: the location where a boundary mode's local bulk energy reaches a (Planckian) UV cut-off. This accomplishes two things: (a) the brick wall is framed as a breakdown criterion for bulk effective field theory, and (b) the definition is boundary-anchored rather than horizon-anchored, aligning it with holography. Near the horizon, spacetime effectively gets cut-off due to blueshift relative to the boundary, and leads to normal modes. By directly computing these new modes for the BTZ black hole, we show that they are qualitatively unchanged from conventional 't Hooftian brick wall normal modes in the relevant part of the spectrum -- successfully reproducing black hole thermodynamics and exterior smooth-horizon correlators, under similar approximations. However, unlike 't Hooft's (and our own previous) calculations, we also do an $exact$ numerical evaluation of the normal mode partition function. This allows us to identify a "little hierarchy" problem in the brick wall paradigm, irrespective of whether it is horizon-anchored or boundary-anchored: because the modes are not exactly degenerate in the $J$-direction, the coefficient of the area law is slightly subleading, unless the brick wall is slightly trans-Planckian. One way to evade the problem is to increase the number of active species. While this is certainly a possibility in string theory, we argue that a natural resolution is to take into account the degrees of freedom intrinsic to the (stretched) horizon, as suggested by the recent results in arXiv:2601.18775. We argue that this will lead to a dominant contribution from a quantum number associated to the radial direction, while retaining the successes of the $J$-degenerate toy model. We discuss the possible significance of these observations for (a) quantum chaos in black holes, and (b) the fuzzball program.

Using $(10087 \pm 44)\times 10^{6}$ $\jpsi$ events collected with the BESIII detector, we perform the first amplitude analysis of the process $\jpsi\toγη\piz$. The decay is dominated by the intermediate processes $\jpsi\to\piz \bo \left( \toγη\right)$, $\jpsi\to\pizρ(1450)^0 \left( \toγη\right)$ and $\jpsi\toηh_1(1170) \left( \toγ\piz\right)$. Contributions from $\jpsi\toγa_0(980)^0(\toη\piz)$, $\jpsi\toγa_2(1320)^0(\toη\piz)$ and $\jpsi\toγa_2(1700)^0(\toη\piz)$ are observed with a statistical significance exceeding $5σ$, constituting the first observation of radiative transitions of $\jpsi$ to isospin-triplet scalar mesons. The total branching fraction of $\jpsi\toγη\piz$ is measured to be \num{25.7\pm0.3\pm1.5e-6}, where the first uncertainty is statistical and the second systematic. This result is consistent with the previous measurement, with the precision improved by more than a factor of two.

We derive new integral identities for AdS propagators and further develop the Wilson network expansion for AdS Feynman diagrams. In particular, we demonstrate that four-point contact and exchange scalar diagrams in two dimensions can be expanded into several infinite series of matrix elements of Wilson line network operators with running conformal weights. Each series is characterized by specific multi-trace operators associated with the external and intermediate edges of the corresponding graphs. The resulting expansions near the conformal boundary reproduce the well-known decompositions of the corresponding four-point Witten diagrams into conformal blocks.

The origin of neutrino mass and the nature of dark matter (DM) remain unresolved puzzles in particle physics, and an appealing possibility is to address both in a unified picture. This paper explores a light fermionic DM candidate within the scotogenic inverse seesaw model, which can simultaneously provide a mechanism for neutrino mass generation. By incorporating constraints from neutrino oscillation data, charged lepton flavor violating processes, invisible decays of the Higgs and $Z$ bosons, DM relic density, and direct detection of DM, we uncover a light fermionic DM window in the mass range $58\,{\rm GeV} \lesssim m_{\tt DM} \lesssim 63\,{\rm GeV}$ that can satisfy all of the aforementioned constraints. We find that this window can be jointly tested by next-generation ton-scale DM direct detection experiments including PandaX-xT and XENONnT, Higgs invisible decays, and future lepton colliders such as ILC.

Using the Drell-Levy-Yan relation, the pion and kaon elementary fragmentation functions (EFFs) are obtained from their hadron-scale parton distribution functions (DFs). These EFFs serve as driving terms in the hadron cascade equations, whose solution yields the complete array of hadron-scale fragmentation functions (FFs) for pion and kaon production in high energy reactions. Evolved to experimental scales, the continuum Schwinger function methods (CSMs) predictions satisfy QCD endpoint behavior: nonsinglet FFs vanish at $z=0$, singlet FFs diverge faster than $1/z$. Jet multiplicity predictions reveal SU(3) symmetry breaking in the charged/neutral kaon ratio, decreasing with energy, and show the pion/kaon ratio in $e^+e^-$ collisions asymptotes to a mass-independent value.

Direct detection of nuclear recoils due to sub-GeV dark matter is challenging because of the small kinetic energy of the light dark matter particles. Although limits down to a few hundred MeV have been reached using specially designed low threshold detectors, further improvements are now constrained more by background event rates than by energy thresholds. However, constraints down to sub-MeV dark matter masses can still be obtained through the boosted dark matter framework. In this scenario, high-energy cosmic rays or neutrinos scatter off dark matter particles, imparting additional kinetic energy and boosting them beyond the typical velocities expected from the non-relativistic dark matter halo. These boosted dark matter particles can then be detected even by experiments with higher energy thresholds. In this work, we present a catalog of dark matter - nucleon scattering cross sections corresponding to a heavy mediator limit for spin zero, one half and one dark matter and for scalar and vector mediators with even or odd parity. Based on these results, we present projected constraints on the dark matter - nucleon cross section for the COSINUS experiment, assuming an exposure of 100 kg d, demonstrating the potential sensitivity to sub-GeV boosted dark matter.

We study the evolution of neutrinos electroweakly interacting with a rotating matter. The description of neutrinos is based on the Dirac equation in the corotating noninertial frame where matter is at rest. We find solution of this Dirac equation, where the matter angular velocity is accounted for exactly, for massless neutrinos. In case of massive particles, this solution is obtained for a slowly rotating matter. Our findings are compared with previous research. We consider two applications of our results. First, we compute the electroweak contribution to the vector current of neutrinos along the rotation axis, which is analogous to the chiral vortical effect. This current is shown to be nonzero for both massless and massive particles. Then, we take into account the nonzero mixing between different mass eigenstates. It allows us to study neutrino flavor oscillations in rotating matter and account for noninertial effects. We derive the transition probability which reveals the resonance. These findings generalize the description of neutrino oscillations in a nonmoving matter. Some astrophysical applications are briefly discussed.

A strongly self-interacting component of asymmetric dark matter (DM) particles can form compact dark stars (DSs). These objects have a broad spectrum of masses and radii, with distinct evolution histories from both neutron stars and black holes (BHs). We argue that these differences allow a population of DSs to contribute significantly to the astrophysical merger rate in unique and discernible ways. Specifically, their merger rate could dominate at low redshifts over other sources, while their mass function may populate windows outside known astrophysical processes. We investigate the structure and formation of DSs within a dissipative model, and calculate the enhancement of their merger cross-section due to tidal deformation effects. From this, we derive the present-day merger rate and its differential mass distribution. These findings open a new window to probe DM substructure and particle interactions through present and future gravitational wave (GW) observatories.

We apply the light-cone QCD sum rules with chiral currents to compute transition form factors of the semileptonic decay of the charmed and bottom scalar mesons $D/B$ to $a_{0}(980)l^{+}ν_{l}$ $(l=e$, $μ)$, where the scalar meson $a_{0}(980)$ are firstly regarded as two quark states, and contributions from the tetraquark component are then added via proper modulation of four distribution amplitudes considered. The obtained transition form factors and branching fraction are free of the contribution from the light-cone distribution amplitudes at the level of twist-three and two simple relations connecting their form factors are obtained. Our computations indicates that the decay branch fractions are on the margin of the measurements reported by Beijing Spectrometer \uppercase\expandafter{\romannumeral3} in the pure two-quark scenario and are in good agreement with observations when tetraquark component is considered.

A virtual production of heavy axion-like particles (ALPs) via light-by-light scattering in pp, pPb and Pb collisions at the future 100 TeV collider FCC-hh is studied. Both differential and total cross sections are calculated. The 95\% C.L. exclusion limits, as well as $3σ$ and $5σ$ discovery limits on an ALP coupling constant versus ALP mass $m_a$ are given, using integrated luminosities of 30 ab$^{-1}$, 27 pb$^{-1}$ and 110 nb$^{-1}$. Our results are compared with the current LHC bounds. The strongest limit on the ALP coupling is obtained if $m_a \simeq 250$ GeV for the PbPb collisions, and if $m_a \simeq 1$ TeV for pp or pPb collisions. It allows us to conclude that the FCC-hh has a great physics potential of searching for the heavy ALPs.

A quantitative investigation into the modification of ground-state field structures in two-color QCD (QC$_2$D) is presented at finite chemical potential. Using lattice simulations with Wilson gauge and fermion actions, we explore the chromo-electromagnetic field strengths under varying matter densities. To ensure accurate measurements, we develop and calibrate two highly improved topological charge operators and evaluate four gradient flow actions. Our results reveal a finite-volume crossover in the regime of the anticipated phase boundary at $μ= m_π/2$, with both chromo-electric and chromo-magnetic field strengths suppressed before recovering and exceeding vacuum values at higher chemical potentials. We find the difference between the squared chromo-electric and chromo-magnetic field strengths, $E^2-B^2$, to increase in magnitude monotonically with increasing chemical potential. At $aμ=0.7$, we find an $11\%$ suppression of $E^2$, a relatively small effect. A systematic analysis using sigmoid fits of lattice simulations in the crossover regime is performed to confirm the critical chemical potential obtained from the field structure is in agreement with the phase boundary at $m_π/ 2$. These findings provide new insight into non-Abelian ground-state vacuum field structures and offer a foundation for future studies in real QCD.

A search for the radiative decays $D^0\to γ\bar K_1(1270)^0$ and $D^+\to γK_1(1270)^+$ is conducted using $20.3~\mathrm{fb}^{-1}$ of $e^+e^-$ annihilation data collected at the center-of-mass energy $\sqrt{s}=3.773$ GeV by the BESIII detector operating at the BEPCII collider. No significant signals are observed, and upper limits on the branching fractions of $D^0\to γ\bar K_1(1270)^0$ and $D^+\to γK_1(1270)^+$ at 90\% confidence level are determined to be $7.7\times10^{-4}$ and $3.9\times10^{-5}$, respectively. This represents the first test of the Vector Meson Dominance mechanism in the radiative decays of charmed mesons to axial-vector mesons.

We compute holographic one-point functions for Wilson surfaces in the case of a toroidal surface operator. Compared to the cases of a planar or spherical surface operator, these one-point functions exhibit a more intricate dependence on the shape and position of both the surface and the local operators. Averaging over the moduli space of membranes dual to the surface operator plays a key role in the computations. We obtain both analytical and numerical results. The case of a cylindrical surface operator is also studied.

This is a continuation of the previous paper (arXiv:2508.01944) in this series. We recontextualise Cirio and Martins' work to motivate our fundamental conjecture that the Drinfeld-Kohno (Lie) 2-algebra has trivial cohomology. It is then shown that this conjecture implies the following: given a coherent totally symmetric infinitesimal 2-braiding $t$, every modification endomorphic on the zero transformation vanishes if it is made up of the four-term relationators and whiskerings by $t$. The power of such an implication is that, in our context, one need only construct the data of a braided monoidal 2-category and it will automatically satisfy the axioms. We thus conclude by constructing the pentagonator via Cirio and Martins' Knizhnik-Zamolodchikov 2-connection over the configuration space of 4 distinguishable particles on the complex line, $Y_4$. In particular, we make use of Bordemann, Rivezzi and Weigel's pentagon in $Y_4$.

The CYGNO collaboration is developing a novel strategy for directional Dark Matter searches based on a gaseous Time Projection Chamber (TPC). The detector is optimized for the exploration of light (0.5-50 GeV) WIMPs-like particles and employs a He/CF4 gas mixture at atmospheric pressure, sensitive to both spin-dependent and spin-independent interactions. A key feature of the project is its optical readout, which relies on photon detection rather than charge collection. In CYGNO detectors, electrons released by ionizing tracks drift toward an amplification stage of three Gas Electron Multipliers (GEMs). The electron avalanches generate scintillation light that is captured by scientific CMOS (sCMOS) cameras for high-resolution two-dimensional imaging and by Photomultiplier Tubes (PMTs) that provide a precise time profile along the drift direction. This allows a 3D event reconstruction, detailed energy deposition mapping, and effective topology and head-to-tail discrimination. Building on the achievements of the 50 L prototype (LIME), which successfully operated underground at LNGS, the next step is the deployment of a 0.4 m3 demonstrator, CYGNO-04, to be completed in 2026. The demonstrator will validate scalability and confirm the advantages of the proposed technique. Recent results from LIME highlight strong progress in 3D tracking and particle identification. The current status of CYGNO-04 and its role in advancing the program will be presented as well.

