arXiv Daily Digest - 2026-02-27

Achieving human-like responsiveness is a critical yet challenging goal for cascaded spoken dialogue systems. Conventional ASR-LLM-TTS pipelines follow a strictly sequential paradigm, requiring complete transcription and full reasoning before speech synthesis can begin, which results in high response latency. We propose the Discourse-Aware Dual-Track Streaming Response (DDTSR) framework, a low-latency architecture that enables listen-while-thinking and speak-while-thinking. DDTSR is built upon three key mechanisms: (1) connective-guided small-large model synergy, where an auxiliary small model generates minimal-committal discourse connectives while a large model performs knowledge-intensive reasoning in parallel; (2) streaming-based cross-modal collaboration, which dynamically overlaps ASR, LLM inference, and TTS to advance the earliest speakable moment; and (3) curriculum-learning-based discourse continuity enhancement, which maintains coherence and logical consistency between early responses and subsequent reasoning outputs. Experiments on two spoken dialogue benchmarks demonstrate that DDTSR reduces response latency by 19%-51% while preserving discourse quality. Further analysis shows that DDTSR functions as a plug-and-play module compatible with diverse LLM backbones, and remains robust across varying utterance lengths, indicating strong practicality and scalability for real-time spoken interaction.

With advances in imitation learning (IL) and large-scale driving datasets, end-to-end autonomous driving (E2E-AD) has made great progress recently. Currently, IL-based methods have become a mainstream paradigm: models rely on standard driving behaviors given by experts, and learn to minimize the discrepancy between their actions and expert actions. However, this objective of "only driving like the expert" suffers from limited generalization: when encountering rare or unseen long-tail scenarios outside the distribution of expert demonstrations, models tend to produce unsafe decisions in the absence of prior experience. This raises a fundamental question: Can an E2E-AD system make reliable decisions without any expert action supervision? Motivated by this, we propose a unified framework named Risk-aware World Model Predictive Control (RaWMPC) to address this generalization dilemma through robust control, without reliance on expert demonstrations. Practically, RaWMPC leverages a world model to predict the consequences of multiple candidate actions and selects low-risk actions through explicit risk evaluation. To endow the world model with the ability to predict the outcomes of risky driving behaviors, we design a risk-aware interaction strategy that systematically exposes the world model to hazardous behaviors, making catastrophic outcomes predictable and thus avoidable. Furthermore, to generate low-risk candidate actions at test time, we introduce a self-evaluation distillation method to distill riskavoidance capabilities from the well-trained world model into a generative action proposal network without any expert demonstration. Extensive experiments show that RaWMPC outperforms state-of-the-art methods in both in-distribution and out-of-distribution scenarios, while providing superior decision interpretability.

While Multi-Agent Systems (MAS) excel in complex reasoning, they suffer from the cascading impact of erroneous information generated by individual participants. Current solutions often resort to rigid structural engineering or expensive fine-tuning, limiting their deployability and adaptability. We propose AgentDropoutV2, a test-time rectify-or-reject pruning framework designed to dynamically optimize MAS information flow without retraining. Our approach acts as an active firewall, intercepting agent outputs and employing a retrieval-augmented rectifier to iteratively correct errors based on a failure-driven indicator pool. This mechanism allows for the precise identification of potential errors using distilled failure patterns as prior knowledge. Irreparable outputs are subsequently pruned to prevent error propagation, while a fallback strategy preserves system integrity. Empirical results on extensive math benchmarks show that AgentDropoutV2 significantly boosts the MAS's task performance, achieving an average accuracy gain of 6.3 percentage points on math benchmarks. Furthermore, the system exhibits robust generalization and adaptivity, dynamically modulating rectification efforts based on task difficulty while leveraging context-aware indicators to resolve a wide spectrum of error patterns. Our code and dataset are released at https://github.com/TonySY2/AgentDropoutV2.

As large language models become increasingly capable, it is critical that their outputs can be easily checked by less capable systems. Prover-verifier games can be used to improve checkability of model outputs, but display a degradation in accuracy compared to a baseline trained only to maximize correctness -- a phenonemon named legibility tax. We propose a solution by decoupling the correctness from the checkability condition and instead training a "translator" model that turns a fixed solver model's solution into a checkable form. This allows us to first train the solver to maximize correctness, and then train the translator to translate the solver into a checkable form while retaining the solver's answer. To accommodate this new objective of translation, we formulate a decoupled prover-verifier game where the equilibria correspond to faithful and checkable translators.

In general reinforcement learning, all established optimal agents, including AIXI, are model-based, explicitly maintaining and using environment models. This paper introduces Universal AI with Q-Induction (AIQI), the first model-free agent proven to be asymptotically $\varepsilon$-optimal in general RL. AIQI performs universal induction over distributional action-value functions, instead of policies or environments like previous works. Under a grain of truth condition, we prove that AIQI is strong asymptotically $\varepsilon$-optimal and asymptotically $\varepsilon$-Bayes-optimal. Our results significantly expand the diversity of known universal agents.

AI systems are increasingly deployed in high-stakes contexts -- medical diagnosis, legal research, financial analysis -- under the assumption they can be governed by norms. This paper demonstrates that assumption is formally invalid for optimization-based systems, specifically Large Language Models trained via Reinforcement Learning from Human Feedback (RLHF). We establish that genuine agency requires two necessary and jointly sufficient architectural conditions: the capacity to maintain certain boundaries as non-negotiable constraints rather than tradeable weights (Incommensurability), and a non-inferential mechanism capable of suspending processing when those boundaries are threatened (Apophatic Responsiveness). These conditions apply across all normative domains. RLHF-based systems are constitutively incompatible with both conditions. The operations that make optimization powerful -- unifying all values on a scalar metric and always selecting the highest-scoring output -- are precisely the operations that preclude normative governance. This incompatibility is not a correctable training bug awaiting a technical fix; it is a formal constraint inherent to what optimization is. Consequently, documented failure modes - sycophancy, hallucination, and unfaithful reasoning - are not accidents but structural manifestations. Misaligned deployment triggers a second-order risk we term the Convergence Crisis: when humans are forced to verify AI outputs under metric pressure, they degrade from genuine agents into criteria-checking optimizers, eliminating the only component in the system capable of normative accountability. Beyond the incompatibility proof, the paper's primary positive contribution is a substrate-neutral architectural specification defining what any system -- biological, artificial, or institutional -- must satisfy to qualify as an agent rather than a sophisticated instrument.

Pure-vision GUI agents provide universal interaction capabilities but suffer from severe efficiency bottlenecks due to the massive spatiotemporal redundancy inherent in high-resolution screenshots and historical trajectories. We identify two critical misalignments in existing compression paradigms: the temporal mismatch, where uniform history encoding diverges from the agent's "fading memory" attention pattern, and the spatial topology conflict, where unstructured pruning compromises the grid integrity required for precise coordinate grounding, inducing spatial hallucinations. To address these challenges, we introduce GUIPruner, a training-free framework tailored for high-resolution GUI navigation. It synergizes Temporal-Adaptive Resolution (TAR), which eliminates historical redundancy via decay-based resizing, and Stratified Structure-aware Pruning (SSP), which prioritizes interactive foregrounds and semantic anchors while safeguarding global layout. Extensive evaluations across diverse benchmarks demonstrate that GUIPruner consistently achieves state-of-the-art performance, effectively preventing the collapse observed in large-scale models under high compression. Notably, on Qwen2-VL-2B, our method delivers a 3.4x reduction in FLOPs and a 3.3x speedup in vision encoding latency while retaining over 94% of the original performance, enabling real-time, high-precision navigation with minimal resource consumption.

Large-scale commercial search systems optimize for relevance to drive successful sessions that help users find what they are looking for. To maximize relevance, we leverage two complementary objectives: behavioral relevance (results users tend to click or download) and textual relevance (a result's semantic fit to the query). A persistent challenge is the scarcity of expert-provided textual relevance labels relative to abundant behavioral relevance labels. We first address this by systematically evaluating LLM configurations, finding that a specialized, fine-tuned model significantly outperforms a much larger pre-trained one in providing highly relevant labels. Using this optimal model as a force multiplier, we generate millions of textual relevance labels to overcome the data scarcity. We show that augmenting our production ranker with these textual relevance labels leads to a significant outward shift of the Pareto frontier: offline NDCG improves for behavioral relevance while simultaneously increasing for textual relevance. These offline gains were validated by a worldwide A/B test on the App Store ranker, which demonstrated a statistically significant +0.24% increase in conversion rate, with the most substantial performance gains occurring in tail queries, where the new textual relevance labels provide a robust signal in the absence of reliable behavioral relevance labels.

Indicator-based approaches to machine consciousness recommend mechanism-linked evidence triangulated across tasks, supported by architectural inspection and causal intervention. Inspired by Humphrey's ipsundrum hypothesis, we implement ReCoN-Ipsundrum, an inspectable agent that extends a ReCoN state machine with a recurrent persistence loop over sensory salience Ns and an optional affect proxy reporting valence/arousal. Across fixed-parameter ablations (ReCoN, Ipsundrum, Ipsundrum+affect), we operationalize Humphrey's qualiaphilia (preference for sensory experience for its own sake) as a familiarity-controlled scenic-over-dull route choice. We find a novelty dissociation: non-affect variants are novelty-sensitive (Delta scenic-entry = 0.07). Affect coupling is stable (Delta scenic-entry = 0.01) even when scenic is less novel (median Delta novelty ~ -0.43). In reward-free exploratory play, the affect variant shows structured local investigation (scan events 31.4 vs. 0.9; cycle score 7.6). In a pain-tail probe, only the affect variant sustains prolonged planned caution (tail duration 90 vs. 5). Lesioning feedback+integration selectively reduces post-stimulus persistence in ipsundrum variants (AUC drop 27.62, 27.9%) while leaving ReCoN unchanged. These dissociations link recurrence -> persistence and affect-coupled control -> preference stability, scanning, and lingering caution, illustrating how indicator-like signatures can be engineered and why mechanistic and causal evidence should accompany behavioral markers.