We present a search for muon neutrino disappearance in the Booster Neutrino Beam (BNB) at Fermilab using the ICARUS detector. Neutrino interactions identified as muon neutrinos interacting with argon nuclei via the charged current interaction and having only a muon and at least one proton in the final state (1$μ$Np) have been selected from data collected in 2022-2023 (ICARUS Run 2) and compared with a simulation-based expectation. In the context of a fit to a two-neutrino approximation of the sterile 3+1 model, including the impact of systematic uncertainty from the flux, neutrino interaction, and detector models, we find no statistically significant muon neutrino disappearance at the ICARUS baseline of 600 meters from the BNB target. Corresponding 90% C.L. exclusion contours in $Δm^2_{41}$ - sin$^22θ_{μμ}$ space are presented. This is the first oscillation analysis produced by ICARUS exposed to the BNB. We note that the analysis is systematics limited due to large unconstrained uncertainties from the flux and interaction models. In future joint analyses, data from ICARUS and the SBND detector, exposed to the BNB at 110 meters from target, will be combined to provide significant constraint of these uncertainties, enabling a robust, world-leading two-detector analysis.

We investigate the enhanced dynamically assisted electron-positron pair production in differently polarized electric fields with frequency chirps within the real-time Dirac-Heisenberg-Wigner formalism. The combined influence of the chirp parameter and the field polarization on the momentum distribution and the total number density of the created pairs is studied in detail for one-color field as well as dynamically assisted two-color combined fields. The frequency chirps lead to strong interference effects and significantly enhanced the peak values in the momentum distribution for both one-color field and two-color combined fields. In the dynamically assisted case, the number density can be enhanced significantly over 2-3 orders when large frequency chirp is applied to both strong and weak fields. Furthermore, we observe that sensitivity of the number density to field polarization progressively diminishes as the chirp parameter increases, a trend that holds for both one-color field and the assisted two-color combined fields. These results provide a valuable foundation for the optimal control of pair production, offering guidance for maximizing particle yield within a constrained set of field parameters.

The NA61/SHINE spectrometer at the CERN Super Proton Synchrotron (SPS) scans particle production in collisions of nuclei with various sizes at a set of energies covering the SPS energy range towards various physics goals. This paper presents the first differential production measurements of deuterons at energies relevant for cosmic-ray studies, produced in inelastic p+p interactions at incident projectile momentum of 158 GeV/c ($\sqrt{s}$ = 17.3 GeV). The double-differential spectra are presented as functions of rapidity and transverse momentum and are compared to predictions of the thermal and coalescence models. These measurements are essential for improving our understanding of cosmic (anti)nuclei production, as detecting cosmic antinuclei can be a breakthrough approach to identifying dark matter. The primary source of cosmic antinuclei background is interactions between cosmic-ray protons and interstellar hydrogen gas. Gaining a deeper insight into the deuteron production mechanism in p+p interactions is an essential first step in modeling cosmic antinuclei production.

We present the FERMIACC, a scaffolded reasoning model built on OpenAI agents designed to autonomously generate and quantitatively validate theory hypotheses for high energy physics data at scale.

In generalizations of the Standard Models with extended scalar sectors, pseudoscalar particles will contribute to associated production of the Higgs boson discovered at the LHC, $h$, and the $Z$ boson. The pseudoscalar can be produced via gluon-gluon fusion, and as such may give contributions to $Zh$ production comparable to the value predicted by the Standard Model. We analyze this possibility in two different models, the Two Higgs Doublet Model, with and without an added complex singlet scalar, computing both the total and differential cross sections for this process. We focus on two pseudoscalar mass regions: (i) $m_A \sim 95$ GeV, to describe the di-photon excesses observed by CMS and ATLAS; (ii) 100 GeV $ \le m_A \le $ 1000 GeV, as a generic heavier pseudoscalar. A pseudoscalar with $m_A \sim 95$ GeV tends to give a contribution to the $pp \to Zh$ cross section that is too small to set new limits on the parameter spaces of the two models. Heavier pseudoscalars, on the other hand, can yield cross-section contributions larger than allowed by current LHC measurements and thus restricting regions of parameter space of these models which are in agreement with theoretical and all other experimental constraints. The increase of LHC luminosity will yield even substantially tighter constraints.

In this paper, we consider three compact objects (HESS J1731-347, PSR J1231-1411, XTE J1814-338) with anomalous mass-radius relation to analyze the possibility of being dark matter admixed neutron stars. We try to infer the dark matter particle properties, under the assumption of behaving as a free Fermi gas. The main novelty relies on the use of a baryonic equation of state obtained from first principles in the whole density range, that allows to eliminate the model dependence of the baryonic part of the calculation. Once the possible Dark Matter Admixed Neutron Star configurations are obtained, we check their stability and whether it is feasible for a Neutron Star to capture the necessary dark matter fraction. We show that two of the anomalous compact objects (HESS J1731-347 and PSR J1231-1411) can be explained with a small fraction of fermionic dark matter content in the star. The other compact object (XTE J1814-338) cannot be explained as a dark matter admixed neutron star, and becomes a potential candidate for a twin star.

Differential cross-section measurements are presented for the production of a $W$ boson in association with a photon. The analysis is performed using proton--proton collision data collected by the ATLAS experiment at $\sqrt{s}= 13$ TeV, corresponding to an integrated luminosity of 140~fb$^{-1}$. The differential cross sections are measured in the $Wγ\rightarrow \ell νγ$ decay channel ($\ell=e,μ$) as a function of 16 observables. Collectively, these observables probe the kinematic properties of the $Wγ$ system, the radiation amplitude zero effect predicted for the $Wγ$ final state, the polarisation of the $W$ boson, the charge conjugation and parity structure of the $WWγ$ triple gauge coupling, and the parton distribution functions of the proton. The data are corrected for the effects of detector inefficiency and resolution and are sufficiently precise that they can be used to distinguish between different state-of-the-art theoretical predictions provided by SHERPA, MADGRAPH5_aMC@NLO, and GENEVA. The differential cross sections are used to search for anomalous weak-boson self-interactions induced by dimension-six operators within an effective field theory. For CP-odd operators, dedicated detector-corrected observables based on the outputs of neural networks are found to be particularly sensitive to the interference between the Standard Model and dimension-six scattering amplitudes. Constraints are placed on the Wilson coefficients of the $\mathcal {o}_{W}$, $\mathcal{o}_{HWB}$, $\mathcal{o}_{\tilde{W}}$ and $\mathcal{o}_{H{\tilde{W}B}}$ operators in the effective field theory. The sensitivity to the $\mathcal{o}_{H{\tilde{W}B}}$ operator is improved by a factor of 2.5 compared to previous measurements in other final states.

We employ {PhysMaster}, an autonomous agentic AI system integrating theoretical reasoning, numerical computation, and exploitation strategies towards ultra-long horizon automation, to tackle long-standing challenges in non-perturbative lattice analyzes, including low signal-to-noise ratio at large transverse separation, complex systematic uncertainties, and labor-intensive manual workflows. Using the extraction of the CS kernel from quasi-transverse-momentum-dependent wave functions (quasi-TMDWFs) via large-momentum effective theory (LaMET) as a showcase, we demonstrate that \textsc{PhysMaster} automates high-dimensional fitting, renormalization, continuum-chiral extrapolation, and non-perturbative reconstruction in a fully autonomous manner. This framework drastically reduces the duration of the workflow from months to hours without compromising precision, stabilizes signals in the large-$b_\perp$ region to $1~\rm fm$, and produces results consistent with perturbative QCD and state-of-the-art traditional lattice calculations. This work validates the effectiveness of physicist-AI collaboration for first-principles QCD research and establishes a generalizable, reproducible paradigm for automated studies of parton structure and other non-perturbative observables from lattice QCD.

The consistent construction of quantum field theories beyond the simplest cases requires a precise characterization of the propagating degrees of freedom. These are encoded in the single-pole structure of two-point functions, a connection firmly established through foundational theoretical work of the last century. While high-level programs for spectral analysis are publicly available, we felt the need to complement the existing landscape with a tool that reaches the gauge-invariant propagator by the shortest possible algorithmic path. This is the purpose of \texttt{Kummitus}, an open-source Wolfram Mathematica toolbox designed to do precisely that: compute the (gauge-invariant) propagator. Beyond its utility in research applications, \texttt{Kummitus} is intended as an accessible and transparent resource for the theoretical community, with particular value for pedagogical purposes.

This paper explores 3d $\mathcal{N}=4$ quiver gauge theories whose moduli spaces represent nilpotent orbits, Słodowy slices or, more generally, Słodowy intersections, which span the Special Pieces of nilcones of Classical or Exceptional algebras. We introduce a map between magnetic and electric quivers containing symmetric group actions, such as wreathings (or loops), bouquets, and/or non-simply laced foldings, which can be related to symmetric subgroups of Lusztig's canonical quotient groups for Special Pieces. The map on quivers induces a map on nilpotent orbits that partially resolves the obstruction to quiver dualities presented by the non-involutive nature of the Lusztig Spaltenstein and Barbasch Vogan maps. We use Coulomb and Higgs branch quiver methods complemented by localisation formulae. Some new quivers for intersections within Exceptional nilcones are presented.

We present the calculation of three-loop massless QCD and QED helicity amplitudes for light-by-light scattering. We make use of Lorentz tensor decomposition in the 't Hooft-Veltman dimensional regularisation scheme to reduce the complexity of the computation. Our analytic amplitude results are remarkably compact and can be efficiently evaluated numerically. We employ them to compute the corresponding NNLO differential cross-section predictions in the invariant mass and rapidity distributions of the di-photon system, for which we find agreement with the experimental ATLAS data from ultra-peripheral heavy-ion collisions.

We examine the impact on the phenomenology of the minimal supersymmetric SU(5) Grand Unified Theory (GUT) of dimension-5 operators with coefficients suppressed by the Planck mass scale, with particular emphasis on predictions for nucleon decay. We incorporate dimension-5 operators in both the Higgs sector and the Yukawa interactions in the theory, and take account of the constraints from gauge coupling measurements, the mass of the Higgs boson, fermion masses and the cold dark matter density. We consider two scenarios for soft supersymmetry breaking: the constrained minimal supersymmetric extension of the Standard Model (CMSSM) and the Non-Universal Higgs Model (NUHM). We present predictions for the nucleon decay modes $p \to π^0 e^+, π^0 μ^+, K^+ \bar ν, π^+ \bar ν$, $K^0 e^+, K^0 μ^+$ and $n\to π^0 \bar ν$, $π^- e^+, K^0 \bar ν$, which we compare with both the present experimental sensitivities and those projected for the JUNO and Hyper-Kamiokande experiments. We find that these experiments may have interesting possibilities for discovering several of these decay modes.

Primordial non-Gaussianity is one of the most powerful probes of the inflationary epoch. The particle spectrum relevant to inflation, including masses and spins, is encoded in the precise form of statistical correlations of the adiabatic modes. Yet, in the presence of nonlinear structure formation, the optimal approach to measuring these signals remains unclear. Accurate modeling becomes crucial as late-time non-Gaussianty can become degenerate with primordial physics. Moreover, scale-dependent bias shows that information can move from non-Gaussian initial conditions to the amplitude of the Gaussian fluctuations. In this paper, we aim to clarify how primordial information is encoded in maps of galaxies. We use the field-level Cramer-Rao bound to investigate the ultimate limit of what can be extracted from realistic maps of the Universe. For local non-Gaussianity, we show that multi-tracer scale-dependent bias can exceed the sensitivity of conservative higher-point analyses. However, as expected, the multi-tracer analysis falls short of the optimal constraint when all the modes at the scale of the dark matter halos are included. We then forecast the potential reach of future surveys for equilateral and local non-Gaussianity. Equilateral in particular is highly sensitive to priors and modeling assumptions and can benefit dramatically from theoretical input such as the redshift evolution of the bias.

In this study we present a comprehensive sensitivity atlas for Big Bang Nucleosynthesis (BBN) in which we quantify the dependence of the primordial abundances of helium-4, deuterium, and lithium-7 as well as $N_{\rm{eff}}$ on variations in 14 fundamental particle physics and cosmological parameters and 63 thermonuclear reaction rates. We use the publicly available BBN code \faGithub \href{https://github.com/vallima/PRyMordial}{\,\texttt{PRyMordial}} to compute each sensitivity using two nuclear reaction rate compilations and two weak-rate normalization schemes, and provide a model independent reference applicable to Beyond the Standard Model (BSM) models in which MeV scale physics is modified. In addition, we rank each parameter's contribution to the theoretical uncertainty budget. We compare our predictions against the latest observational determinations of the primordial abundances, including a recent LBT measurement of the helium-4 abundance \cite{Aver:2026dxv} which roughly halves the observational uncertainty relative to previous determinations. We present these results both fixing $ΔN_{\rm eff}$ at its Standard Model (SM) value, and allowing it to be a free parameter using the latest uncertainty from the combined CMB+BAO+BBN 2026 value \cite{Goldstein:2026iuu}. When $ΔN_{\rm eff}$ is allowed to be a free parameter, it dominates the theoretical uncertainty of the helium-4 abundance, highlighting the importance of upcoming observations from the Simons Observatory \cite{SimonsObservatory:2025wwn}. As illustrative applications, we examine the deuterium tension and the lithium problem in light of our sensitivity analysis. The full set of numerical results and figures is publicly available on GitHub \faGithub \href{https://github.com/Anne-KatherineBurns/bbn-sensitivity-atlas}{\,\texttt{bbn-sensitivity-atlas}

We report new constraints on the spin-dependent scattering cross section between low-mass dark matter and protons using data collected by the COSINE-100 experiment. By implementing a specialized event selection process using a multi-layer perceptron and robust noise mitigation, this analysis pioneers a detection threshold of 3 and 4 isolated peaks, corresponding to the reconstructed photoelectrons, which is significantly lower than the 8 photoelectron threshold used in previous analyses. In this unstudied few-photoelectron regime, where PMT-induced noise and phosphorescence are prevalent, we utilize a phenomenological background model to search for the annual modulation signal expected from the Standard Halo Model. No statistically significant annual modulation is observed in our data. We derive new 90% confidence level upper limits for the spin-dependent DM-proton cross section, establishing the world's most stringent constraints in the 1.75-2.25 GeV/c$^2$ mass range. Furthermore, by incorporating the Migdal effect, we extend the experimental sensitivity to the sub-GeV/c$^2$ regime, setting world-leading limits in the 15-58 MeV/c$^2$ range. These results demonstrate the capability of NaI(Tl) target materials to probe previously unexplored regions of the dark matter parameter space.

We combine fast amplitude surrogates with neural importance sampling to accelerate NLO calculations. For virtual corrections, a learned ratio to the Born matrix element with calibrated uncertainties guarantees reliable precision across phase space. For real emission, we stick to the standard FKS subtraction and train sector-conditioned surrogates of the regularized integrands away from divergences. MadNIS then uses multi-channel mappings and FKS sectors as conditions. We validate our approach for electron-positron scattering to three and four jets and find significant speed-ups and variance reduction in the integration.

We study slow-down effects for bubbles formed in a cosmological first-order phase transition (PT) focusing on deflagrations and hybrids, where the bubble wall is preceded by a shockwave of heated plasma. Slow-down has been observed in multi-bubble simulations together with a suppression of gravitational wave (GW) emission, mostly for slow walls. We study the impact of the shock waves on the wall velocity around percolation, by considering steady-state single-bubble solutions and incorporating the possible heating effects by two different mechanisms. First, we investigate the slow-down experienced by a bubble expanding into an impeding shockwave, where the temperature is higher than at nucleation, and the fluid is no longer at rest. Taking into account such heating and kinematic effects, we find that the most significant slow-down occurs for the fastest walls, and thus cannot explain the suppression of the GWs observed in the simulations. However, these effects are stronger for PTs with a sizeable change in degrees of freedom unlike what is usually implemented in simulations, suggesting that the degrees of freedom can be an important additional parameter for characterizing the GW spectrum. For the second slow-down mechanism, we study heated droplets of false vacuum that shrink towards the end of the PT. By implementing a suitable boundary condition motivated by energy conservation, we show how the droplet velocity, interpreted here as the late-time velocity of the bubble walls, can be predicted from the properties of the initial deflagration/hybrid, in remarkable agreement with numerical simulations. Droplets are found to shrink more slowly for stronger PTs and slower deflagrations, with mild dependence on the change of degrees of freedom. Such slow droplets naturally correlate with a suppression of GWs, while geometrical properties such as the shock width play an important role as well.

We formulate a new class of primordial perturbations called $\textit{kinetic isocurvature perturbations}$, where the mass density of dark matter is constant relative to the photon number density while the kinetic energy of dark matter fluctuates in space. Such perturbations naturally arise in scenarios where a nonrelativistic heavy field decays into relativistic dark matter particles with a spatially modulated rate. As dark matter cools and becomes nonrelativistic, these fluctuations in kinetic energy leave large-scale density perturbations essentially unaffected and therefore evade the Cosmic Microwave Background bounds on isocurvature perturbations, yet survive as spatial variations in the free-streaming scale, resulting in patch-by-patch variation of the matter power spectrum.

The vast majority of extensions of the Standard Model affecting the number of effective relativistic neutrino species ($N_{\rm eff}$) do so additively, namely, they enhance this quantity with some light state contributing to dark radiation. In this work, we consider precisely the opposite case: new physics scenarios that can lead to $N_{\rm eff} < 3$ that are consistent with all known cosmological, astrophysical, and laboratory data. We are motivated by three main reasons: 1) a recent measurement from ACT and SPT in combination with Planck that leads to $N_{\rm eff} = 2.81\pm0.12$, 2) by a new and powerful measurement of the primordial helium abundance, which anchors $N_{\rm eff}$ to be very close to the Standard Model value one second after the Big Bang, 3) by the deployment of the Simons Observatory which will provide precise tests of the radiation content in the Universe and which may detect with a high significance cosmologies with $N_{\rm eff}<3$. We survey the main theoretical possibilities and find that only a few simple scenarios can consistently give $N_{\rm eff}=2.81\pm0.12$. One class consists of thermal electrophilic relics with masses $m\sim 8\!-\!13\,{\rm MeV}$. Another consists of out-of-equilibrium particles decaying to $e^+e^-$ or $γγ$, with a rather particular lifetime $0.05\,{\rm s}\lesssim τ\lesssim 3\,{\rm min}$, mass $250\,{\rm MeV}\lesssim m \lesssim 600\,{\rm MeV}$, and abundance $ρ/ρ_γ\sim 0.1$ at decay. Thermal electrophilic particles are especially interesting because they can account for the dark matter in the Universe and can be tested in experiments such as SENSEI, DAMIC-M, and Oscura. We conclude that if the Simons Observatory confirms that $N_{\rm eff} \simeq 2.8$, it will point to very specific extensions of the Standard Model.

We present a concept for a tabletop-scale detector with an amorphous target designed to search for dark matter absorption into phonon excitations. In crystalline materials, absorption occurs only at narrow resonances where the dark matter mass matches a zero momentum optical phonon mode, whereas amorphous targets provide a broadband response that can substantially enhance the absorption rate away from these resonances. The predicted backgrounds arise from the relaxation of disorder-induced metastable defects in the amorphous target, as well as from low-energy noise intrinsic to superconducting phonon sensors. A prototype detector with a target mass of only a few $μ$g could provide broadband sensitivity to dark photon absorption across the 50 meV-200 meV mass range, probing up to two orders of magnitude beyond existing constraints.

Understanding the link between correlation functions (CFs) of local operators and measurable collider correlators has emerged as a new opportunity in the study of gauge theory dynamics at colliders. While in Conformal Field Theories (CFTs) this connection is established by the light transform, the non-conformal nature of QCD complicates its use beyond the lowest perturbative order. We show that a continuation of the CFs to the Wilson-Fisher fixed point can be used as a method to overcome these obstacles, serving as a conformal bridge for the evaluation of the light transform. At the fixed point, the renormalized CF of four local operators features a variable drop and only depends on two conformal cross ratios, in line with a genuine CFT quantity. This allows us to exploit CFT techniques to perform, for the first time, its light transform at higher loop orders. Remarkably, the resulting collider correlator in four dimensions can be recovered from this result simply by using lower-loop data. We demonstrate this method by computing the back-to-back limit of the charge-charge correlation (QQC) at two loops in QCD through the light transform of the CF of four vector currents in the sequential light-cone limit, reproducing a recent prediction.

A search for Higgs boson (\PH) production at high transverse momentum ($p_\mathrm{T}$) in the WW decay channel is presented. The analysis uses proton-proton collisions at $\sqrt{s}$ = 13 TeV recorded by the CMS experiment in 2016$-$2018, corresponding to an integrated luminosity of 138 fb$^{-1}$. The visible decay products of the Higgs boson are reconstructed as a single large-radius jet with one isolated lepton or none (1$\ell$ and 0$\ell$, respectively; $\ell$ = e, $μ$). The \PH-candidate jets are identified using an advanced transformer-based algorithm and are calibrated with the Lund jet plane reweighting technique. The 1$\ell$ channel is further split into gluon fusion, vector boson fusion, and associated production with hadronically decaying vector boson categories, while the 0$\ell$ channel considers all production processes inclusively. The measured cross section times the H $\to$ WW branching fraction relative to the standard model expectation is $μ$ = $-$0.19$^{+0.48}_{-0.46}$, indicating no evidence of a signal above the background. This measurement represents the first dedicated study of highly Lorentz-boosted H $\to$ WW decays, complementing earlier searches for high-$p_\mathrm{T}$ Higgs boson in other decay channels.

We show that the leading non-analytic terms in the small-t expansion of the energy momentum tensor (EMT) form factors of an electrically charged particle in QED can be correctly derived in a classical model of the electron by Bialynicki-Birula. Based on the lucidity of the employed exactly solvable model, we comment also on the recently proposed concept of a regularized proton D-term.

Evidence from lattice and continuum studies supports the existence of a fully nonperturbative Higgs mechanism generating mass for gluons in the gauge sector of the strong interactions in linear covariant gauges. The broken charge is the Kugo-Ojima charge. The corresponding unphysical Goldstone boson is a bound-state superposition of two gluons, three gluons and a ghost-antighost pair. Mass generation occurs via the Schwinger mechanism, triggered by the formation of the Goldstone boson. Once corrected for symmetry breaking, the color charge operator is unbroken and confining.

We consider Abelian extensions of global symmetries of the form $A \to G \to K$, with $A$ finite (and similar higher-group structures). For a quantum field theory $\mathcal{T}$ with symmetry $G$, we compare gauging $G$ directly with gauging first $A$ and then $K$, and show that for finite Abelian groups and for $K \simeq U(1)$ the two procedures are equivalent as expected, $\mathcal{T}/G \simeq \mathcal{T}/A/K$. In the continuous case $K=U(1)$, after gauging the full extension, the dual symmetry $\widehat{\mathbb{Z}}_q^{(d-2)}$ fits into an extension characterizing the topological data of the magnetic $U(1)_m^{(d-3)}$ symmetry. This is better described using differential cohomology. We also briefly comment on the relation to symmetry fractionalization.

Strongly coupled theories are of phenomenological interest, for example as dark matter candidates. Theories that can undergo first order thermal phase transitions are particularly appealing as potential sources of a stochastic gravitational wave background. Determining the expected gravitational wave signal from a first order phase transition requires accurate information on the bubble nucleation rate, but thus far for strongly coupled models these have relied on semiclassical methods. As a first step towards determining the nucleation rate, in this paper we study the confinement-deconfinement phase transition in a 4D SU(8) pure gauge model, using multicanonical Monte Carlo. Resolving the critical bubble for the first time in a pure Yang-Mills model, we determine the critical bubble probability and compare it to results from thin wall calculations. We also compare the effectiveness of different lattice pseudo-order parameters at resolving the condensation transition between the metastable phase and critical bubble branch, and point out the choice of order parameter is crucial to accurately resolve the critical configurations.

The weak decays of charmonium, involving $J/ψ$ and $ψ(2S)$ states, are instrumental in probing both non-perturbative QCD dynamics and flavor structure of the standard model~(SM). The extremely rare nature of charmonium weak decays makes them highly sensitive to new physics beyond the SM, particularly in channels heavily suppressed in the SM, such as flavor-changing neutral current~(FCNC) decays. This review highlights the critical role of the BESIII experiment, which leverages an unprecedented dataset of over $10^{10}$ $J/ψ$ events and $2.7\times10^{9}$ $ψ(2S)$ events to push the sensitivity of charmonium weak decay searches. We present the latest and most stringent upper limits established by BESIII on various semi-leptonic, non-leptonic, and FCNC charmonium weak decay channels.

Understanding the infrared sensitivity of perturbative predictions in QCD is important for assessing the magnitude of possible non-perturbative power corrections to processes with large momentum transfer. In renormalon models, this sensitivity can be related to computable dependences of perturbative quantities on a small gluon mass. However, this procedure cannot be applied to collider processes with gluons at the Born level. To address this problem, we promote the gluon mass to a parameter of a consistent non-Abelian quantum field theory where the gauge symmetry is spontaneously broken through the Higgs mechanism. Working in the limit in which the gluon mass $m_\mathrm{g}$ is the smallest dimensionful parameter, we compute through two loops the $\mathcal{O}(m_\mathrm{g})$ contributions to the relation between the pole and $\overline{\rm MS}$ masses of a heavy quark and to the relation between corresponding field counterterms. We expect that the proposed framework will provide a useful laboratory for probing linear infrared sensitivity of collider observables in QCD.