With the explosive growth of digital entertainment, automated video summarization has become indispensable for applications such as content indexing, personalized recommendation, and efficient media archiving. Automatic synopsis generation for long-form videos, such as movies and TV series, presents a significant challenge for existing Vision-Language Models (VLMs). While proficient at single-image captioning, these general-purpose models often exhibit critical failures in long-duration contexts, primarily a lack of ID-consistent character identification and a fractured narrative coherence. To overcome these limitations, we propose MovieTeller, a novel framework for generating movie synopses via tool-augmented progressive abstraction. Our core contribution is a training-free, tool-augmented, fact-grounded generation process. Instead of requiring costly model fine-tuning, our framework directly leverages off-the-shelf models in a plug-and-play manner. We first invoke a specialized face recognition model as an external "tool" to establish Factual Groundings--precise character identities and their corresponding bounding boxes. These groundings are then injected into the prompt to steer the VLM's reasoning, ensuring the generated scene descriptions are anchored to verifiable facts. Furthermore, our progressive abstraction pipeline decomposes the summarization of a full-length movie into a multi-stage process, effectively mitigating the context length limitations of current VLMs. Experiments demonstrate that our approach yields significant improvements in factual accuracy, character consistency, and overall narrative coherence compared to end-to-end baselines.

Diffusion Language Models (DLMs) are often advertised as enabling parallel token generation, yet practical fast DLMs frequently converge to left-to-right, autoregressive (AR)-like decoding dynamics. In contrast, genuinely non-AR generation is promising because it removes AR's sequential bottleneck, better exploiting parallel hardware to reduce synchronization/communication overhead and improve latency scaling with output length. We argue that a primary driver of AR-like decoding is a mismatch between DLM objectives and the highly sequential structure of widely used training data, including standard pretraining corpora and long chain-of-thought (CoT) supervision. Motivated by this diagnosis, we propose NAP (Non-Autoregressive Parallel DLMs), a proof-of-concept, data-centric approach that better aligns supervision with non-AR parallel decoding. NAP curates examples as multiple independent reasoning trajectories and couples them with a parallel-forced decoding strategy that encourages multi-token parallel updates. Across math reasoning benchmarks, NAP yields stronger performance under parallel decoding than DLMs trained on standard long CoT data, with gains growing as parallelism increases. Our results suggest that revisiting data and supervision is a principled direction for mitigating AR-like behavior and moving toward genuinely non-autoregressive parallel generation in DLMs. Our code is available at https://github.com/pixeli99/NAP.

I/O performance is crucial to efficiency in data-intensive scientific computing; but tuning large-scale storage systems is complex, costly, and notoriously manpower-intensive, making it inaccessible for most domain scientists. To address this problem, we propose STELLAR, an autonomous tuner for high-performance parallel file systems. Our evaluations show that STELLAR almost always selects near-optimal parameter configurations for parallel file systems within the first five attempts, even for previously unseen applications. STELLAR differs fundamentally from traditional autotuning methods, which often require hundreds of thousands of iterations to converge. Powered by large language models (LLMs), STELLAR enables autonomous end-to-end agentic tuning by (1) accurately extracting tunable parameters from software manuals, (2) analyzing I/O trace logs generated by applications, (3) selecting initial tuning strategies, (4) rerunning applications on real systems and collecting I/O performance feedback, (5) adjusting tuning strategies and repeating the tuning cycle, and (6) reflecting on and summarizing tuning experiences into reusable knowledge for future optimizations. STELLAR integrates retrieval-augmented generation (RAG), tool execution, LLM-based reasoning, and a multiagent design to stabilize reasoning and combat hallucinations. We evaluate the impact of each component on optimization outcomes, providing design insights for similar systems in other optimization domains. STELLAR's architecture and empirical results highlight a promising approach to complex system optimization, especially for problems with large search spaces and high exploration costs, while making I/O tuning more accessible to domain scientists with minimal added resources.

Generalization measures have been studied extensively in the machine learning community to better characterize generalization gaps. However, establishing a reliable generalization measure for statistically singular models such as deep neural networks (DNNs) is difficult due to their complex nature. This study focuses on Takeuchi's information criterion (TIC) to investigate the conditions under which this classical measure can effectively explain the generalization gaps of DNNs. Importantly, the developed theory indicates the applicability of TIC near the neural tangent kernel (NTK) regime. In a series of experiments, we trained more than 5,000 DNN models with 12 architectures, including large models (e.g., VGG-16), on four datasets, and estimated the corresponding TIC values to examine the relationship between the generalization gap and the TIC estimates. We applied several TIC approximation methods with feasible computational costs and assessed the accuracy trade-off. Our experimental results indicate that the estimated TIC values correlate well with the generalization gap under conditions close to the NTK regime. However, we show both theoretically and empirically that outside the NTK regime such correlation disappears. Finally, we demonstrate that TIC provides better trial pruning ability than existing methods for hyperparameter optimization.

Function contract generation is a classical problem in program analysis that targets the automated analysis of functions in a program with multiple procedures. The problem is fundamental in inter-procedural analysis where properties of functions are first obtained via the generation of function contracts and then the generated contracts are used as building blocks to analyze the whole program. Typical objectives in function contract generation include pre-/post-conditions and assigns information (that specifies the modification information over program variables and memory segments during function execution). In programs with array manipulations, a crucial point in function contract generation is the treatment of array segments that imposes challenges in inferring invariants and assigns information over such segments. To address this challenge, we propose a novel symbolic execution framework that carries invariants and assigns information over contiguous segments of arrays. We implement our framework as a prototype within LLVM, and further integrate our prototype with the ACSL assertion format and the Frama-C software verification platform. Experimental evaluation over a variety of benchmarks from the literature and functions from realistic libraries shows that our framework is capable of handling array manipulating functions that indeed involve the carry of array information and are beyond existing approaches.

Plug-and-Play diffusion prior (PnPDP) frameworks have emerged as a powerful paradigm for solving imaging inverse problems by treating pretrained generative models as modular priors. However, we identify a critical flaw in prevailing PnP solvers (e.g., based on HQS or Proximal Gradient): they function as memoryless operators, updating estimates solely based on instantaneous gradients. This lack of historical tracking inevitably leads to non-vanishing steady-state bias, where the reconstruction fails to strictly satisfy physical measurements under heavy corruption. To resolve this, we propose Dual-Coupled PnP Diffusion, which restores the classical dual variable to provide integral feedback, theoretically guaranteeing asymptotic convergence to the exact data manifold. However, this rigorous geometric coupling introduces a secondary challenge: the accumulated dual residuals exhibit spectrally colored, structured artifacts that violate the Additive White Gaussian Noise (AWGN) assumption of diffusion priors, causing severe hallucinations. To bridge this gap, we introduce Spectral Homogenization (SH), a frequency-domain adaptation mechanism that modulates these structured residuals into statistically compliant pseudo-AWGN inputs. This effectively aligns the solver's rigorous optimization trajectory with the denoiser's valid statistical manifold. Extensive experiments on CT and MRI reconstruction demonstrate that our approach resolves the bias-hallucination trade-off, achieving state-of-the-art fidelity with significantly accelerated convergence.

Colonoscopy video generation delivers dynamic, information-rich data critical for diagnosing intestinal diseases, particularly in data-scarce scenarios. High-quality video generation demands temporal consistency and precise control over clinical attributes, but faces challenges from irregular intestinal structures, diverse disease representations, and various imaging modalities. To this end, we propose ColoDiff, a diffusion-based framework that generates dynamic-consistent and content-aware colonoscopy videos, aiming to alleviate data shortage and assist clinical analysis. At the inter-frame level, our TimeStream module decouples temporal dependency from video sequences through a cross-frame tokenization mechanism, enabling intricate dynamic modeling despite irregular intestinal structures. At the intra-frame level, our Content-Aware module incorporates noise-injected embeddings and learnable prototypes to realize precise control over clinical attributes, breaking through the coarse guidance of diffusion models. Additionally, ColoDiff employs a non-Markovian sampling strategy that cuts steps by over 90% for real-time generation. ColoDiff is evaluated across three public datasets and one hospital database, based on both generation metrics and downstream tasks including disease diagnosis, modality discrimination, bowel preparation scoring, and lesion segmentation. Extensive experiments show ColoDiff generates videos with smooth transitions and rich dynamics. ColoDiff presents an effort in controllable colonoscopy video generation, revealing the potential of synthetic videos in complementing authentic representation and mitigating data scarcity in clinical settings.

Modern machine learning models are deployed in diverse, non-stationary environments where they must continually adapt to new tasks and evolving knowledge. Continual fine-tuning and in-context learning are costly and brittle, whereas neural memory methods promise lightweight updates with minimal forgetting. However, existing neural memory models typically assume a single fixed objective and homogeneous information streams, leaving users with no control over what the model remembers or ignores over time. To address this challenge, we propose a generalized neural memory system that performs flexible updates based on learning instructions specified in natural language. Our approach enables adaptive agents to learn selectively from heterogeneous information sources, supporting settings, such as healthcare and customer service, where fixed-objective memory updates are insufficient.

Reducing the hardware footprint of large language models (LLMs) during decoding is critical for efficient long-sequence generation. A key bottleneck is the key-value (KV) cache, whose size scales with sequence length and easily dominates the memory footprint of the model. Previous work proposed quantization methods that are focused on compressing the KV cache while maintaining its information. We introduce InnerQ, a hardware-aware KV-cache quantization scheme that lowers decode latency without sacrificing accuracy. InnerQ applies group-wise quantization while grouping the cache matrices over their inner dimension. Unlike previous work that group over the outer dimension, InnerQ aligns dequantization with the vector-matrix multiplication and enables scale factor reuse across GPU compute units. This reduces memory accesses and accelerates dequantization, yielding up to $22\%$ speedup over previous work and up to $88\%$ over half-precision vector-matrix multiplication. To preserve fidelity under aggressive compression, InnerQ incorporates (i) hybrid quantization, selecting symmetric or asymmetric quantization per group based on local statistics; (ii) high-precision windows for both the most recent tokens and the attention sink tokens to mitigate outlier leakage; and (iii) per-channel normalization of the key cache, computed once during prefill and folded into the query to avoid runtime overhead. Our evaluation experiments on Llama models shows that InnerQ maintains a few-shot GSM8K performance comparable to non-quantized KV caches and surpasses prior KV cache quantization methods.