We prove that all geometric helices in the derived category of coherent sheaves on a del Pezzo surface are related by a sequence of elementary operations: rotation, shifting, orthogonal reordering, derived dualization, tensoring by a line bundle, and tilting. As a consequence, any two non-commutative crepant resolutions of the affine cone over a del Pezzo surface are related by mutations. The proof relies on a geometric interpretation of tilting operations as cluster transformations acting on toric models of a log Calabi--Yau surface mirror to the del Pezzo surface.

We propose a novel model in which the Universe is created as a pair of coordinate-reversed counterparts, forming a globally CPT-symmetric system that permits local CPT violations within each sector. This framework naturally introduces a mirror universe with opposite chiralities and reversed microscopic time coordinates, providing a geometric interpretation of time reversal without relying on initial-final state interchange. We investigate the consequences of local CPT violation in each universe, which induces a mass difference between the real inflaton and anti-inflaton fields. Such an asymmetry can modify reheating temperatures and naturally generate the observed matter-antimatter asymmetry in both universes.

We study QCD with three degenerate flavors of dynamical quarks using first-principles lattice simulations. For a specific choice of imaginary isospin chemical potential, this theory possesses an exact center symmetry, just like pure gauge theory. This exact symmetry is expected to be intact at low temperatures and spontaneously broken in the high-temperature regime. By analyzing the finite-size scaling of the Polyakov loop distribution, obtained with a dedicated multi-histogram approach, we demonstrate that there is a first-order deconfinement phase transition in between. Our results are obtained employing stout-smeared rooted staggered quarks at one lattice spacing. Using simulations at different quark masses we sketch the behavior of QCD in the mass-isospin chemical potential plane, shedding new light on this corner of the fundamental phase diagram of the strong interactions and the relationship between chiral symmetry breaking and deconfinement.

In this paper we investigate the existence of internal modes of vortices in the gauged $\mathbb{C}P^1$ sigma model. We develop a clean geometric formalism that highlights the symmetries of the Jacobi operator, obtained from the second variation of the energy functional. The formalism and subsequent results fundamentally rely on the Bogomol'nyi decomposition of the energy functional, and can therefore be extended to other models with such a decomposition. We prove the existence of at least one shape mode for a general $\mathbb{C}P^1$ vortex solution on $\mathbb{R}^2$, and find numerically the shape modes and corresponding frequencies of a radially symmetric vortex. A surprising result is that the shape mode eigenvalues are very close to the scattering threshold, suggesting weakly bound shape modes could be characteristic of the $\mathbb{C}P^1$ model.

We investigate whether the type-I seesaw mechanism can rescue the $Z_3$ Froggatt--Nielsen framework for neutrinos and find that it cannot. With right-handed Majorana masses carrying the $Z_3$ charge structure dictated by the Majorana bilinear -- where suppression powers follow $(q_i+q_j)\bmod 3$ -- the mass matrix contains an unsuppressed off-diagonal entry whose dominance in $M_R^{-1}$, combined with the hierarchical column texture of $M_D$, over-suppresses the two lightest neutrino masses to $\mathcal{O}(\varepsilon^3)$ while $m_3$ remains $\mathcal{O}(1)$. This pushes the solar-to-atmospheric mass ratio to a median $Δm^2_{21}/Δm^2_{31}\sim 4\times 10^{-11}$ -- eight orders of magnitude below the observed value of $0.030$. We prove this failure is universal across all six permutations of the charges $(2,1,0)$ and show analytically that the generic ratio scales as $Δm^2_{21}/Δm^2_{31}\sim\mathcal{O}(\varepsilon^6) \sim 10^{-11}$, with fewer than $0.01\%$ of parameter-space points exceeding $\varepsilon^2\approx 2\times 10^{-4}$. The PMNS angles remain Haar-random, carrying no information from the expansion parameter. We then show that $A_4$, the alternating group of order 12, is the minimal discrete symmetry resolving both failures. Its triplet representation provides two independent vacuum parameters controlling the solar and atmospheric mass scales separately, while constraining the PMNS matrix to the trimaximal TM$_1$ pattern. The TM$_1$ solar sum rule predicts $\sin^2θ_{12}=0.318$ ($1.2σ$ from NuFit 6.0, $1.0σ$ from JUNO), and the atmospheric sum rule yields a parameter-free $(\sin^2θ_{23},\,\cosδ)$ correlation predicting $δ\approx -71^\circ$, testable at DUNE and T2HK.

Numerous microlensing survey programs have constrained the possibility of dark matter existing in the form of compact objects within the Galactic halo. These constraints on the dark matter fraction were derived under the assumption of isolated, widely separated objects. This work investigates microlensing by primordial black holes (PBHs) organized into clusters. In this scenario, it is necessary to account for both the influence of neighboring PBHs and the collective gravitational potential of the entire cluster, which significantly complicates the microlensing light curve. Events exhibiting such complex light curves elude detection in observational experiments such as MACHO, EROS, OGLE, POINT-AGAPE, and HSC. It is demonstrated that a significant fraction of PBH dark matter (up to 93\% for the models studied) remains undetected in these observational data. However, for all considered cluster models, a substantial population of PBHs still behaves as isolated lenses. Consequently, the clustering of PBHs does not completely eliminate the microlensing constraints on the PBH contribution to dark matter.

We present a high-energy spectral analysis of GRB 250129A, which was triggered by the Swift-BAT. The burst exhibits a complex, multi-peaked temporal structure characterized by two distinct emission episodes, with the main peak occurring approximately 180 seconds after the BAT trigger. The time-integrated spectral analysis in the 15 - 150 keV energy range indicates that a broken power-law (BPL) model provides the best fit, signifying a non thermal origin of the prompt emission. A time resolved spectral analysis, performed using the Bayesian block technique, shows that the intervals around the main emission peak are well described by the BPL model, while the fits for low count intervals remain less constrained. An evident intensity tracking behavior is observed between the flux and the spectral peak energy (Ep). Furthermore, both the Amati relation and hardness - intensity correlation suggest that GRB~250129A occupies an intermediate regime, acting as a bridge between long and ultra long GRBs.

The cosmic distance duality relation (CDDR), expressed as $d_L(z) = (1+z)^2 D_A(z)$, is a fundamental relation in modern cosmology. In this work, we apply a method to test the CDDR using simulated strongly lensed gravitational-wave (SLGW) signals from massive binary black holes (MBBH) as observed by proposed space-based detector networks. Our analysis is conducted under the point-mass lens model, considering the strong lensing scenario that produces two images. We generate 90 days of simulated SLGW data for 10 events based on the Population III stellar formation model, with source redshifts in the range $z_s \in [2,6]$ and lens redshifts in $z_L \in [0.2,1]$. The deviation of CDDR is parameterized by $η_1(z) = 1 + η_0 z$ and $η_2(z) = 1 + η_0 z/(1+z)$, and we incorporate the deviation parameter $η_0$ directly into the waveform model. Parameter estimation is performed within a Bayesian statistical framework, combining simulated data from both Taiji and LISA. For a single lensed event, the joint Taiji+LISA analysis improves the measurement precision of $η_0$ by roughly a factor of two compared with Taiji-only observations. By combining 10 simulated events, the population-level constraints on $η_0$, quantified by the half width of the $95\%$ credible interval, reach approximately $2.61\times10^{-4}$ ($1.72\times10^{-4}$) for the $η_1(z)$ parameterization and $1.22\times10^{-3}$ ($6.86\times10^{-4}$) for $η_2(z)$ in the Taiji-only (Taiji+LISA) scenario, respectively. The inferred values of $η_0$ remain consistent with $η_0 = 0$ within the estimated uncertainties, with no statistically significant evidence for deviations from the CDDR at the achieved precision. These results demonstrate the significant advantage of joint space-based observations for high-precision tests of the CDDR.

Gamma-Ray Burst (GRB) afterglows arise from the interaction of relativistic ejecta with the circumburst medium and are observed across the electromagnetic spectrum. Afterglow polarisation is expected at early and late phases depending on the presence of reverse shocks (RS) and the observer's viewing geometry relative to the jet. Polarimetric observations of GRB afterglows provide a unique diagnostic tool to probe the geometry and structure of magnetic fields in the emitting region, which cannot be inferred from photometric or spectroscopic data alone. We report late-time (~19 hours post-burst) spectropolarimetric observations of GRB 250129A using the Southern African Large Telescope (SALT). The data reveal a hint of linear polarisation, with no evidence for rotation in the polarisation angle across wavelengths. Polarisation is typically expected during the early afterglow (<100 s) when the RS dominates. However, multi-wavelength modelling shows no indication of RS contribution at late times. Modelling incorporating both forward shock (FS) and RS components confirms that the RS fades rapidly after ~100 s. The afterglow emission is best explained by an off-axis viewing geometry of a jet with a Gaussian core and wings evolving in a uniform density environment. GRB 250129A thus provides rare observational evidence linking late-time polarisation to jet geometry and structure.

The 3D structure of The Milky Way's Central Molecular Zone (CMZ) informs our understanding of star formation cycles, black hole accretion, and the evolution of galactic nuclei. However, a comprehensive 3D model has remained elusive, as no singular dataset nor theory contains the requisite information to describe the orbital motion of the gas. We implement a Bayesian framework to flexibly combine datasets across the electromagnetic spectrum for molecular clouds in our CMZ catalog. We develop near/far metrics for each dataset, including dust extinction, absorption, stellar densities, X-ray echoes, and proper motions; and report a posterior positional probability density function (PPDF) for each cloud. We then use the posterior PPDF distributions for all CMZ clouds to search for a best fitting x$_2$ orbit. We find that no single orbit is a perfect fit, but the structure can overall be represented by nested x$_2$ orbits, with major axes ranging from about $72 < a < 146$ pc. We also present projected line of sight distance estimates for all 31 clouds in the catalog. Our results highlight asymmetries along the line of sight, with most clouds lying on the near side of the Galactic Center, and agree overall with current near/far assumptions for most CMZ clouds, including those in the Sgr A region, which may be much closer to the center. We conclude that the CMZ can be well-described by x$_2$ orbital families, and that the overall gas distribution is more complex than a single closed or open elliptical orbit.

Class II protoplanetary discs feature numerous non-axisymmetric substructures like spirals and the underlying mechanisms for their formation are still highly debated. Coincidentally, early stage, massive discs are subject to the gravitational instability that causes them to collapse into denser substructures. However, like for most instabilities, real systems usually remain marginally stable, here with Toomre parameter $Q \gtrsim 1$. We study how the self-gravity of the gas triggers the growth of spiral structures in the disc. We specifically focus on discs that are considered stable, that is, with respect to the gravitational instability (with $Q > 1$), as these discs remain unstable to non-axisymmetric perturbations like spirals. After a linear stability analysis, we produce high-resolution 2D shearing sheet simulations with the GPU-accelerated code \idefix of self-gravitating discs. We probe different initial densities and thermodynamical models of Toomre-stable discs. The initial transient growth of the spiral wave matches the linear theory provided we take into account the time dependency of the amplification. The spirals are then rapidly non-linearly amplified with growth rate $\approx 10$ orbital time scale. After this time spiral large scale mode are amplified up to 1000 times more than linear theory predicts. At later times, low density discs reach a weak gravito-turbulent state with $α\approx 10^{-3}$ and discs with higher density undergo runaway collapse of the spiral arms. All discs exhibit dominant large-scale spirals.

Cosmographic reconstructions provide a model-agnostic approach towards constraining cosmic evolution. In this work, we develop a semi-cosmographic framework that adopts a Padé-rational fraction parametrization of the Luminosity distance, but also invokes a phenomenology-motivated two-body decaying dark matter (DDM) sector. In this approach, we do not assume any model for the dark energy. However, we consider the dark matter sector to comprise a non-relativistic parent particle that decays into a massless and a massive daughter. Assuming a cosmographic expansion history and the DDM background evolution, a semi-cosmographic dark energy equation of state is inferred. The various cosmological observables, hence computed, are fitted to the data. We use DESI DR2 BAO data and with a forecasted 21\,-cm intensity-mapping power spectrum at $z\simeq 1.75$ with a SKA1-Mid-like instrument. Posterior constraints on the Padé and DDM parameters are obtained using Markov Chain Monte Carlo (MCMC) analysis. This allows us to reconstruct the equations of state of the massive daughter and dark energy.

3I/ATLAS is the third interstellar comet identified as passing through the Solar System. Its high outgassing activity and favourable perihelion passage on October 29, 2025 UT provided an excellent opportunity to investigate the composition of its coma gases through millimeter spectroscopy. We present observations undertaken with the IRAM 30-m telescope on November 1--3, 2025 at an heliocentric distance of 1.36--1.37 au. Lines of HCN, CH$_3$OH, CO, and H$_2$CO are well detected, and $\sim$4$σ$ detections are obtained for CS and CH$_3$CN. The search for H$_2$S was unsuccessful. Abundances of CO, H$_2$CO, CH$_3$OH, and CH$_3$CN relative to HCN are in the upper ranges of values measured in Solar System comets. The sulfur-to-carbon abundance ratio in 3I/ATLAS's coma is at most the minimum value observed in comets. The unusually low expansion velocity of coma gases suggests a near-nucleus gas flow driven by heavy molecules such as CO$_2$, and/or a large fraction of the gaseous production coming from subliming icy grains.

Observing the first generation of Population~III (Pop~III) stars is one of the most demanding challenges in astronomy. Indeed, Pop~III stars are expected to predominantly form within faint minihalos at early times with a top-heavy initial mass function, resulting in efficient metal enrichment and a fast transition to Pop II-dominated systems. However, recent surveys with the James Webb Space Telescope (JWST) have identified galaxies at the end of the Epoch of Reionization (EoR) with possible signatures of significant Pop~III star formation even at these later times. We here explore the physical conditions required to produce massive Pop~III starbursts during the EoR, using cosmological radiation-hydrodynamic zoom-in simulations. We specifically focus on galaxies with a virial (dynamical) mass of $M_{\rm vir} \approx 10^{8} \msun $ at $7 \lesssim z \lesssim 8$, i.e., the atomic-cooling halos that could be potential sites for such maximal Pop~III starbursts. In particular, we vary the strength of Lyman-Werner (LW) background radiation up to $J_{\rm LW} \leq 10^4J_{21}$, further imposing a high star formation efficiency (up to $ε_{\rm ff} = 1.0$). Our results show that Pop~III starbursts, observable in strongly-lensed survey fields like GLIMPSE, can occur in the presence of a sufficiently high LW flux (with $\gtrsim 10^3J_{21}$), leading to delayed, but intense Pop~III star formation. However, even for such high LW fluxes, the Pop~III starburst mass is limited to $M_{\star, \rm Pop~III} <10^6\msun$, as strong internal metal enrichment occurs after the first Pop~III supernova explosions within the simulated galaxies. While the conditions favoring observable Pop~III starbursts are expected to be rare, we anticipate that future and ongoing large-volume surveys leveraging gravitational lensing, such as VENUS, will detect multiple cases of Pop~III starbursts in the EoR.

The initial masses of red supergiant (RSG) type II supernova (SN II) progenitors are commonly inferred from pre-explosion imaging by converting the progenitor luminosity into an initial mass estimate using non-rotating stellar evolution models. However, stellar rotation affects the evolution and may influence these estimates. We investigate how the observed distribution of rotational velocities in massive stars influences the progenitor initial masses of SNe II inferred from pre-SN imaging. We compare initial mass estimates obtained from non-rotating models with those derived from rotating models, where the initial rotational velocities of the stellar models are sampled from the observed distribution. We analyse the inferred progenitor initial masses by (i) comparing the results for each SN individually, (ii) examining the overall probability density function, (iii) constructing the cumulative distribution function, and (iv) determining the upper initial-mass boundary. In all cases, the distributions obtained from rotating models are slightly shifted towards lower masses, although the differences remain smaller than the typical uncertainties. When using the observed distribution of initial rotational velocities for massive stars, we infer an upper initial-mass limit for SN II progenitors of 20.4$^{+2.3}_{-1.9} M_{\odot}$. Taken together, these analyses demonstrate that stellar rotation has only a modest impact on progenitor mass estimates from pre-SN imaging within the current observational and model uncertainties when the observed distribution of initial rotational velocities is taken into account. Therefore, adopting this distribution leads to small differences compared to non-rotating models.

Properties of the hot intracluster and intragroup medium are mostly set by the underlying gravitational potential well, although complex astrophysical processes at play during their buildup may leave a significant imprint. Observational constraints on the degree and scales of such non-gravitational processes require well-selected samples of objects and deep observations of their gas content. We aim to study the scaling relation between two global properties of the hot gas, namely its soft-band X-ray luminosity ($L_X$) and its temperature ($T$), by studying a sample of low-mass systems associated with precise redshifts, simultaneously accounting for sample selection biases and associated measurement uncertainties. This work takes as input a large catalogue of X-ray-selected galaxy clusters (X-CLASS). We perform a thorough revision of the redshifts of sources using deep photometric data from the Legacy Surveys and our own tailored spectroscopic follow-up of 52 low-redshift systems. We devise a spectroscopically complete sample of 155 low-redshift ($0.07<z<0.2$) systems, and we measure properties of their X-ray emitting gas, with median $\overline{T}=1.7$ keV and median $\overline{L_X}=10^{43}$ erg s$^{-1}$. We infer the relation between $L_X$ and $T$ in a Bayesian framework. Our sample of groups and clusters with median total mass $\sim 6 \times 10^{13}M_\odot$ reveals a relation $L_X-T$ steeper than predicted by the self-similar model, with a slope $B=3.2 \pm 0.1$. This result fits well within recent studies that together indicate a trend of increasing slope with decreasing median halo mass. This work supports a scenario of a stronger decrease in luminosity with decreasing mass in the group regime than for massive galaxy clusters. This effect is possibly due to strong and sustained feedback expelling gas efficiently from their relatively shallower potential wells.

While many aspects of high-mass star formation have been investigated, the accretion onto the central protostars is one of the most fundamental but less explored physical properties. JWST/MIRI offers a unique opportunity to explore tracers of accretion at less-extincted wavelengths (5 to 27 um) than those studied so far. We probe the MIRI (MRS/IFU) capability to detect and resolve atomic Hydrogen (HI) emission lines in such embedded objects, to subsequently estimate accretion luminosities (Lacc) and accretion rates (Macc) for the first time in a sample of high-mass star forming regions at different evolutionary stages. We use dereddened HI line luminosities as tracers of accretion by applying existing line-to-accretion-luminosity relations (Lacc-calibrations). As they were originally established for low-mass Class II objects, we assess their applicability on our sample prior to estimating Macc. The infrared continuum reveals, at much higher spatial resolution than before, the location of new protostars, toward which we detect a handful of HI lines. While a few lines are secure detections, many are tentative. The most commonly detected line is HI 7-6, followed by HI 8-6 and HI 6-5. Assuming that their line fluxes are dominated by accretion, we find that two of the three existing Lacc-calibrations predict excessively high Lacc that largely exceed the corresponding L_bol, and that the third Lacc-calibration still overpredicts Lacc for some sources. Considering the given uncertainties, estimated accretion rates are only tentative. This work demonstrates the great potential of JWST/MIRI to probe HI line emission originated in the innermost regions of high-mass protostars, setting the ground floor for further investigations into accretion. While this project had the ambitious goal of robustly quantifying Macc, we have shed light on what outstanding methodological challenges remain.

Understanding whether prebiotic molecules can endure and reform through the energetic stages of star formation is essential for tracing the continuity of interstellar chemistry toward life. Glycolamide, an isomer of glycine, was recently detected in the molecular cloud G+0.693-0.027. However, establishing its presence in warm, high-density environments is crucial to evaluate the chemical continuity of amides. Here we report the first detection of glycolamide in a hot molecular core, G358.93-0.03 MM1, using ALMA 1 mm observations. Seven unblended or only mildly blended emission lines were identified, yielding an abundance of (1.7$\pm$0.2)$\times 10^{-10}$ relative to H$_{2}$. The comparable formamide/glycolamide and acetamide/glycolamide abundance ratios in both sources suggest a chemically connected amide network across different environments. These results demonstrate that amides can persist and chemically evolve during massive star formation, tracing the chemical continuity from interstellar to protostellar environments.

We investigate whether the intracluster light (ICL) can serve as a reliable tracer of the shape of the underlying dark matter (DM) haloes in galaxy clusters. Using the cosmological Hydrangea cluster simulations, we measure the 3D and projected shapes of both components with a shape tensor computed in concentric ellipsoidal shells, out to the virial radius $R_\mathrm{200c}$ for each cluster. The ICL and DM are closely aligned, with their major axes typically offset from each other by $\lesssim$10 degrees. Their axis ratios also match closely, with a typical difference of only $\approx\! 0.07$ for both the major-to-minor and major-to-intermediate axes, the DM being slightly rounder than the ICL. These trends are consistent across 2D and 3D measurements and agree well with results from isophotal fitting of mock images. In detail, the axis ratio offset is sensitive to the method used to remove satellites, and may also depend on the choice of subgrid physics models. We demonstrate that the ICL traces the DM shape better than the distribution of satellite galaxies, which exhibits larger scatter in the axis ratio and misalignment angle and is overall more elliptical. Together, these results indicate that the ICL can act as a useful proxy for DM halo ellipticity and orientation.

Fast radio bursts (FRBs) are millisecond-duration pulses originating from cosmological distances. Multi-wavelength counterparts associated with FRBs are important for unveiling their physical origins. Recent observations provide strong evidence that the sources of some active FRBs are residing in massive star binaries. In this paper, we study the electromagnetic counterparts of FRBs, including the persistent radio sources (PRSs) and the bow shock radiation from wind collisions for FRBs residing in magnetar - massive star binaries. We find that the PRSs with luminosity $10^{38}-10^{39}$ erg s$^{-1}$ can be generated by young magnetar wind nebulae (MWN). The age of magnetars is a few decades. The observed long-term variation of flux density for PRSs can be explained by the internal magnetic field decay of magnetars. The bow shock radiation can account for the less luminous PRS of FRB 20201124A. The multi-wavelength emission arising from synchrotron radiation and inverse-Compton scattering in the bow shock can be the electromagnetic counterpart of FRBs. The emission at keV, GeV and TeV bands from the binary model can be detected at the distances of $\sim10-100$ Mpc, $\sim 1-10$ Mpc and $\sim0.1$ Mpc by current instruments, respectively.

The time-delay model is the way we interpret the diagram [X/Fe] vs. [Fe/H], where X is the abundance of a generic element from carbon to uranium. This interpretation is based on the lifetimes of stars of different masses producing different elements. The abundance of Fe ([Fe/H]) traces the "stellar metallicity" and is due to supernovae Type Ia, which are believed to be the major producers of Fe, and in part to supernovae core-collapse. In particular, if X is an alpha-element, produced on short timescales from massive stars, the ratio [alpha/Fe] will show an overabundance of the alpha-elements relative to Fe at low metallicity. In fact, the bulk of Fe is produced with a time delay relative to alpha-elements, since Type Ia supernovae are white dwarfs in binary systems and they can have lifetimes as long as the age of the Universe. In this paper, I will show how powerful is the time-delay model in order to interpret the abundance patterns observed in stars and interstellar gas, since it allows us to put constraints on stellar nucleosynthesis as well as on the star formation histories of galaxies. I will present some applications of the time-delay model, in particular to the chemical evolution of the Milky Way and galaxies of different morphological type as well as to the identification of high redshift objects by means of their abundances.

Neutral sodium is an important tracer of the Galactic chemical evolution, a powerful diagnostic of different stellar populations, and the subject of detailed studies of exoplanet atmospheres via transmission spectroscopy. This work aims to study and quantify the errors in stellar analyses of Na I lines caused by the use of one-dimensional (1D) hydrostatic model atmospheres and the assumption of local thermodynamic equilibrium (LTE). We studied the line formation of nine Na I lines in FGK dwarfs and giants via, for the first time, 3D non-LTE (NLTE) radiative transfer post-processing with the code Balder on 3D radiation hydrodynamic stellar atmospheres from the Stagger grid spanning Teff= 4000 to 6500 K, log g = 1.5 to 5.0, and [Fe/H]=-4 to +0.5. We find that the 3D NLTE abundance corrections relative to 1D LTE tend to be negative, and more positive than the corresponding 1D NLTE corrections. This reflects more efficient overionisation in the steeper temperature gradient of the 3D models. The corrections are typically less severe than -0.1 dex for weak lines, but become much larger for saturated lines in low-gravity giants (log g < 2.0), even reaching -0.7 dex. However, for the D resonance lines, the 3D NLTE corrections relative to 1D LTE become slightly positive at the lowest metallicities in our grid, typically around +0.05 dex at [Fe/H]=-4. We make our 3D NLTE grid, together with interpolation routines based on radial basis functions and fully connected feedforward neural networks, publicly available. This will enable more accurate determination of sodium abundances in present and forthcoming stellar spectroscopic surveys, particularly for metal-poor stars, as well as a better characterisation of the Na I D lines in exoplanet atmospheres.

The Large High Altitude Air Shower Observatory (LHAASO) Collaboration has recently reported the first detection at TeV energies of a low-luminosity radio galaxy, NGC 4278. The aim of this work is to investigate the high-energy properties of NGC 4278 during the flaring and subsequent quasi-quiescent states with the Florian Goebel Major Atmospheric Gamma Imaging Cherenkov (MAGIC) telescopes. NGC 4278 is located in the field of view of two blazars, 1ES 1215+303 and 1ES 1218+304, previously observed by the MAGIC telescopes. Therefore, we re-analyzed MAGIC observations made between 2010 and 2024 on these sources. We also modeled the broadband spectral energy distribution of the source during and after the flaring state at TeV energies. We did not detect any statistically significant $γ$-ray emission from NGC 4278 with MAGIC. The corresponding upper limits obtained using the entire MAGIC dataset ($F_{{\rm UL, }\, >150\, \mathrm{GeV}}=1.5 \times 10^{-12}\, \mathrm{ph \, s^{-1}\, cm^{-2}}$) are consistent with the LHAASO results. The best-fit models obtained for both emission states suggest that the emitting region is strongly particle-dominated, and an efficient acceleration mechanism has to be in action in order to reach TeV energies. The transition between the flaring and quasi-quiescent state cannot be explained by a simple radiative cooling of the emitting particles. The inferred jet power, of the order of $L_{\rm jet}\sim 10^{42}\, \mathrm{erg\,s^{-1}}$, is dominated by the kinetic component in both states and it is in a good agreement with previous, time-averaged observational estimates, supporting the idea that such high-energy flares might be recurrent. The jet, however, remains too weak to break the host-galaxy confinement.