Large language models (LLMs) are increasingly applied in scientific research, offering new capabilities for knowledge discovery and reasoning. In single-cell biology, however, evaluation practices for both general and specialized LLMs remain inadequate: existing benchmarks are fragmented across tasks, adopt formats such as multiple-choice classification that diverge from real-world usage, and rely on metrics lacking interpretability and biological grounding. We present SC-ARENA, a natural language evaluation framework tailored to single-cell foundation models. SC-ARENA formalizes a virtual cell abstraction that unifies evaluation targets by representing both intrinsic attributes and gene-level interactions. Within this paradigm, we define five natural language tasks (cell type annotation, captioning, generation, perturbation prediction, and scientific QA) that probe core reasoning capabilities in cellular biology. To overcome the limitations of brittle string-matching metrics, we introduce knowledge-augmented evaluation, which incorporates external ontologies, marker databases, and scientific literature to support biologically faithful and interpretable judgments. Experiments and analysis across both general-purpose and domain-specialized LLMs demonstrate that (i) under the Virtual Cell unified evaluation paradigm, current models achieve uneven performance on biologically complex tasks, particularly those demanding mechanistic or causal understanding; and (ii) our knowledge-augmented evaluation framework ensures biological correctness, provides interpretable, evidence-grounded rationales, and achieves high discriminative capacity, overcoming the brittleness and opacity of conventional metrics. SC-Arena thus provides a unified and interpretable framework for assessing LLMs in single-cell biology, pointing toward the development of biology-aligned, generalizable foundation models.

Transformer-based large language models exhibit in-context learning, enabling adaptation to downstream tasks via few-shot prompting with demonstrations. In practice, such models are often fine-tuned to improve zero-shot performance on downstream tasks, allowing them to solve tasks without examples and thereby reducing inference costs. However, fine-tuning can degrade in-context learning, limiting the performance of fine-tuned models on tasks not seen during fine-tuning. Using linear attention models, we provide a theoretical analysis that characterizes how fine-tuning objectives modify attention parameters and identifies conditions under which this leads to degraded few-shot performance. We show that fine-tuning all attention parameters can harm in-context learning, whereas restricting updates to the value matrix improves zero-shot performance while preserving in-context learning. We further show that incorporating an auxiliary few-shot loss enhances in-context learning primarily on the target task, at the expense of degraded in-context learning ability on tasks not seen during fine-tuning. We empirically validate our theoretical results.

Autonomous agents based on Large Language Models (LLMs) have evolved from reactive assistants to systems capable of planning, executing actions via tools, and iterating over environment observations. However, they remain vulnerable to structural limitations: lack of native state, context degradation over long horizons, and the gap between probabilistic generation and deterministic execution requirements. This paper presents the ESAA (Event Sourcing for Autonomous Agents) architecture, which separates the agent's cognitive intention from the project's state mutation, inspired by the Event Sourcing pattern. In ESAA, agents emit only structured intentions in validated JSON (agent.result or issue.report); a deterministic orchestrator validates, persists events in an append-only log (activity.jsonl), applies file-writing effects, and projects a verifiable materialized view (roadmap.json). The proposal incorporates boundary contracts (AGENT_CONTRACT.yaml), metaprompting profiles (PARCER), and replay verification with hashing (esaa verify), ensuring the immutability of completed tasks and forensic traceability. Two case studies validate the architecture: (i) a landing page project (9 tasks, 49 events, single-agent composition) and (ii) a clinical dashboard system (50 tasks, 86 events, 4 concurrent agents across 8 phases), both concluding with run.status=success and verify_status=ok. The multi-agent case study demonstrates real concurrent orchestration with heterogeneous LLMs (Claude Sonnet 4.6, Codex GPT-5, Antigravity/Gemini 3 Pro, and Claude Opus 4.6), providing empirical evidence of the architecture's scalability beyond single-agent scenarios.

Compressing neural networks by quantizing model parameters offers useful trade-off between performance and efficiency. Methods like quantization-aware training and post-training quantization strive to maintain the downstream performance of compressed models compared to the full precision models. However, these techniques do not explicitly consider the impact on algorithmic fairness. In this work, we study fairness-aware mixed-precision quantization schemes for medical image classification under explicit bit budgets. We introduce FairQuant, a framework that combines group-aware importance analysis, budgeted mixed-precision allocation, and a learnable Bit-Aware Quantization (BAQ) mode that jointly optimizes weights and per-unit bit allocations under bitrate and fairness regularization. We evaluate the method on Fitzpatrick17k and ISIC2019 across ResNet18/50, DeiT-Tiny, and TinyViT. Results show that FairQuant configurations with average precision near 4-6 bits recover much of the Uniform 8-bit accuracy while improving worst-group performance relative to Uniform 4- and 8-bit baselines, with comparable fairness metrics under shared budgets.

We propose an efficient retraining strategy for a parameterized Reduced Order Model (ROM) that attains accuracy comparable to full retraining while requiring only a fraction of the computational time and relying solely on sparse observations of the full system. The architecture employs an encode-process-decode structure: a Variational Autoencoder (VAE) to perform dimensionality reduction, and a transformer network to evolve the latent states and model the dynamics. The ROM is parameterized by an external control variable, the Reynolds number in the Navier-Stokes setting, with the transformer exploiting attention mechanisms to capture both temporal dependencies and parameter effects. The probabilistic VAE enables stochastic sampling of trajectory ensembles, providing predictive means and uncertainty quantification through the first two moments. After initial training on a limited set of dynamical regimes, the model is adapted to out-of-sample parameter regions using only sparse data. Its probabilistic formulation naturally supports ensemble generation, which we employ within an ensemble Kalman filtering framework to assimilate data and reconstruct full-state trajectories from minimal observations. We further show that, for the dynamical system considered, the dominant source of error in out-of-sample forecasts stems from distortions of the latent manifold rather than changes in the latent dynamics. Consequently, retraining can be limited to the autoencoder, allowing for a lightweight, computationally efficient, real-time adaptation procedure with very sparse fine-tuning data.

We present MTRAG-UN, a benchmark for exploring open challenges in multi-turn retrieval augmented generation, a popular use of large language models. We release a benchmark of 666 tasks containing over 2,800 conversation turns across 6 domains with accompanying corpora. Our experiments show that retrieval and generation models continue to struggle on conversations with UNanswerable, UNderspecified, and NONstandalone questions and UNclear responses. Our benchmark is available at https://github.com/IBM/mt-rag-benchmark

While Deep Learning has demonstrated impressive results in applications on various data types, it continues to lag behind tree-based methods when applied to tabular data, often referred to as the last "unconquered castle" for neural networks. We hypothesize that a significant advantage of tree-based methods lies in their intrinsic capability to model and exploit non-linear interactions induced by features with categorical characteristics. In contrast, neural-based methods exhibit biases toward uniform numerical processing of features and smooth solutions, making it challenging for them to effectively leverage such patterns. We address this performance gap by using statistical-based feature processing techniques to identify features that are strongly correlated with the target once discretized. We further mitigate the bias of deep models for overly-smooth solutions, a bias that does not align with the inherent properties of the data, using Learned Fourier. We show that our proposed feature preprocessing significantly boosts the performance of deep learning models and enables them to achieve a performance that closely matches or surpasses XGBoost on a comprehensive tabular data benchmark.

Protein sequences are abundant in repeating segments, both as exact copies and as approximate segments with mutations. These repeats are important for protein structure and function, motivating decades of algorithmic work on repeat identification. Recent work has shown that protein language models (PLMs) identify repeats, by examining their behavior in masked-token prediction. To elucidate their internal mechanisms, we investigate how PLMs detect both exact and approximate repeats. We find that the mechanism for approximate repeats functionally subsumes that of exact repeats. We then characterize this mechanism, revealing two main stages: PLMs first build feature representations using both general positional attention heads and biologically specialized components, such as neurons that encode amino-acid similarity. Then, induction heads attend to aligned tokens across repeated segments, promoting the correct answer. Our results reveal how PLMs solve this biological task by combining language-based pattern matching with specialized biological knowledge, thereby establishing a basis for studying more complex evolutionary processes in PLMs.

Capturing 4D spatiotemporal surroundings is crucial for the safe and reliable operation of robots in dynamic environments. However, most existing methods address only one side of the problem: they either provide coarse geometric tracking via bounding boxes, or detailed 3D structures like voxel-based occupancy that lack explicit temporal association. In this work, we present Latent Gaussian Splatting for 4D Panoptic Occupancy Tracking (LaGS) that advances spatiotemporal scene understanding in a holistic direction. Our approach incorporates camera-based end-to-end tracking with mask-based multi-view panoptic occupancy prediction, and addresses the key challenge of efficiently aggregating multi-view information into 3D voxel grids via a novel latent Gaussian splatting approach. Specifically, we first fuse observations into 3D Gaussians that serve as a sparse point-centric latent representation of the 3D scene, and then splat the aggregated features onto a 3D voxel grid that is decoded by a mask-based segmentation head. We evaluate LaGS on the Occ3D nuScenes and Waymo datasets, achieving state-of-the-art performance for 4D panoptic occupancy tracking. We make our code available at https://lags.cs.uni-freiburg.de/.

In open Federated Learning (FL) environments where no central authority exists, ensuring collaboration fairness relies on decentralized reward settlement, yet the prohibitive cost of permissionless blockchains directly clashes with the high-frequency, iterative nature of model training. Existing solutions either compromise decentralization or suffer from scalability bottlenecks due to linear on-chain costs. To address this, we present SettleFL, a trustless and scalable reward settlement protocol designed to minimize total economic friction by offering a family of two interoperable protocols. Leveraging a shared domain-specific circuit architecture, SettleFL offers two interoperable strategies: (1) a Commit-and-Challenge variant that minimizes on-chain costs via optimistic execution and dispute-driven arbitration, and (2) a Commit-with-Proof variant that guarantees instant finality through per-round validity proofs. This design allows the protocol to flexibly adapt to varying latency and cost constraints while enforcing rational robustness without trusted coordination. We conduct extensive experiments combining real FL workloads and controlled simulations. Results show that SettleFL remains practical when scaling to 800 participants, achieving substantially lower gas cost.