Spectral energy distribution (SED) models are widely used to infer the physical properties of galaxies from multi-wavelength photometry, but their accuracy is difficult to assess because the true properties of observed galaxies are generally unknown. We address this by fitting synthetic SEDs of ~31,000 star-forming galaxies drawn from the EAGLE cosmological simulations, post-processed with the SKIRT radiative transfer code, using the MAGPHYS SED modelling framework. This provides a controlled testbed with known intrinsic parameters, enabling a direct assessment of model accuracy and the origin of systematic biases. Under idealised conditions, fitting well-sampled ultraviolet-to-submillimetre SEDs at z=0.1, z=2, and z=5, MAGPHYS recovers stellar mass, star formation rate, specific star formation rate, dust mass, and dust luminosity to within <~0.14 dex, while mass-weighted stellar ages are not robustly constrained. We find that mismatches between the assumed star formation history (SFH) priors and the intrinsic SFHs of the simulated galaxies introduce systematic biases in stellar mass estimates, even when the fits provide good statistical agreement. To assess performance under realistic survey conditions, we construct a WAVES-like mock sample using optical and near-infrared photometry with realistic uncertainties. In this case, stellar masses and star formation rates remain well constrained (systematic offsets <~0.1 dex; scatters ~0.07 and ~0.15 dex, respectively), whereas dust properties degrade significantly without far-infrared data: dust luminosities show offsets of ~0.30 dex and scatters ~0.25 dex, and dust masses exhibit scatters ~0.3 dex. We conclude that MAGPHYS is a reliable tool for recovering key galaxy properties from broad-band photometry, but that SFH assumptions and limited wavelength coverage introduce significant uncertainties, particularly for dust and stellar ages.

The interpretation of ultraviolet Galactic interstellar extinction curves is obscured today by accumulated assumptions, such as a purported link between the 2200 A bump and metallicity, that are not firmly supported by observations. In this paper I define extinction curves as the ratio F*/F0 of the near-infrared-to-ultraviolet spectrum of a reddened star to that of the same star without intervening material, rather than in terms of a magnitude difference, and revisit their observed properties. Special attention is given to the connection that Galactic extinction curves with a 2200 A bump retain with the ultraviolet extrapolation of the exponential extinction law defined by their near-infrared-to-optical segment. This connection leads to the classification of all extinction curves into three types. A graphical representation of these types together with their underlying exponential extinction laws demonstrates that interstellar extinction curves can be interpreted in two ways. Either they result from the mixing of distinct extinction laws associated with different particles, as traditionally assumed, or Galactic ultraviolet curves with a bump are not extinction laws proper but instead deviate from a universal exponential extinction law owing to an additional contribution from coherently forward-scattered starlight. Given the observational constraints on the interpretation of extinction curves, such as their dependence on just two parameters, and the fact that bump-like extinction curves are barely observed outside the Galaxy, the latter interpretation emerges as the only logically consistent one.

A subset of Be stars, typified by the naked-eye star gamma Cas, exhibits unusually bright and hard X-ray emission, the origin of which has remained debated for five decades. We performed high-resolution X-ray spectroscopic monitoring of gamma Cas with the Resolve instrument aboard the X-Ray Imaging and Spectroscopy Mission (XRISM). X-ray lines from the ultra-hot plasma and fluorescence from cooler material exhibit Doppler shifts consistent with orbital motion, not of the Be star itself, but of its low-mass companion (previously shown to be a white dwarf). This first evidence of orbital motion for the hard X-ray emitting plasma uniquely links it to the scenario of accretion onto the white dwarf companion. The modest line broadening further indicates that fluorescence occurs on the white dwarf surface and excludes X-ray generation in the inner parts of an accretion disc. Our findings identify gamma Cas and its analogues as the previously elusive, but long predicted class of binaries composed of a Be and a white dwarf. Identifying the origin of the hard X-rays from gamma Cas and its analogues, which represent about 10% of early-type Be stars, provides a key input for population synthesis models of massive binary evolution.

It has long been predicted that the Milky Way (MW) will eventually merge with Andromeda (M31), a view reinforced by \textit{HST} measurements indicating a small M31 transverse velocity. However, using updated \textit{Gaia}-based proper motions (PMs) and including the dynamical influence of the Large Magellanic Cloud (LMC) and M33, Sawala et al. reported an MW--M31 merger probability of $\sim$50\% within 10 Gyr, leaving the fate of the Local Group (LG) uncertain. Adopting their semi-analytic framework, we revisit this problem with the latest and most precise \textit{Gaia}-based PMs for M31 and M33, corrected for systematic offsets in \textit{Gaia} astrometry. In our fiducial model, the MW--M31 merger probability rises to 90\%, with a median merger time of $6.5_{-1.5}^{+1.3}$ Gyr, broadly restoring the classical picture. A sensitivity analysis shows that the merger probability depends strongly on the adopted M31 PM through two channels: a direct effect via the radial-tangential balance of the MW-M31 orbit, and a satellite-mediated effect, where the M31 PM fixes the orbital plane and determines how satellite-induced barycentric reflex motions project onto it, either promoting or suppressing a merger. Given this sensitivity, current measurements, while favoring a high merger probability, remain inconclusive, spanning from 64.7\% to 100\% across the 2$σ$ PM region. Future PM measurements with uncertainty of $\lesssim2\,\upmu\mathrm{as\,yr^{-1}}$ will be required to reach a firm conclusion, i.e., to constrain the probability range within 10\% at the 2$σ$ level.

Large-scale structure (LSS) and tracer bias connect observable populations to the cosmic matter distribution. While galaxies are standard tracers, transient events such as gravitational-wave sources can also probe LSS despite large localization uncertainties. Fast radio bursts (FRBs), owing to their cosmological distances and dispersion-measure information, provide a promising complementary tracer of LSS. However, most FRBs lack precise localization and redshift measurements, introducing severe angular and radial errors that dilute the clustering signal. Here we construct an end-to-end framework to infer the linear large-scale bias of unlocalized FRB populations using the isotropic two-point correlation function. Our pipeline adopts the Landy-Szalay estimator with noise-matched random catalogs, a Monte Carlo forward model accounting for localization smearing, and likelihood-based inference with covariance matrices from lognormal mock samples. We test the method on synthetic FRB samples at redshifts z=0.3, 0.5, and 0.7 with injected bias values b=1.2, 1.5, and 2.0. The measured correlation functions closely follow smeared theoretical predictions, confirming that positional uncertainty dominates clustering suppression. Despite sample variance, the inferred bias posteriors recover the true inputs and preserve relative bias ordering. Discrimination is strongest at low redshift and weakens at higher redshift, where low-bias populations become poorly constrained. Our results demonstrate that meaningful large-scale clustering information can be extracted from poorly localized FRBs when smearing effects are properly modeled, establishing a practical route for future FRB-based LSS investigations.

A key goal in large-scale structure analysis is to extract multi-scale information to improve cosmological parameter constraints. In particular, higher-order derivative fields are especially valuable as they capture the geometric and topological information of the cosmic web that is highly sensitive to cosmological parameters. Traditional derivative-based methods, such as finite-difference or Fourier approaches, suffer from noise amplification at small scales and cannot stably capture multi-scale features. We present a robust two-step framework: first, stable multi-scale arbitrary-order derivatives are obtained via Hermite-Gaussian convolutional filters that suppress small-scale noise; second, a tanh nonlinear transformation compresses extreme density contrasts and enhances the visibility of cosmic web structures. Using the Quijote simulations, we show that combining multi-scale first-order spectra yields improvements of 1.2-3.0 times across all seven cosmological parameters, while multi-order spectra at a fixed scale provide 1.3-2.9 times gains. The most comprehensive combination achieves nominal gains of 2.0-5.3 times. Our method offers a robust approach to extracting additional cosmological information for future surveys.

Magnetars are highly magnetized neutron stars (NSs) whose evolution and radiation are governed by the decay and/or reconfiguration of their magnetic fields. The origin of magnetars remains an open question, with proposed progenitor scenarios including core-collapse (CC) of very massive stars ($\ge 25~M_\odot$) or non-very massive stars ($8<M_*<25~M_\odot$), mergers of stellar systems, and accretion-induced collapse (AIC) of white dwarfs (WDs). Investigating the environments of magnetars can offer valuable clues to this issue. In this work, we study the local (a radius of $0.87^\circ$, $\sim 100$ pc at 6.6 kpc) stellar environment of SGR 1935+2154, which is spatially associated with the supernova remnant (SNR) G57.2+0.8, based on astrometry from Gaia DR3 and multi-band photometry from optical to infrared (IR). We discover that the upper limit of the surface density of massive stars around SGR 1935+2154 is only a quarter of that of the solar neighborhood, where the star formation rate is modest in the Galaxy. This quiet environment implies that the magnetar was likely formed by the CC of either a non-very massive star or a binary merger product rather than the CC of a very massive star. Although alternative channels cannot be excluded, their probabilities may be substantially lower. The studies of magnetars associated with SNRs consistently favor non-very massive progenitors, implying that such progenitors may produce a considerable fraction of magnetars. We also backtrack the trajectories of SGR 1935+2154 and its surrounding stars to search for its potential massive companions, yet no such companions are found.

Atomic hydrogen (HI) plays a fundamental role in fueling star formation in galaxies. However, the behavior of HI gas in interacting systems, particularly galaxy pairs, remains elusive. In this work, we investigate the HI content of major mergers by cross-matching the extragalactic HI catalog from the FAST All-Sky HI Survey (FASHI) with a previously established sample of isolated galaxy pairs. With the superior sensitivity of FAST, we have constructed the largest sample of major mergers with HI detections, consisting of $440$ galaxy pairs: $364$ spiral-spiral (S+S) and $76$ spiral-elliptical (S+E) systems. We examine the HI gas fraction ($f_{\mathrm{HI}}$), star formation rate (SFR) and HI star formation efficiency ($\mathrm{SFE_{HI}}=\mathrm{SFR}/M_{\rm HI}$) for individual galaxies in pairs. The control sample is matched in both stellar mass and redshift. We find that paired galaxies, particularly those in pairs with small projected separations ($d_{\mathrm{p}}<50\ h^{-1}\mathrm{kpc}$), exhibit systematically lower (by $8.8\%$) HI gas fractions compared to the control galaxies. The SFR is enhanced for galaxies in S+S pairs. $\mathrm{SFE_{HI}}$ is $\sim15\%$ higher for galaxies in S+S pairs than in the control galaxies, while spiral galaxies in S+E pairs show no significant difference in $\mathrm{SFE_{HI}}$ compared to the control sample. These findings suggest that the merging process triggers efficient HI gas depletion and enhances star formation, especially in close S+S pairs. Notably, our sample includes $26$ red spirals in paired systems. These galaxies exhibit HI deficiency and suppressed star formation activity compared to the isolated galaxies, indicating that interactions may affect quiescent spirals differently, potentially due to mechanisms similar to ellipticals.

The outskirts of galaxy clusters present a promising avenue for constraining cluster masses in a way that is robust to the impact of baryonic physics. We assess the accuracy to which the cluster infall regions can be used to for cluster mass calibration. Building on previous work, we parameterize the velocity distribution $P(v_{\rm r},v_{\rm tan}|r,M)$ of dark matter halos on scales $r \geq 5\ h^{-1}\ \rm{Mpc}$ as the product of the marginalized distribution $P(v_{\rm r}|r,M)$ and the conditional distribution $P(v_{\rm tan}|v_{\rm r},r,M)$, calibrating the radial and mass dependence of these distributions in numerical simulations. We then project our model along the line-of-sight to obtain accurate predictions for the distributions of line-of-sight velocities at a given projected radius and cluster mass $P(v_{\rm LOS}|R,M)$, which we can observe with spectroscopic survey data. With our model, we forecast that spectra from the Dark Energy Spectroscopic Instrument (DESI) can constrain cluster masses with sub-percent level precision, comparable to that of Stage IV weak lensing surveys.