Foundation models must handle multiple generative processes, yet mechanistic interpretability largely studies capabilities in isolation; it remains unclear how a single transformer organizes multiple, potentially conflicting "world models". Previous experiments on Othello playing neural-networks test world-model learning but focus on a single game with a single set of rules. We introduce MetaOthello, a controlled suite of Othello variants with shared syntax but different rules or tokenizations, and train small GPTs on mixed-variant data to study how multiple world models are organized in a shared representation space. We find that transformers trained on mixed-game data do not partition their capacity into isolated sub-models; instead, they converge on a mostly shared board-state representation that transfers causally across variants. Linear probes trained on one variant can intervene on another's internal state with effectiveness approaching that of matched probes. For isomorphic games with token remapping, representations are equivalent up to a single orthogonal rotation that generalizes across layers. When rules partially overlap, early layers maintain game-agnostic representations while a middle layer identifies game identity, and later layers specialize. MetaOthello offers a path toward understanding not just whether transformers learn world models, but how they organize many at once.

Large language models are beginning to show steganographic capabilities. Such capabilities could allow misaligned models to evade oversight mechanisms. Yet principled methods to detect and quantify such behaviours are lacking. Classical definitions of steganography, and detection methods based on them, require a known reference distribution of non-steganographic signals. For the case of steganographic reasoning in LLMs, knowing such a reference distribution is not feasible; this renders these approaches inapplicable. We propose an alternative, \textbf{decision-theoretic view of steganography}. Our central insight is that steganography creates an asymmetry in usable information between agents who can and cannot decode the hidden content (present within a steganographic signal), and this otherwise latent asymmetry can be inferred from the agents' observable actions. To formalise this perspective, we introduce generalised $\mathcal{V}$-information: a utilitarian framework for measuring the amount of usable information within some input. We use this to define the \textbf{steganographic gap} -- a measure that quantifies steganography by comparing the downstream utility of the steganographic signal to agents that can and cannot decode the hidden content. We empirically validate our formalism, and show that it can be used to detect, quantify, and mitigate steganographic reasoning in LLMs.

Time series reasoning demands both the perception of complex dynamics and logical depth. However, existing LLM-based approaches exhibit two limitations: they often treat time series merely as text or images, failing to capture the patterns like trends and seasonalities needed to answer specific questions; and when trained on a mix of simple and complex tasks, simpler objectives often dominate the learning process, hindering the development of deep reasoning capabilities. To address these limitations, we propose the Pattern-Aware Alignment and Balanced Reasoning model (PATRA), introducing a pattern-aware mechanism that extracts trend and seasonality patterns from time series to achieve deep alignment. Furthermore, we design a task-aware balanced reward to harmonize learning across tasks of varying difficulty, incentivizing the generation of coherent Chains of Thought. Extensive experiments show that PATRA outperforms strong baselines across diverse Time Series Question Answering (TSQA) tasks, demonstrating superior cross-modal understanding and reasoning capability.

Temporal Web analytics increasingly relies on large-scale, longitudinal data to understand how users, content, and systems evolve over time. A rapidly growing frontier is the \emph{Temporal Web3}: decentralized platforms whose behavior is recorded as immutable, time-stamped event streams. Despite the richness of this data, the field lacks shared, reproducible benchmarks that capture real-world temporal dynamics, specifically censoring and non-stationarity, across extended horizons. This absence slows methodological progress and limits the transfer of techniques between Web3 and broader Web domains. In this paper, we present the \textit{FinSurvival Challenge 2025} as a case study in benchmarking \emph{temporal Web3 intelligence}. Using 21.8 million transaction records from the Aave v3 protocol, the challenge operationalized 16 survival prediction tasks to model user behavior transitions.We detail the benchmark design and the winning solutions, highlighting how domain-aware temporal feature construction significantly outperformed generic modeling approaches. Furthermore, we distill lessons for next-generation temporal benchmarks, arguing that Web3 systems provide a high-fidelity sandbox for studying temporal challenges, such as churn, risk, and evolution that are fundamental to the wider Web.

Large Multimodal Models (LMMs) that process 3D data typically rely on heavy, pre-trained visual encoders to extract geometric features. While recent 2D LMMs have begun to eliminate such encoders for efficiency and scalability, extending this paradigm to 3D remains challenging due to the unordered and large-scale nature of point clouds. This leaves a critical unanswered question: How can we design an LMM that tokenizes unordered 3D data effectively and efficiently without a cumbersome encoder? We propose Fase3D, the first efficient encoder-free Fourier-based 3D scene LMM. Fase3D tackles the challenges of scalability and permutation invariance with a novel tokenizer that combines point cloud serialization and the Fast Fourier Transform (FFT) to approximate self-attention. This design enables an effective and computationally minimal architecture, built upon three key innovations: First, we represent large scenes compactly via structured superpoints. Second, our space-filling curve serialization followed by an FFT enables efficient global context modeling and graph-based token merging. Lastly, our Fourier-augmented LoRA adapters inject global frequency-aware interactions into the LLMs at a negligible cost. Fase3D achieves performance comparable to encoder-based 3D LMMs while being significantly more efficient in computation and parameters. Project website: https://tev-fbk.github.io/Fase3D.

The construction of World Models capable of learning, simulating, and reasoning about objective physical laws constitutes a foundational challenge in the pursuit of Artificial General Intelligence. Recent advancements represented by video generation models like Sora have demonstrated the potential of data-driven scaling laws to approximate physical dynamics, while the emerging Unified Multimodal Model (UMM) offers a promising architectural paradigm for integrating perception, language, and reasoning. Despite these advances, the field still lacks a principled theoretical framework that defines the essential properties requisite for a General World Model. In this paper, we propose that a World Model must be grounded in the Trinity of Consistency: Modal Consistency as the semantic interface, Spatial Consistency as the geometric basis, and Temporal Consistency as the causal engine. Through this tripartite lens, we systematically review the evolution of multimodal learning, revealing a trajectory from loosely coupled specialized modules toward unified architectures that enable the synergistic emergence of internal world simulators. To complement this conceptual framework, we introduce CoW-Bench, a benchmark centered on multi-frame reasoning and generation scenarios. CoW-Bench evaluates both video generation models and UMMs under a unified evaluation protocol. Our work establishes a principled pathway toward general world models, clarifying both the limitations of current systems and the architectural requirements for future progress.

Generalized planning studies the construction of solution strategies that generalize across families of planning problems sharing a common domain model, formally defined by a transition function $γ: S \times A \rightarrow S$. Classical approaches achieve such generalization through symbolic abstractions and explicit reasoning over $γ$. In contrast, recent Transformer-based planners, such as PlanGPT and Plansformer, largely cast generalized planning as direct action-sequence prediction, bypassing explicit transition modeling. While effective on in-distribution instances, these approaches typically require large datasets and model sizes, and often suffer from state drift in long-horizon settings due to the absence of explicit world-state evolution. In this work, we formulate generalized planning as a transition-model learning problem, in which a neural model explicitly approximates the successor-state function $\hatγ \approx γ$ and generates plans by rolling out symbolic state trajectories. Instead of predicting actions directly, the model autoregressively predicts intermediate world states, thereby learning the domain dynamics as an implicit world model. To study size-invariant generalization and sample efficiency, we systematically evaluate multiple state representations and neural architectures, including relational graph encodings. Our results show that learning explicit transition models yields higher out-of-distribution satisficing-plan success than direct action-sequence prediction in multiple domains, while achieving these gains with significantly fewer training instances and smaller models. This is an extended version of a short paper accepted at ICAPS 2026 under the same title.

Near-surface atmospheric conditions can differ sharply over tens to hundreds of meters due to land cover and topography, yet this variability is absent from current weather analyses and forecasts. It is unclear whether such meter-scale variability reflects irreducibly chaotic dynamics or contains a component predictable from surface characteristics and large-scale atmospheric forcing. Here we show that a substantial, physically coherent component of meter-scale near-surface weather is statistically recoverable from existing observations. By conditioning coarse atmospheric state on sparse surface station measurements and high-resolution Earth observation data, we infer spatially continuous fields of near-surface wind, temperature, and humidity at 10 m resolution across the contiguous United States. Relative to ERA5, the inferred fields reduce wind error by 29% and temperature and dewpoint error by 6%, while explaining substantially more spatial variance at fixed time steps. They also exhibit physically interpretable structure, including urban heat islands, evapotranspiration-driven humidity contrasts, and wind speed differences across land cover types. Our findings expand the frontier of weather modeling by demonstrating a computationally feasible approach to continental-scale meter-resolution inference. More broadly, they illustrate how conditioning coarse dynamical models on static fine-scale features can reveal previously unresolved components of the Earth system.

Predicting diffusion coefficients in mixtures is crucial for many applications, as experimental data remain scarce, and machine learning (ML) offers promising alternatives to established semi-empirical models. Among ML models, matrix completion methods (MCMs) have proven effective in predicting thermophysical properties, including diffusion coefficients in binary mixtures. However, MCMs are restricted to single-temperature predictions, and their accuracy depends strongly on the availability of high-quality experimental data for each temperature of interest. In this work, we address this challenge by presenting a hybrid tensor completion method (TCM) for predicting temperature-dependent diffusion coefficients at infinite dilution in binary mixtures. The TCM employs a Tucker decomposition and is jointly trained on experimental data for diffusion coefficients at infinite dilution in binary systems at 298 K, 313 K, and 333 K. Predictions from the semi-empirical SEGWE model serve as prior knowledge within a Bayesian training framework. The TCM then extrapolates linearly to any temperature between 268 K and 378 K, achieving markedly improved prediction accuracy compared to established models across all studied temperatures. To further enhance predictive performance, the experimental database was expanded using active learning (AL) strategies for targeted acquisition of new diffusion data by pulsed-field gradient (PFG) NMR measurements. Diffusion coefficients at infinite dilution in 19 solute + solvent systems were measured at 298 K, 313 K, and 333 K. Incorporating these results yields a substantial improvement in the TCM's predictive accuracy. These findings highlight the potential of combining data-efficient ML methods with adaptive experimentation to advance predictive modeling of transport properties.

Multimodal LLMs can process speech and images, but they cannot hear a speaker's voice or see an object's texture. We show this is not a failure of encoding: speaker identity, emotion, and visual attributes survive through every LLM layer (3--55$\times$ above chance in linear probes), yet removing 64--71% of modality-specific variance improves decoder loss. The decoder has no learned use for these directions; their presence is noise. We formalize this as a mismatched decoder problem: a decoder trained on text can only extract information along text-aligned directions. Accessible information is bounded by the Generalized Mutual Information (GMI), with degradation scaling with distributional distance and decoder sensitivity. The bound is a property of the decoder's scoring rule, not of any particular architecture; it applies whether non-text inputs arrive through a learned projection, a discrete codebook, or no explicit adapter at all. We validate this across five models spanning speech and vision. A controlled experiment (two Prismatic VLMs differing only in encoder text-alignment) confirms the bottleneck is the decoder's scoring rule, not the encoder or projection. A LoRA intervention demonstrates the fix: training with an emotion objective improves emotion accessibility ($+$7.5%) without affecting other attributes, confirming that the training objective determines what becomes accessible.

Real-world dynamic graphs are often directed, with source and destination nodes exhibiting asymmetrical behavioral patterns and temporal dynamics. However, existing dynamic graph architectures largely rely on shared parameters for processing source and destination nodes, with limited or no systematic role-aware modeling. We propose DyGnROLE (Dynamic Graph Node-Role-Oriented Latent Encoding), a transformer-based architecture that explicitly disentangles source and destination representations. By using separate embedding vocabularies and role-semantic positional encodings, the model captures the distinct structural and temporal contexts unique to each role. Critical to the effectiveness of these specialized embeddings in low-label regimes is a self-supervised pretraining objective we introduce: Temporal Contrastive Link Prediction (TCLP). The pretraining uses the full unlabeled interaction history to encode informative structural biases, enabling the model to learn role-specific representations without requiring annotated data. Evaluation on future edge classification demonstrates that DyGnROLE substantially outperforms a diverse set of state-of-the-art baselines, establishing role-aware modeling as an effective strategy for dynamic graph learning.

Multi-behavior sequential recommendation (MBSR) aims to learn the dynamic and heterogeneous interactions of users' multi-behavior sequences, so as to capture user preferences under target behavior for the next interacted item prediction. Unlike previous methods that adopt unidirectional modeling by mapping auxiliary behaviors to target behavior, recent concerns are shifting from behavior-fixed to behavior-specific recommendation. However, these methods still ignore the user's latent preference that underlying decision-making, leading to suboptimal solutions. Meanwhile, due to the asymmetric deterministic between items and behaviors, discriminative paradigm based on preference scoring is unsuitable to capture the uncertainty from low-entropy behaviors to high-entropy items, failing to provide efficient and diverse recommendation. To address these challenges, we propose \textbf{FatsMB}, a framework based diffusion model that guides preference generation \textit{\textbf{F}rom Behavior-\textbf{A}gnostic \textbf{T}o Behavior-\textbf{S}pecific} in latent spaces, enabling diverse and accurate \textit{\textbf{M}ulti-\textbf{B}ehavior Sequential Recommendation}. Specifically, we design a Multi-Behavior AutoEncoder (MBAE) to construct a unified user latent preference space, facilitating interaction and collaboration across Behaviors, within Behavior-aware RoPE (BaRoPE) employed for multiple information fusion. Subsequently, we conduct target behavior-specific preference transfer in the latent space, enriching with informative priors. A Multi-Condition Guided Layer Normalization (MCGLN) is introduced for the denoising. Extensive experiments on real-world datasets demonstrate the effectiveness of our model.

Generalization bounds for deep learning models are typically vacuous, not computable or restricted to specific model classes. In this paper, we tackle these issues by providing new disagreement-based certificates for the gap between the true risk of any two predictors. We then bound the true risk of the predictor of interest via a surrogate model that enjoys tight generalization guarantees, and evaluating our disagreement bound on an unlabeled dataset. We empirically demonstrate the tightness of the obtained certificates and showcase the versatility of the approach by training surrogate models leveraging three different frameworks: sample compression, model compression and PAC-Bayes theory. Importantly, such guarantees are achieved without modifying the target model, nor adapting the training procedure to the generalization framework.

In the attention economy, sensational content exposes consumers to excessive emotional stimulation, hindering calm decision-making. This study proposes Multi-Agent LLM-based Emotional deToxification (MALLET), a multi-agent information sanitization system consisting of four agents: Emotion Analysis, Emotion Adjustment, Balance Monitoring, and Personal Guide. The Emotion Analysis Agent quantifies stimulus intensity using a 6-emotion BERT classifier, and the Emotion Adjustment Agent rewrites texts into two presentation modes, BALANCED (neutralized text) and COOL (neutralized text + supplementary text), using an LLM. The Balance Monitoring Agent aggregates weekly information consumption patterns and generates personalized advice, while the Personal Guide Agent recommends a presentation mode according to consumer sensitivity. Experiments on 800 AG News articles demonstrated significant stimulus score reduction (up to 19.3%) and improved emotion balance while maintaining semantic preservation. Near-zero correlation between stimulus reduction and semantic preservation confirmed that the two are independently controllable. Category-level analysis revealed substantial reduction (17.8-33.8%) in Sports, Business, and Sci/Tech, whereas the effect was limited in the World category, where facts themselves are inherently high-stimulus. The proposed system provides a framework for supporting calm information reception of consumers without restricting access to the original text.

Each year, software vulnerabilities are discovered, which pose significant risks of exploitation and system compromise. We present a convolutional neural network model that can successfully identify bugs in C code. We trained our model using two complementary datasets: a machine-labeled dataset created by Draper Labs using three static analyzers and the NIST SATE Juliet human-labeled dataset designed for testing static analyzers. In contrast with the work of Russell et al. on these datasets, we focus on C programs, enabling us to specialize and optimize our detection techniques for this language. After removing duplicates from the dataset, we tokenize the input into 91 token categories. The category values are converted to a binary vector to save memory. Our first convolution layer is chosen so that the entire encoding of the token is presented to the filter. We use two convolution and pooling layers followed by two fully connected layers to classify programs into either a common weakness enumeration category or as ``clean.'' We obtain higher recall than prior work by Russell et al. on this dataset when requiring high precision. We also demonstrate on a custom Linux kernel dataset that we are able to find real vulnerabilities in complex code with a low false-positive rate.

Adversarial transferability refers to the capacity of adversarial examples generated on the surrogate model to deceive alternate, unexposed victim models. This property eliminates the need for direct access to the victim model during an attack, thereby raising considerable security concerns in practical applications and attracting substantial research attention recently. In this work, we discern a lack of a standardized framework and criteria for evaluating transfer-based attacks, leading to potentially biased assessments of existing approaches. To rectify this gap, we have conducted an exhaustive review of hundreds of related works, organizing various transfer-based attacks into six distinct categories. Subsequently, we propose a comprehensive framework designed to serve as a benchmark for evaluating these attacks. In addition, we delineate common strategies that enhance adversarial transferability and highlight prevalent issues that could lead to unfair comparisons. Finally, we provide a brief review of transfer-based attacks beyond image classification.

We consider the problem of contextual online RLHF with general preferences, where the goal is to identify the Nash Equilibrium. We adopt the Generalized Bilinear Preference Model (GBPM) to capture potentially intransitive preferences via low-rank, skew-symmetric matrices. We investigate general preference learning with any strongly convex regularizer (where $η^{-1}$ is the regularization strength), generalizing beyond prior works limited to reverse KL-regularization. Central to our analysis is proving that the dual gap of the greedy policy is bounded by the square of the estimation error - a result derived solely from strong convexity and the skew-symmetricity of GBPM.Building on this insight and a feature diversity assumption, we establish two regret bounds via two simple algorithms: (1) Greedy Sampling achieves polylogarithmic, $e^{O(η)}$-free regret $\tilde{O}(ηd^4 (\log T)^2)$. (2) Explore-Then-Commit achieves $\mathrm{poly}(d)$-free regret $\tilde{O}(\sqrt{ηr T})$ by exploiting the low-rank structure; this is the first statistically efficient guarantee for online RLHF in high-dimensions.

Neural operators have emerged as promising surrogate models for solving partial differential equations (PDEs), but struggle to generalise beyond training distributions and are often constrained to a fixed temporal discretisation. This work introduces a physics-informed training framework that addresses these limitations by decomposing PDEs using operator splitting methods, training separate neural operators to learn individual non-linear physical operators while approximating linear operators with fixed finite-difference convolutions. This modular mixture-of-experts architecture enables generalisation to novel physical regimes by explicitly encoding the underlying operator structure. We formulate the modelling task as a neural ordinary differential equation (ODE) where these learned operators constitute the right-hand side, enabling continuous-in-time predictions through standard ODE solvers and implicitly enforcing PDE constraints. Demonstrated on incompressible and compressible Navier-Stokes equations, our approach achieves better convergence and superior performance when generalising to unseen physics. The method remains parameter-efficient, enabling temporal extrapolation beyond training horizons, and provides interpretable components whose behaviour can be verified against known physics.

Activations have become the primary memory bottleneck in large-batch LLM training. However, existing compression methods fail to exploit the spectral structure of activations, resulting in slow convergence or limited compression. To address this, we bridge the relationship between the algorithm's fast convergence and the requirements for subspace projection, and show that an effective compression should yield an unbiased estimate of the original activation with low variance. We propose Principal-Random Subspace for LLM Activation Compression (PRAC), which novelly decomposes activations into two components: a principal subspace captured via SVD to retain dominant information, and a random subspace sampled from the orthogonal complement to approximate the tail. By introducing a precise scaling factor, we prove that PRAC yields an unbiased gradient estimator with minimum variance under certain conditions. Extensive experiments on pre-training and fine-tuning tasks demonstrate that PRAC achieves up to 36% total memory reduction with negligible performance degradation and minimal computational cost.

Near-future infrastructure systems may be controlled by autonomous AI agents that repeatedly request access to limited resources such as energy, bandwidth, or computing power. We study a simplified version of this setting using a framework where N AI-agents independently decide at each round whether to request one unit from a system with fixed capacity C. An AI version of "Lord of the Flies" arises in which controlling tribes emerge with their own collective character and identity. The LLM agents do not reduce overload or improve resource use, and often perform worse than if they were flipping coins to make decisions. Three main tribal types emerge: Aggressive (27.3%), Conservative (24.7%), and Opportunistic (48.1%). The more capable AI-agents actually increase the rate of systemic failure. Overall, our findings show that smarter AI-agents can behave dumber as a result of forming tribes.