Understanding the thermal structure of the outer crust of accreting neutron stars is important to interpret astronomical X-ray observations. Ground-state to ground-state $β$-decay transitions of neutron-rich nuclei comprising the crust enable Urca neutrino cooling processes that affect this thermal structure. Here we constrain the ground-state to ground-state transition strengths for the decays of $^{57}$Sc, $^{57}$Ti, and $^{59}$Ti based on experimental data. The data were obtained by combining total absorption $γ$-spectroscopy data from the SuN detection system with $β$-delayed neutron emission data from the NERO detection system at Michigan State University's National Superconducting Cyclotron Laboratory. We find $\log ft=$5.8$^{+0.3}_{-0.2}$ and $\log ft=$5.34$^{+0.08}_{-0.24}$ for the decays of $^{57}$Ti and $^{59}$Ti, respectively, and find no evidence for ground-state feeding in the decay of $^{57}$Sc. The results indicate weaker transitions than predicted by theory and indicated by previous measurements, resulting in reduced efficiency of neutrino cooling in accreted neutron star crusts in systems that exhibit X-ray superbursts.

We present Ly$α$, He II and C IV observations of 7 redshift ~ 2 radio-loud quasars observed using the Keck Cosmic Web Imager (KCWI) and compare it to observed radio jet emission using archival VLA and ALMA radio observations. We detect 80-120 kpc diameter Ly$α$ and 10-40 kpc He II and C IV emission around the targets. We find the Ly$α$ emission to be brighter in the inner 30 kpc by factors of 2-10 compared to other literature samples. We reproduce the trend for increased total luminosity for a larger area on sky, but find our targets tend to be brighter for a given area when compared to literature observations, even when adjusting for the observational sensitivity. We infer that the He II and C IV is likely powered by quasar photoionization, with the ionizing radiation likely escaping along the radio jet axis which is aligned with the He II and C IV emission. The observations agree with a two component model of the CGM where the inner CGM (< 30 kpc) is directly influenced by the host galaxies, whereas the gas motion in the outer CGM (> 30 kpc) is influenced by gas turbulence and the larger environment around the host galaxies.

The last five years have shown us that ultra-high-energy (UHE; $>$100 TeV) gamma-ray sources are ubiquitous, but the nature of these sources remain highly uncertain. UHE gamma rays can be produced via either leptonic (Inverse compton) or hadronic (pion decay) emission mechanisms. To decisively determine the emission mechanisms, multimessenger searches are essential. Neutrinos are of particular interest as they are only created via hadronic channels. In this work, we describe a metric to select high-quality UHE events from the High Altitude Water Cherenkov (HAWC) Observatory. We use this metric to search for correlations between HAWC archival data and IceCube public neutrino alerts. 24 spatial coincidences are found, which is higher than the number of events expected by random chance. Therefore, we conclude that there are likely associations between HAWC gamma rays and IceCube neutrinos, but the angular resolutions of the two instruments prevent us from conclusively making any definitive associations between the coincidences and specific astrophysical sources. More sensitive detectors are needed.

We provide a description of the code implementation and structure of Cosmology Likelihood for Observables in Euclid (CLOE), developed by members of the Euclid Consortium. CLOE is a modular Python code for computing the theoretical predictions of cosmological observables and evaluating them against state-of-the-art data from galaxy surveys such as Euclid in a unified likelihood. This primarily includes the core observables of weak gravitational lensing, photometric galaxy clustering, galaxy-galaxy lensing, and spectroscopic galaxy clustering, but also extended probes such as the clusters of galaxies and cross-correlations of galaxy positions and shapes with the cosmic microwave background. While CLOE has been developed to serve as the unified framework for the parameter inferences in Euclid, it has general capabilities that can serve the broader cosmological community. It is different from other comparable cosmological tools in that it is written entirely in Python, performs the full likelihood calculation, and includes both photometric and spectroscopic observables. We will focus on the primary probes of Euclid and will describe the overall code structure, rigorous code development practices, extensive documentation, unique features, speed optimization, and future development plans. CLOE is publicly available at https://github.com/cloe-org/cloe.

The Thomson scattering optical depth to reionisation, $τ$, one of the six parameters of the $Λ$CDM model, is primarily constrained by the large-scale E-mode polarisation of the Cosmic Microwave Background (CMB). In this work, we present the E-mode Likelihood for Cross-Analysis (elica), a multi-frequency, harmonic-space likelihood that combines all currently available large-scale satellite polarisation data, namely the Planck LFI 70 GHz channel, the Planck HFI 100 and 143 GHz channels processed with the SRoll2 map-making algorithm, and the WMAP Ka, Q, and V bands. The likelihood is built on an extension of the Hamimeche-Lewis formalism to multi-field partial-sky observations. We validate the pipeline using 500 realistic simulations and find that retaining all cross-spectra and the WMAP-LFI auto-spectrum eliminates the significant bias present when all spectra are retained, while preserving comparable uncertainties in the recovered value of $τ$. From the low-$\ell$ E-mode power spectrum alone, we obtain $τ= 0.0575_{-0.0058}^{+0.0048}$ (68% CL). Combining elica with the Planck low-$\ell$ temperature likelihood and the CamSpec high-$\ell$ likelihood, we find $τ= 0.0581_{-0.0059}^{+0.0048}$ and $\ln(10^{10}A_{\mathrm{s}}) = 3.048_{-0.012}^{+0.011}$. Including ACT{} DR6 + Planck CMB lensing and DESI DR2 BAO measurements, we derive an upper bound on the total neutrino mass of $\sum m_ν< 0.069$ eV (95% CL). Our results, obtained through careful cross-validation of all available large-scale polarisation datasets, robustly confirm that the optical depth remains relatively low. This severely constrains the possibility of explaining, or even significantly reducing, the tension between DESI-BAO and CMB observations with a high value of $τ$. The elica likelihood is publicly available.

The Cosmological Principle assumes a statistically isotropic Universe, but the Cosmic Microwave Background (CMB) exhibits some anomalous statistical features, such as the hemispherical power asymmetry, that challenge this core assumption. We aim to expand the characterization of this anomaly by investigating the uniformity of the CMB's morphological and topological properties, testing whether the asymmetry extends beyond the variance of the temperature field. We evaluate the three Minkowski Functionals (MFs) on local patches of the \textit{Planck} SMICA temperature map and compare them against realistic simulations. By fitting the local MFs to the analytical expectations for Gaussian isotropic fields, we extract local estimates of the temperature variance, the variance of the field gradients, and the goodness-of-fit. We then evaluate the amplitude and alignment of dipoles in these quantities. We confirm the highly significant variance dipole (p-value $\sim0.3\%-1.0\%$). Furthermore, we report a moderately significant dipole in the gradient variance (p-value $\sim2.6\%-3.7\%$) that is statistically independent of the previous dipole under the $Λ$CDM model. We also find a mild spatial variation in the goodness-of-fit to the Gaussian isotropic predictions (p-value $\sim2.1\%-5.6\%$). Remarkably, the \textit{Planck} dipoles for all three quantities point toward the same region of the sky. We find that the all asymmetries are well described by dipoles. The Hemispherical Asymmetry in the CMB extends beyond its local variance, as we found it to be also present in its local morphology, with a possible hint of non-Gaussianity. These results provide a more complete characterization of the Hemispherical Asymmetry, and will therefore contribute to better determine the nature of the physical effect behind it, whether cosmological, residual foregrounds, or unknown systematics. (Abridged)

Ubiquitously found in the Universe, atomic hydrogen represents up to 70% of the neutral gas composition of the Milky Way. As an adatom, hydrogen can physisorb or chemisorb onto interstellar dust grains and icy mantles, thereby contributing to the formation of H2 and, potentially, to the synthesis of more complex hydrogenated species. In addition, structures of relatively large specific surface areas -- such as silicates, amorphous carbon, graphene sheets, or water ice-host heterogeneous chemistry that is thought to facilitate the emergence of complex organic matter in astrophysical environments. Although the fundamental physical and chemical processes occurring at dust/gas interfaces are well characterized, current understanding of dust properties governing the formation of H2 and complex molecules remains incomplete. In this context, we introduce graphitic-like two-dimensional carbon nitride monolayer structures (2D-CN) as a putative molecular family of potential relevance to astrochemistry. The physicochemical and electronic properties of these materials have been extensively examined in recent years for industrial and technological applications. Here, we propose that their importance may likewise extend to interstellar and circumstellar environments. To explore this possibility, we employed Density Functional Theory (DFT) calculations to investigate the characteristics and extent of H adsorption onto C2N1, C3N1, C3N2, C3N4, C4N3, C6N6, C6N8, C9N4, and C9N7 monolayer nanosheets. We identify multiple adsorption sites over C-C bonds, above C and N atoms, and hollow (macropore) locations at which energetically favorable binding of atomic hydrogen could occur in the interstellar medium (ISM). From an astrochemical perspective, these 2D-CN structures, if formed, could therefore contribute to the physicochemical processing and evolution of hydrogen in the ISM.

In most galaxies, the central black hole accounts for no more than a percent of the total mass in stars. Recently, however, extremely over-massive black holes with ratios of 10% have been reported in dwarf galaxies at z<1 and at cosmic dawn (z>5.5) by JWST. Both findings have been interpreted as signatures of the still mysterious origins of super-massive black holes, such that most of the black hole mass was built at birth rather than through black hole accretion. Here we show that among evolved galaxies over-massive black holes are also present, indicating that overmassive BHs are not a signature unique to black hole formation channels. The first large-area sky survey of the eROSITA X-ray telescope on board SpectrRG identified 200 quasars by their luminous hard X-ray radiation. These signpost rapidly growing black holes. Complementary optical spectroscopy from the Sloan Digital Sky Survey and archival UV to IR photometric data combined with galaxy-quasar decomposition techniques allow us unbiased estimates of cosmological distances, black hole masses and host galaxy stellar masses. We securely identify a sample of over-massive black holes with BH-to-host ratios of more than 5%, which may have undergone exponential accretion spurts lasting about a billion years. Our survey identified a high space density of at least 4/Gpc^3 of overmassive black holes near cosmic noon. This indicates an accretion channel disconnected from the stellar population that cause strong deviations from galaxy scaling relations. This channel is currently not part of galaxy evolution models. The identified channel, if applicable also for the first billion years of cosmic time, can explain JWST AGN without requiring them to signify imprints of black hole seeding mechanism.

We present the VLA C-array S-band (2--4 GHz) radio continuum observations of the nearby edge-on spiral galaxy NGC 4565, a target from the Continuum Halos in Nearby Galaxies - an EVLA (CHANG-ES) Survey. We conduct rotation measure synthesis to probe the magnetic field structure and analyze the vertical radio continuum intensity profiles using the 1-D cosmic ray transportation models. The radio continuum emission of NGC 4565 is vertically compact, with a vertical-to-radial extent ratio of $\sim 1/6$. Its vertical profile is optimally described by a two-component Gaussian distribution, yielding a mean Gaussian halo scale height of $\sim 3.0$ kpc. The magnetic field is weak, predominantly disk-parallel, with an equipartition strength of $\lesssim 5\ μ$G and a rotation measure profile indicative of an axisymmetric spiral structure. Nevertheless, we identify a localized, faint vertical magnetic field component in the northeastern region, hinting at an X-shaped structure that spatially coincides with extraplanar structures detected in H I and soft X-ray emission. The CR transport modeling favors a flux-tube advection scenario, with a slow initial velocity of $v_0 \approx 60$ km s$^{-1}$, consistent with a limited energy input from star formation. Therefore, the absence of an extended radio halo can be explained by the low star formation rate, the weak magnetic field, and the inefficient CR transport. The localized X-shaped field may trace a weak, magnetically guided outflow or a tidal perturbation induced by the nearby companion. NGC 4565 is thus a key quiescent benchmark for understanding the physical conditions required to drive large-scale outflows and generate extended radio halos.

Recently, we have performed a systematic study of AGN feedback in a disk galaxy within the MACER framework. Various model predictions, including the AGN duty cycle, the correlation between black hole accretion rates and star formation rates, and the (cold) gas fraction, have been compared with observations and will be presented in a series of papers. As the second paper in this series, without adjusting any model parameters, we directly use the simulation data introduced in Paper I to compute the predicted X-ray surface brightness profile and compare it with eROSITA observations of circumgalactic medium (CGM) emission around galaxies, which provide important constraints on AGN feedback models. For this comparison, we adopt two stacked eROSITA radial profiles of X-ray surface brightness: (1) distant galaxies with log(M*/M_sun) = 10.5-11.0 at z ~ 0.02-0.10 from Y. Zhang et al. (2024), and (2) nearby L* galaxies within 50 Mpc from L. He & Z. Li (2026). We find that the average simulated profile over time is in good agreement with the stacked measurements of Y. Zhang et al. (2024) over a broad radial range (out to ~ 100 kpc). Our model predictions also match the results of L. He & Z. Li (2026) at projected radii from ~ 20 kpc to 120 kpc. Overall, the consistency between our simulations and the eROSITA data indicates that the X-ray emission detected by eROSITA is predominantly thermal in origin, rather than nonthermal, as supported by the spectral analysis presented by L. He & Z. Li (2026).