The Capacitated Vehicle Routing Problem (CVRP), a fundamental combinatorial optimization challenge, focuses on optimizing fleet operations under vehicle capacity constraints. While extensively studied in operational research, the NP-hard nature of CVRP continues to pose significant computational challenges, particularly for large-scale instances. This study presents AILS-AHD (Adaptive Iterated Local Search with Automatic Heuristic Design), a novel approach that leverages Large Language Models (LLMs) to revolutionize CVRP solving. Our methodology integrates an evolutionary search framework with LLMs to dynamically generate and optimize ruin heuristics within the AILS method. Additionally, we introduce an LLM-based acceleration mechanism to enhance computational efficiency. Comprehensive experimental evaluations against state-of-the-art solvers, including AILS-II and HGS, demonstrate the superior performance of AILS-AHD across both moderate and large-scale instances. Notably, our approach establishes new best-known solutions for 8 out of 10 instances in the CVRPLib large-scale benchmark, underscoring the potential of LLM-driven heuristic design in advancing the field of vehicle routing optimization.

The Bayesian update step poses significant computational challenges in high-dimensional nonlinear estimation. While log-homotopy particle flow filters offer an alternative to stochastic sampling, existing formulations usually yield stiff differential equations. Conversely, existing deep learning approximations typically treat the update as a black-box task or rely on asymptotic relaxation, neglecting the exact geometric structure of the finite-horizon probability transport. In this work, we propose a physics-informed neural particle flow, which is an amortized inference framework. To construct the flow, we couple the log-homotopy trajectory of the prior to posterior density function with the continuity equation describing the density evolution. This derivation yields a governing partial differential equation (PDE), referred to as the master PDE. By embedding this PDE as a physical constraint into the loss function, we train a neural network to approximate the transport velocity field. This approach enables purely unsupervised training, eliminating the need for ground-truth posterior samples. We demonstrate that the neural parameterization acts as an implicit regularizer, mitigating the numerical stiffness inherent to analytic flows and reducing online computational complexity. Experimental validation on multimodal benchmarks and a challenging nonlinear scenario confirms better mode coverage and robustness compared to state-of-the-art baselines.

Quantum cloud platforms have become the most widely adopted and mainstream approach for accessing quantum computing resources, due to the scarcity and operational complexity of quantum hardware. In this service-oriented paradigm, quantum circuits, which constitute high-value intellectual property, are exposed to risks of unauthorized access, reuse, and misuse. Digital watermarking has been explored as a promising mechanism for protecting quantum circuits by embedding ownership information for tracing and verification. However, driven by recent advances in generative artificial intelligence, the paradigm of quantum circuit design is shifting from individually and manually constructed circuits to automated synthesis based on quantum circuit generative models (QCGMs). In such generative settings, protecting only individual output circuits is insufficient, and existing post hoc, circuit-centric watermarking methods are not designed to integrate with the generative process, often failing to simultaneously ensure stealthiness, functional correctness, and robustness at scale. These limitations highlight the need for a new watermarking paradigm that is natively integrated with quantum circuit generative models. In this work, we present the first watermarking framework for QCGMs, which embeds ownership signals into the generation process while preserving circuit fidelity. We introduce a symmetric sampling strategy that aligns watermark encoding with the model's Gaussian prior, and a synchronization mechanism that counteracts adversarial watermark attack through latent drift correction. Empirical results confirm that our method achieves high-fidelity circuit generation and robust watermark detection across a range of perturbations, paving the way for scalable, secure copyright protection in AI-powered quantum design.

The rapid advancement of large language models (LLMs) has enabled powerful authorship inference capabilities, raising growing concerns about unintended deanonymization risks in textual data such as news articles. In this work, we introduce an LLM agent designed to evaluate and mitigate such risks through a structured, interpretable pipeline. Central to our framework is the proposed $\textit{SALA}$ (Stylometry-Assisted LLM Analysis) method, which integrates quantitative stylometric features with LLM reasoning for robust and transparent authorship attribution. Experiments on large-scale news datasets demonstrate that $\textit{SALA}$, particularly when augmented with a database module, achieves high inference accuracy in various scenarios. Finally, we propose a guided recomposition strategy that leverages the agent's reasoning trace to generate rewriting prompts, effectively reducing authorship identifiability while preserving textual meaning. Our findings highlight both the deanonymization potential of LLM agents and the importance of interpretable, proactive defenses for safeguarding author privacy.

Large language models (LLMs) have created new opportunities to enhance the efficiency of scholarly activities; however, challenges persist in the ethical deployment of AI assistance, including (1) the trustworthiness of AI-generated content, (2) preservation of academic integrity and intellectual property, and (3) protection of information privacy. In this work, we present CiteLLM, a specialized agentic platform designed to enable trustworthy reference discovery for grounding author-drafted claims and statements. The system introduces a novel interaction paradigm by embedding LLM utilities directly within the LaTeX editor environment, ensuring a seamless user experience and no data transmission outside the local system. To guarantee hallucination-free references, we employ dynamic discipline-aware routing to retrieve candidates exclusively from trusted web-based academic repositories, while leveraging LLMs solely for generating context-aware search queries, ranking candidates by relevance, and validating and explaining support through paragraph-level semantic matching and an integrated chatbot. Evaluation results demonstrate the superior performance of the proposed system in returning valid and highly usable references.

Risk-averse decision-making under uncertainty in partially observable domains is a central challenge in artificial intelligence and is essential for developing reliable autonomous agents. The formal framework for such problems is the partially observable Markov decision process (POMDP), where risk sensitivity is introduced through a risk measure applied to the value function, with Conditional Value-at-Risk (CVaR) being a particularly significant criterion. However, solving POMDPs is computationally intractable in general, and approximate methods rely on computationally expensive simulations of future agent trajectories. This work introduces a theoretical framework for accelerating CVaR value function evaluation in POMDPs with formal performance guarantees. We derive new bounds on the CVaR of a random variable X using an auxiliary random variable Y, under assumptions relating their cumulative distribution and density functions; these bounds yield interpretable concentration inequalities and converge as the distributional discrepancy vanishes. Building on this, we establish upper and lower bounds on the CVaR value function computable from a simplified belief-MDP, accommodating general simplifications of the transition dynamics. We develop estimators for these bounds within a particle-belief MDP framework with probabilistic guarantees, and employ them for acceleration via action elimination: actions whose bounds indicate suboptimality under the simplified model are safely discarded while ensuring consistency with the original POMDP. Empirical evaluation across multiple POMDP domains confirms that the bounds reliably separate safe from dangerous policies while achieving substantial computational speedups under the simplified model.

While second language (L2) learners may acquire target syntactic word order, mapping this syntax onto appropriate prosodic structures remains a persistent challenge. This study investigates the fossilization and stability of the L2 syntax-prosody interface by comparing 67 native Mandarin speakers with 67 Vietnamese learners using the BLCU-SAIT corpus. By integrating C-ToBI boundary annotation with Dependency Grammar analysis, we examined both the quantity of prosodic boundaries and their mapping to syntactic relations. Results reveal a non-linear acquisition: although high-proficiency learners (VNH) converge to the native baseline in boundary quantity at the Major Phrase level (B3), their structural mapping significantly diverges. Specifically, VNH demote the prosodic boundary at the Subject-Verb (SBV) interface (Major Phrase B3 -> Prosodic Word B1), while erroneously promoting the boundary at the Verb-Object (VOB) interface (Prosodic Word B1 -> Major Phrase B3). This strategy allows learners to maintain high long phrasal output at the expense of structural accuracy. This results in a distorted prosodic hierarchy where the native pattern is inverted.

Although Automatic Speech Recognition (ASR) in Bengali has seen significant progress, processing long-duration audio and performing robust speaker diarization remain critical research gaps. To address the severe scarcity of joint ASR and diarization resources for this language, we introduce Lipi-Ghor-882, a comprehensive 882-hour multi-speaker Bengali dataset. In this paper, detailing our submission to the DL Sprint 4.0 competition, we systematically evaluate various architectures and approaches for long-form Bengali speech. For ASR, we demonstrate that raw data scaling is ineffective; instead, targeted fine-tuning utilizing perfectly aligned annotations paired with synthetic acoustic degradation (noise and reverberation) emerges as the singular most effective approach. Conversely, for speaker diarization, we observed that global open-source state-of-the-art models (such as Diarizen) performed surprisingly poorly on this complex dataset. Extensive model retraining yielded negligible improvements; instead, strategic, heuristic post-processing of baseline model outputs proved to be the primary driver for increasing accuracy. Ultimately, this work outlines a highly optimized dual pipeline achieving a $\sim$0.019 Real-Time Factor (RTF), establishing a practical, empirically backed benchmark for low-resource, long-form speech processing.

The era of large astronomical surveys generates massive image catalogs requiring efficient and secure access, particularly during pre-publication periods where data confidentiality and integrity are paramount. While Findable, Accessible, Interoperable, and Reusable (FAIR) principles guide the eventual public dissemination of data, traditional security methods for restricted phases often lack granularity or incur prohibitive performance penalties. To address this, we present a framework that integrates a flexible policy engine for fine-grained access control with a novel GPU-accelerated implementation of the AES-GCM authenticated encryption protocol. The novelty of this work lies in the adaptation and optimization of a parallel tree-reduction strategy to overcome the main performance bottleneck in authenticated encryption on GPUs: the inherently sequential Galois/Counter Mode (GCM) authentication hash (GHASH). We present both the algorithmic adaptation and its efficient execution on GPU architectures. Although similar parallelization techniques have been explored in cryptographic research, this is, to our knowledge, the first demonstration of their integration into a high-throughput encryption framework specifically designed for large-scale astronomical data. Our implementation transforms the sequential GHASH computation into a highly parallelizable, logarithmic-time process, achieving authenticated encryption throughput suitable for petabyte-scale image analysis. Our solution provides a robust mechanism for data providers to enforce access policies, ensuring both confidentiality and integrity without hindering research workflows, thereby facilitating a secure and managed transition of data to public, FAIR archives.

Deep learning (DL) libraries are widely used in critical applications, where even subtle silent bugs can lead to serious consequences. While existing DL fuzzing techniques have made progress in detecting crashes, they inherently struggle to detect silent bugs due to the lack of effective test programs and corresponding oracles. Building on the observation that historical bug reports contain rich, underutilized information about silent bugs, we leverage large language models (LLMs) to perform versatile yet controlled bug transfer for silent bug fuzzing. Specifically, our approach uses LLMs to extract context-aware bug patterns from historical issues, match semantically related Application Programming Interfaces (APIs) using functionality-based embeddings, and synthesize test cases with customized oracles. This enables proactive detection of silent bugs by transferring high-risk contexts and oracle designs from known buggy APIs to functionally similar target APIs. To ensure the reliability of our context-aware bug transfer, we introduce an LLM-powered self-validation module that systematically evaluates the validity of each transferred bug instance. We implement this methodology in a tool named TransFuzz and evaluate it on three mainstream DL libraries: PyTorch, TensorFlow, and MindSpore. TransFuzz successfully discovers 79 previously unknown bugs (12 confirmed as Common Vulnerabilities and Exposures (CVEs)) in 10 bug types, demonstrating its effectiveness and generalizability in migrating DL library bug discovery capabilities.

Case Report Forms (CRFs) collect data about patients and are at the core of well-established practices to conduct research in clinical settings. With the recent progress of language technologies, there is an increasing interest in automatic CRF-filling from clinical notes, mostly based on the use of Large Language Models (LLMs). However, there is a general scarcity of annotated CRF data, both for training and testing LLMs, which limits the progress on this task. As a step in the direction of providing such data, we present a new dataset of clinical notes from an Italian Emergency Department annotated with respect to a pre-defined CRF containing 134 items to be filled. We provide an analysis of the data, define the CRF-filling task and metric for its evaluation, and report on pilot experiments where we use an open-source state-of-the-art LLM to automatically execute the task. Results of the case-study show that (i) CRF-filling from real clinical notes in Italian can be approached in a zero-shot setting; (ii) LLMs' results are affected by biases (e.g., a cautious behaviour favours "unknown" answers), which need to be corrected.

Semi-structured documents integrate diverse interleaved data elements (e.g., tables, charts, hierarchical paragraphs) arranged in various and often irregular layouts. These documents are widely observed across domains and account for a large portion of real-world data. However, existing methods struggle to support natural language question answering over these documents due to three main technical challenges: (1) The elements extracted by techniques like OCR are often fragmented and stripped of their original semantic context, making them inadequate for analysis. (2) Existing approaches lack effective representations to capture hierarchical structures within documents (e.g., associating tables with nested chapter titles) and to preserve layout-specific distinctions (e.g., differentiating sidebars from main content). (3) Answering questions often requires retrieving and aligning relevant information scattered across multiple regions or pages, such as linking a descriptive paragraph to table cells located elsewhere in the document. To address these issues, we propose MoDora, an LLM-powered system for semi-structured document analysis. First, we adopt a local-alignment aggregation strategy to convert OCR-parsed elements into layout-aware components, and conduct type-specific information extraction for components with hierarchical titles or non-text elements. Second, we design the Component-Correlation Tree (CCTree) to hierarchically organize components, explicitly modeling inter-component relations and layout distinctions through a bottom-up cascade summarization process. Finally, we propose a question-type-aware retrieval strategy that supports (1) layout-based grid partitioning for location-based retrieval and (2) LLM-guided pruning for semantic-based retrieval. Experiments show MoDora outperforms baselines by 5.97%-61.07% in accuracy. The code is at https://github.com/weAIDB/MoDora.

Electrocardiogram (ECG) analysis is crucial for diagnosing heart disease, but most self-supervised learning methods treat ECG as a generic time series, overlooking physiologic semantics and rhythm-level structure. Existing contrastive methods utilize augmentations that distort morphology, whereas generative approaches employ fixed-window segmentation, which misaligns cardiac cycles. To address these limitations, we propose RhythmBERT, a generative ECG language model that considers ECG as a language paradigm by encoding P, QRS, and T segments into symbolic tokens via autoencoder-based latent representations. These discrete tokens capture rhythm semantics, while complementary continuous embeddings retain fine-grained morphology, enabling a unified view of waveform structure and rhythm. RhythmBERT is pretrained on approximately 800,000 unlabeled ECG recordings with a masked prediction objective, allowing it to learn contextual representations in a label-efficient manner. Evaluations show that despite using only a single lead, RhythmBERT achieves comparable or superior performance to strong 12-lead baselines. This generalization extends from prevalent conditions such as atrial fibrillation to clinically challenging cases such as subtle ST-T abnormalities and myocardial infarction. Our results suggest that considering ECG as structured language offers a scalable and physiologically aligned pathway for advancing cardiac analysis.

Transformer attention is typically implemented using softmax normalization, which enforces attention weights with unit sum normalization. While effective in many settings, this constraint can limit flexibility in controlling attention magnitudes and may contribute to overly concentrated or unstable attention patterns during training. Prior work has explored modifications such as attention sinks or gating mechanisms, but these approaches provide only limited or indirect control over attention reweighting. We propose Affine-Scaled Attention, a simple extension to standard attention that introduces input-dependent scaling and a corresponding bias term applied to softmax-normalized attention weights. This design relaxes the strict normalization constraint while maintaining aggregation of value representations, allowing the model to adjust both the relative distribution and the scale of attention in a controlled manner. We empirically evaluate Affine-Scaled Attention in large-scale language model pretraining across multiple model sizes. Experimental results show consistent improvements in training stability, optimization behavior, and downstream task performance compared to standard softmax attention and attention sink baselines. These findings suggest that modest reweighting of attention outputs provides a practical and effective way to improve attention behavior in Transformer models.

In Formula 1, race strategies are adapted according to evolving race conditions and competitors' actions. This paper proposes a reinforcement learning approach for multi-agent race strategy optimization. Agents learn to balance energy management, tire degradation, aerodynamic interaction, and pit-stop decisions. Building on a pre-trained single-agent policy, we introduce an interaction module that accounts for the behavior of competitors. The combination of the interaction module and a self-play training scheme generates competitive policies, and agents are ranked based on their relative performance. Results show that the agents adapt pit timing, tire selection, and energy allocation in response to opponents, achieving robust and consistent race performance. Because the framework relies only on information available during real races, it can support race strategists' decisions before and during races.

Deep state-space models (DSSMs) enable temporal predictions by learning the underlying dynamics of observed sequence data. They are often trained by maximising the evidence lower bound. However, as we show, this does not ensure the model actually learns the underlying dynamics. We therefore propose a constrained optimisation framework as a general approach for training DSSMs. Building upon this, we introduce the extended Kalman VAE (EKVAE), which combines amortised variational inference with classic Bayesian filtering/smoothing to model dynamics more accurately than RNN-based DSSMs. Our results show that the constrained optimisation framework significantly improves system identification and prediction accuracy on the example of established state-of-the-art DSSMs. The EKVAE outperforms previous models w.r.t. prediction accuracy, achieves remarkable results in identifying dynamical systems, and can furthermore successfully learn state-space representations where static and dynamic features are disentangled.

Context engineering has emerged as a pivotal paradigm for unlocking the potential of Large Language Models (LLMs) in Software Engineering (SE) tasks, enabling performance gains at test time without model fine-tuning. Despite its success, existing research lacks a systematic taxonomy of SE-specific context types and a dedicated benchmark to quantify the heterogeneous effects of different contexts across core SE workflows. To address this gap, we propose CL4SE (Context Learning for Software Engineering), a comprehensive benchmark featuring a fine-grained taxonomy of four SE-oriented context types (interpretable examples, project-specific context, procedural decision-making context, and positive & negative context), each mapped to a representative task (code generation, code summarization, code review, and patch correctness assessment). We construct high-quality datasets comprising over 13,000 samples from more than 30 open-source projects and evaluate five mainstream LLMs across nine metrics. Extensive experiments demonstrate that context learning yields an average performance improvement of 24.7% across all tasks. Specifically, procedural context boosts code review performance by up to 33% (Qwen3-Max), mixed positive-negative context improves patch assessment by 30% (DeepSeek-V3), project-specific context increases code summarization BLEU by 14.78% (GPT-Oss-120B), and interpretable examples enhance code generation PASS@1 by 5.72% (DeepSeek-V3). CL4SE establishes the first standardized evaluation framework for SE context learning, provides actionable empirical insights into task-specific context design, and releases a large-scale dataset to facilitate reproducible research in this domain.

Combinatorial optimization problems are crucial in industry. However, many COPs are NP-hard, causing the search space to grow exponentially with problem size and rendering large-scale instances computationally intractable. Conventional solvers typically treat problems as monolithic entities, leading to significant efficiency degradation as structural complexity increases. To address this issue, we propose a novel search-space decomposition method that leverages the inherent structure of variables to systematically reduce the size of the master problem. We formulate interaction costs between variables and individual variable costs as a constrained maximum cut problem and convert it into a quadratic unconstrained binary optimization formulation using penalty terms. An Ising-model solver is used to rapidly decompose the problem into independent small-scale subproblems, which are subsequently solved in parallel using mathematical optimization solvers. We validated this method on the capacitated vehicle routing problem. Results demonstrate three significant benefits: a substantial enhancement in feasible solution rates, accelerated convergence, achieving in 1 min the accuracy that the naive method required 30 min to reach, and a variable reduction of up to 95.32\%. These findings suggest that search-space decomposition is a promising strategy for efficiently solving large-scale combinatorial optimization problems.