NGC 2168 (M35) serves as a fundamental benchmark for studying stellar evolution and dynamical environments at the transition between young and intermediate-age populations. We present a comprehensive analysis of the cluster's kinematic, structural, and astrophysical properties utilizing high-precision astrometry and photometry data from Gaia Data Release 3 (DR3), complemented by 2MASS data. A statistical membership assessment yields a clean sample of probable members (N ~ 1397), with mean proper motion components of mu_alpha cos(delta) = 2.278 +/- 0.006 mas/yr and mu_delta = -2.893 +/- 0.006 mas/yr, along with a mean trigonometric parallax of varpi = 1.154 +/- 0.052 mas. We derived the cluster's fundamental parameters via isochrone fitting, determining an age of 190 +/- 12 Myr, a metallicity of [M/H] = -0.048 dex, and a probabilistic distance of 840 +/- 54 pc. The radial density profile is well described by a generalized King model with beta = 1 (rc = 7.97', rcl = 36.69'), revealing the presence of a loosely bound, extended stellar halo. Furthermore, we detect a spatial elongation oriented perpendicular to the Galactic plane, likely a signature of vertical tidal heating or disk shocking. The mass function analysis exhibits a multimodal Gaussian structure, suggesting a complex dynamical formation history beyond a simple power-law distribution. Finally, orbital integration confirms NGC 2168 as a thin disk object with a maximum vertical excursion of ~171 pc, consistent with the observed vertical morphological deformation.

Recent results form the James Webb Space Telescope (JWST) report space densities for bright and massive galaxies at z>7 that far exceed expectations of theoretical models of galaxy formation, prompting a revision of our understanding of the physical processes leading to the assembly of the first luminous structures. In this work we present predictions from a realization of the GAlaxy Evolution and Assembly (GAEA) model, which implements a prescription for a variable stellar initial mass function (IMF). This prescription is inspired by high-resolution numerical simulations that account for the role of cosmic rays (CR) as regulators of the star formation rate (SFR) in giant molecular clouds. In our approach, SFR density is assumed to be a proxy for the CR density, providing a link between the IMF shape and the predicted physical conditions of the star forming interstellar medium. Our results show that, in our model framework, assuming such a variable IMF reproduces several properties of the z>6 galaxy population, with no further modification of the feedback model, including their UV luminosity functions up to z~13. In order to compare model predictions with available estimates for the galaxy stellar mass function (GSMF), we reconstruct stellar masses from the model's synthetic photometry assuming a universal IMF, reflecting standard observational practice. Under this approach, we show that the model can reproduce the evolution of the GSMF up to the highest redshifts accessible. Our findings highlight the need to consider a variable IMF shape in the error budget associated with stellar mass estimates. We show that the evolution of both the slope and normalization of the gas-phase mass metallicity relation can be used as powerful discriminant between models of early galaxy formation assuming different IMF evolution.

The location of $γ$-ray creation and emission within extra-galactic jets is a matter of active debate. One particularly well-suited source to pinpoint the location is the nearby, bright radio galaxy 3C 84, harbouring a powerful jet. Here we investigate the origin of $γ$-rays measured during a recent $γ$-ray flare, by analysing the linear polarisation signal of close-in-time very long baseline interferometry (VLBI) observations at centimetre and millimetre wavelengths. While 3C 84 is overall almost unpolarised, we find that close-in-time to the $γ$-ray flare peak regions at parsec-scale distances from the central engine shows a fractional linear polarisation increase. Under the physically well-motivated assumption of a causal relation between this polarisation enhancement and the $γ$-ray flare, and combined with insights from concurrent X-ray polarisation measurements, the $γ$-rays being created in this region is a physically motivated scenario, in a process consistent with synchrotron self-Compton.

The largest existing simulations of cosmic reionization model radiative transfer with moment methods that require a closure relation. The two most commonly used closure relations are M1 and OTVET; both close the moment hierarchy at the first moment. We explore the properties of a higher, second-order closure. We show that direct generalizations of M1 and OTVET to one higher order are physically unstable - i.e., the closure equations themselves result in unstable solutions, not just their numerical implementation. In fact, a generalization of OTVET to any order higher than the first one is unstable. We are also able to show that any local (i.e., depending only on the local moments of the radiation field, like M1) second-order closure that depends only on the radiation intensity and radiation flux, but does not explicitly depend on the radiation pressure, is physically unstable. This result restricts the choice of possible second-order closure relations.

The Local Bubble (LB) is a hot, low-density cavity in the solar neighborhood, inside which the Solar System is currently located. The X-ray emission from such bubbles is strongly governed by the gas density, temperature, and the effects of line-of-sight column density. Yet the physical processes that control the formation and evolution of this emission remain incompletely understood. We analyze a LB analogue identified within a magnetohydrodynamical simulation to investigate the key physical factors that shape its X-ray properties. In post-processing, we examine the spatial distribution, variability, and observational constraints of the X-ray emission. Our study reveals three main results: (1) Shortly after a supernova (SN), the bulk of the X-ray emission arises from a small fraction of the bubble's volume, concentrated in hot regions around recent SN sites. Approximately 95% of the X-ray luminosity originates from less than 1% of the total bubble volume. During quiescent phases without recent SNe, the emission morphology changes substantially, with X-ray-bright regions becoming more volume-filling. (2) Column density effects strongly modulate the observable X-ray signal. Gas with column densities exceeding $N_\mathrm{H} \gtrsim 10^{20} \,\mathrm{cm}^{-2}$ efficiently absorbs soft X-ray photons, limiting the depth to which observations can probe. This absorption causes a significant fraction of the sky to be obscured from external soft X-rays. Differences between active and quiescent phases further influence how much of the total bubble emission is visible from within. (3) The X-ray flux shows pronounced temporal variability on Myr timescales, with SN events producing rapid, transient luminosity enhancements, followed by steep declines due to adiabatic cooling. The total flux varies by several orders of magnitude, with SN-driven peaks fading within $10^5$ years.

Obscured active galactic nuclei (AGN) are often invoked to explain the rapid emergence of young quasars at high redshift and are crucial for building a complete census of AGN activity and black hole growth. The advent of the James Webb Space Telescope (JWST) extends the discovery space for obscured AGN into the mid-infrared (mid-IR) with unprecedented precision through reprocessed dust emission. In this work, we use deep JWST Mid-Infrared Instrument (MIRI) imaging from the MIRI Early Obscured AGN Wide Survey (MEOW), together with existing JWST Near Infrared Camera (NIRCam), spectroscopic, and Hubble Space Telescope imaging data, to identify a previously unrecognized population of obscured AGN out to z ~ 6. Using spectral energy distribution (SED) modeling of the MIRI-detected sources, we identify 883 AGN over an area of ~ 131 arcmin2 and construct the AGN bolometric luminosity function, including both obscured and unobscured sources, across five redshift bins. We find an excess in AGN abundance relative to UV-selected AGN luminosity functions, indicating a substantial obscured population missed by optical/UV surveys, with the inferred obscured fraction increasing with redshift and reaching ~ 98-99% in our highest-redshift bin, 4.5 < z < 6. We also find higher AGN abundances and obscured fractions than X-ray-based studies, consistent with a previously unrecognized population of heavily obscured, Compton-thick AGN revealed by mid-IR selection. These results suggest that a large fraction of supermassive black hole growth at early times occurs during heavily obscured phases largely inaccessible at other wavelengths.

It has been claimed that a fraction of the so-called Little Red Dots (LRDs) are characterised by exponential broad line profiles, which have been ascribed to broadening from electron scattering by an ionised cocoon. In this work, we investigate the H$α$ broad line profiles of 32 AGN, including Little Red Dots (LRDs), Little Blue Dots (LBDs), and X-ray detected sources, using high SNR and resolution spectroscopy. We find that while single Gaussian models are statistically rejected, the exponential model is not universally preferred. Lorentzian and multi-Gaussian profiles provide equally good or superior fits for the majority of the sample, with no statistical preference for exponential profiles in $\sim$60% of cases across all AGN subtypes. There are indications that exponential profiles are preferred more frequently among LBDs, indicating that exponential profiles are not a prerogative of LRDs, which actually seem to more often favour Lorentzian profiles. Furthermore, we demonstrate that exponential wings can emerge naturally from the stratification of BLR clouds in virial motion, without invoking any scattering process. More generally, we also show that stacking multiple broad lines (either from multiple objects, as done in previous works, or from different BLR components within the same object) generally yields an exponential profile, even if none of the individual profiles are exponential. Explaining the exponential profiles in terms of BLR stratification solves various observational tensions with the electron scattering interpretation. While electron scattering may play a role, there is no evidence that it dominates the line profiles and that it significantly affects the inferred black hole masses.

The abundance discrepancy problem refers to the systematic differences observed between chemical abundances derived from collisionally excited lines (CELs) and recombination lines (RLs) of heavy ions. It remains a major unsolved problem in the study of ionized nebulae and is quantified by the abundance discrepancy factor (ADF). In this work, we present a deep integral field spectroscopic dataset covering the entire Lagoon Nebula (M 8), obtained by the SDSS-V Local Volume Mapper project, at a spatial resolution of 0.21 pc per spaxel. This unique dataset allows us, for the first time, to investigate spatially resolved maps of oxygen RL intensities (O II V1), together with maps of H I RLs, heavy-ion CELs, and dust attenuation across a whole H II region. We map the electron temperature using CELs and RLs of $O^{2+}$, CELs of $N^{+}$, and the electron density using CELs of $S^{+}$. We derive CEL-based ionic and elemental oxygen abundances and, for the first time, a spatially resolved map of the RL-based $O^{2+}$ abundance in an H II region. These measurements enable the construction of the first spatially resolved ADF($O^{2+}$) map of an H II region and yield a global mean ADF of ~0.47 +/- 0.02 dex. Focusing on the central region of M 8, where ionization is dominated by the O-type star Her 36, we find radial variations in the ADF ranging between ~0.35-0.50 dex. Our findings provide novel constraints on the spatial behavior and origin of the abundance discrepancy in H II regions.

Based on abundant data from Insight-HXMT, we conducted a detailed analysis of type-C quasi-periodic oscillations (QPOs) in the black hole X-ray binary MAXI J1348-630. Type-C QPOs were intensively detected over a broad energy band, with frequencies ranging from 0.24 to 10.3 Hz, and several new evolutionary features were identified. First, although type-C QPOs reappear intermittently, they show a stable characteristic frequency around 7 Hz. This implies a characteristic spatial scale for the QPO emission region, despite large variations in outburst intensity. Second, from the hard state to the hard-intermediate state, type-C QPOs display a harder fractional rms spectrum, with the rms peak shifting toward high energies (>20 keV) and an amplitude exceeding 10 %. This hard rms spectrum favors a high-energy origin for type-C QPOs. The spectral hardening occurs simultaneously with the weakening of the compact jet, suggesting a physical connection between these two processes. Finally, we observed hysteresis in the QPO frequency-flux relation, with the hysteresis loop evolving in opposite directions between the main and mini-outbursts. This offers a new perspective on the physical differences between the two outburst types, which may arise from variations in initial magnetic field conditions.

Massive He stars are potential candidates of type Ib/c supernova (SN) progenitors. Understanding their final fates remains a key issue in astrophysics. In this work, we investigate the evolution of He stars with initial masses from 5 $M_\odot$ to 65 $M_\odot$, focusing on the presupernova (pre-SN) core structures to assess their explodability. Our simulations indicate that the final core structure is determined by the CO core mass and the central 12C mass fraction at the end of core He burning, affecting the properties of central C-burning and the locations of convective shells. The location of the last convective C-burning shell sets the mass of the C-free core, constraining the iron core mass and compactness. We found that the final compactness and iron core mass exhibit non-monotonic behavior with initial mass, suggesting that the boundary between neutron star and black hole formation is not a simple mass threshold. This is due to core C/Ne burning becoming neutrino dominated. This process drives stronger core contraction, ultimately increasing the iron core mass and the final compactness. In contrast, earlier core Ne/O/Si ignition and shell mergers inhibit core contraction, reducing both the iron core mass and final compactness. We also discuss the effects of metallicity and overshooting on the pre-SN core structure. These factors potentially affect the explodability of progenitors.

We report the confirmation of a HeII$λ$1640 emitter located at 3 pkpc from the galaxy GN-z11, at z=10.6. The detection, based on JWST NIRSpec-IFU high-resolution spectroscopy, confirms a previous claim based on medium-resolution spectroscopy. The HeII$λ$1640 identification is further supported by the independent detection of H$γ$ obtained by Übler et al. (2026) at the same location. The HeII emission is spectrally resolved in two components separated by 120 km/s. The Equivalent Width of the HeII emission is extremely high ($>$20 A). No metal lines are detected. Population III stars appear to be the most plausible explanation for the observed HeII emission. We also discuss the possible contribution from a Direct Collapse Black Hole, or a Primordial Black Hole - these scenarios are less plausible, but cannot be ruled out completely.