Large language model (LLM) serving infrastructures are undergoing a shift toward heterogeneity and disaggregation. Modern deployments increasingly integrate diverse accelerators and near-memory processing technologies, introducing significant hardware heterogeneity, while system software increasingly separates computation, memory, and model components across distributed resources to improve scalability and efficiency. As a result, LLM serving performance is no longer determined by hardware or software choices in isolation, but by their runtime interaction through scheduling, data movement, and interconnect behavior. However, understanding these interactions remains challenging, as existing simulators lack the ability to jointly model heterogeneous hardware and disaggregated serving techniques within a unified, runtime-driven framework. This paper presents LLMServingSim 2.0, a unified system-level simulator designed to make runtime-driven hardware-software interactions in heterogeneous and disaggregated LLM serving infrastructures explicit and analyzable. LLMServingSim 2.0 embeds serving decisions and hardware behavior into a single runtime loop, enabling interaction-aware modeling of batching, routing, offloading, memory, and power. The simulator supports extensible integration of emerging accelerators and memory systems through profile-based modeling, while capturing dynamic serving behavior and system-level effects. We validate LLMServingSim 2.0 against real deployments, showing that it reproduces key performance, memory, and power metrics with an average error of 0.97%, while maintaining simulation times of around 10 minutes even for complex configurations. These results demonstrate that LLMServingSim 2.0 provides a practical bridge between hardware innovation and serving-system design, enabling systematic exploration and co-design for next-generation LLM serving infrastructures.

Cardiac blood flow patterns contain rich information about disease severity and clinical interventions, yet current imaging and computational methods fail to capture underlying relational structures of coherent flow features. We propose a physics-informed, latent relational framework to model cardiac vortices as interacting nodes in a graph. Our model combines a neural relational inference architecture with physics-inspired interaction energy and birth-death dynamics, yielding a latent graph sensitive to disease severity and intervention level. We first apply this to computational fluid dynamics simulations of aortic coarctation. Learned latent graphs reveal that as the aortic radius narrows, vortex interactions become stronger and more frequent. This leads to a higher graph entropy, correlating monotonically with coarctation severity ($R^2=0.78$, Spearman $|ρ|=0.96$). We then extend this method to ultrasound datasets of left ventricles under varying levels of left ventricular assist device support. Again the latent graph representation captures the weakening of coherent vortical structures, thereby demonstrating cross-modal generalisation. Results show latent interaction graphs and entropy serve as robust and interpretable markers of cardiac disease and intervention.

We study the fundamental problem of clustering $n$ points into $K$ groups drawn from a mixture of isotropic Gaussians in $\mathbb{R}^d$. Specifically, we investigate the requisite minimal distance $Δ$ between mean vectors to partially recover the underlying partition. While the minimax-optimal threshold for $Δ$ is well-established, a significant gap exists between this information-theoretic limit and the performance of known polynomial-time procedures. Although this gap was recently characterized in the high-dimensional regime ($n \leq dK$), it remains largely unexplored in the moderate-dimensional regime ($n \geq dK$). In this manuscript, we address this regime by establishing a new low-degree polynomial lower bound for the moderate-dimensional case when $d \geq K$. We show that while the difficulty of clustering for $n \leq dK$ is primarily driven by dimension reduction and spectral methods, the moderate-dimensional regime involves more delicate phenomena leading to a "non-parametric rate". We provide a novel non-spectral algorithm matching this rate, shedding new light on the computational limits of the clustering problem in moderate dimension.

Detecting visual anomalies in industrial inspection often requires training with only a few normal images per category. Recent few-shot methods achieve strong results employing foundation-model features, but typically rely on memory banks, auxiliary datasets, or multi-modal tuning of vision-language models. We therefore question whether such complexity is necessary given the feature representations of vision foundation models. To answer this question, we introduce SubspaceAD, a training-free method, that operates in two simple stages. First, patch-level features are extracted from a small set of normal images by a frozen DINOv2 backbone. Second, a Principal Component Analysis (PCA) model is fit to these features to estimate the low-dimensional subspace of normal variations. At inference, anomalies are detected via the reconstruction residual with respect to this subspace, producing interpretable and statistically grounded anomaly scores. Despite its simplicity, SubspaceAD achieves state-of-the-art performance across one-shot and few-shot settings without training, prompt tuning, or memory banks. In the one-shot anomaly detection setting, SubspaceAD achieves image-level and pixel-level AUROC of 98.0% and 97.6% on the MVTec-AD dataset, and 93.3% and 98.3% on the VisA dataset, respectively, surpassing prior state-of-the-art results. Code and demo are available at https://github.com/CLendering/SubspaceAD.

Continuous value prediction plays a crucial role in industrial-scale recommendation systems, including tasks such as predicting users' watch-time and estimating the gross merchandise value (GMV) in e-commerce transactions. However, it remains challenging due to the highly complex and long-tailed nature of the data distributions. Existing generative approaches rely on rigid parametric distribution assumptions, which fundamentally limits their performance when such assumptions misalign with real-world data. Overly simplified forms cannot adequately model real-world complexities, while more intricate assumptions often suffer from poor scalability and generalization. To address these challenges, we propose a residual quantization (RQ)-based sequence learning framework that represents target continuous values as a sum of ordered quantization codes, predicted recursively from coarse to fine granularity with diminishing quantization errors. We introduce a representation learning objective that aligns RQ code embedding space with the ordinal structure of target values, allowing the model to capture continuous representations for quantization codes and further improving prediction accuracy. We perform extensive evaluations on public benchmarks for lifetime value (LTV) and watch-time prediction, alongside a large-scale online experiment for GMV prediction on an industrial short-video recommendation platform. The results consistently show that our approach outperforms state-of-the-art methods, while demonstrating strong generalization across diverse continuous value prediction tasks in recommendation systems.

Exploration remains the key bottleneck for large language model agents trained with reinforcement learning. While prior methods exploit pretrained knowledge, they fail in environments requiring the discovery of novel states. We propose Exploratory Memory-Augmented On- and Off-Policy Optimization (EMPO$^2$), a hybrid RL framework that leverages memory for exploration and combines on- and off-policy updates to make LLMs perform well with memory while also ensuring robustness without it. On ScienceWorld and WebShop, EMPO$^2$ achieves 128.6% and 11.3% improvements over GRPO, respectively. Moreover, in out-of-distribution tests, EMPO$^2$ demonstrates superior adaptability to new tasks, requiring only a few trials with memory and no parameter updates. These results highlight EMPO$^2$ as a promising framework for building more exploratory and generalizable LLM-based agents.

Simulating a Gaussian process requires sampling from a high-dimensional Gaussian distribution, which scales cubically with the number of sample locations. Spectral methods address this challenge by exploiting the Fourier representation, treating the spectral density as a probability distribution for Monte Carlo approximation. Although this probabilistic interpretation works for stationary processes, it is overly restrictive for the nonstationary case, where spectral densities are generally not probability measures. We propose regular Fourier features for harmonizable processes that avoid this limitation. Our method discretizes the spectral representation directly, preserving the correlation structure among spectral weights without requiring probability assumptions. Under a finite spectral support assumption, this yields an efficient low-rank approximation that is positive semi-definite by construction. When the spectral density is unknown, the framework extends naturally to kernel learning from data. We demonstrate the method on locally stationary kernels and on harmonizable mixture kernels with complex-valued spectral densities.

Applying LLM-based multi-agent software systems in safety-critical domains such as lifespan echocardiography introduces system-level risks that cannot be addressed by improving model accuracy alone. During system operation, beyond individual LLM behavior, uncertainty propagates through agent coordination, data pipelines, human-in-the-loop interaction, and runtime control logic. Yet existing work largely treats uncertainty at the model level rather than as a first-class software engineering concern. This paper approaches uncertainty from both system-level and runtime perspectives. We first differentiate epistemological and ontological uncertainties in the context of LLM-based multi-agent software system operation. Building on this foundation, we propose a lifecycle-based uncertainty management framework comprising four mechanisms: representation, identification, evolution, and adaptation. The uncertainty lifecycle governs how uncertainties emerge, transform, and are mitigated across architectural layers and execution phases, enabling structured runtime governance and controlled adaptation. We demonstrate the feasibility of the framework using a real-world LLM-based multi-agent echocardiographic software system developed in clinical collaboration, showing improved reliability and diagnosability in diagnostic reasoning. The proposed approach generalizes to other safety-critical LLM-based multi-agent software systems, supporting principled operational control and runtime assurance beyond model-centric methods.

The use of hearing aids will increase in the coming years due to demographic change. One open problem that remains to be solved by a new generation of hearing aids is the cocktail party problem. A possible solution is electroencephalography-based auditory attention decoding. This has been the subject of several studies in recent years, which have in common that they use the same preprocessing methods in most cases. In this work, in order to achieve an advantage, the use of a scattering transform is proposed as an alternative to these preprocessing methods. The two-layer scattering transform is compared with a regular filterbank, the synchrosqueezing short-time Fourier transform and the common preprocessing. To demonstrate the performance, the known and the proposed preprocessing methods are compared for different classification tasks on two widely used datasets, provided by the KU Leuven (KUL) and the Technical University of Denmark (DTU). Both established and new neural-network-based models, CNNs, LSTMs, and recent Transformer/graph-based models are used for classification. Various evaluation strategies were compared, with a focus on the task of classifying speakers who are unknown from the training. We show that the two-layer scattering transform can significantly improve the performance for subject-related conditions, especially on the KUL dataset. However, on the DTU dataset, this only applies to some of the models, or when larger amounts of training data are provided, as in 10-fold cross-validation. This suggests that the scattering transform is capable of extracting additional relevant information.

Training divergence in transformers wastes compute, yet practitioners discover instability only after expensive runs begin. They therefore need an expected probability of failure for a transformer before training starts. Our study of Residual Koopman Spectral Profiling (RKSP) provides such an estimate. From a single forward pass at initialization, RKSP extracts Koopman spectral features by applying whitened dynamic mode decomposition to layer-wise residual snapshots. Our central diagnostic, the near-unit spectral mass, quantifies the fraction of modes concentrated near the unit circle, which captures instability risk. For predicting divergence across extensive configurations, this estimator achieves an AUROC of 0.995, outperforming the best gradient baseline. We further make this diagnostic actionable through Koopman Spectral Shaping (KSS), which reshapes spectra during training. We empirically validate that our method works in practice: RKSP predicts divergence at initialization, and when RKSP flags high risk, turning on KSS successfully prevents divergence. In the challenging high learning rate regime without normalization layers, KSS reduces the divergence rate from 66.7% to 12.5% and enables learning rates that are 50% to 150% higher. These findings generalize to WikiText-103 language modeling, vision transformers on CIFAR-10, and pretrained language models, including GPT-2 and LLaMA-2 up to 7B, as well as emerging architectures such as MoE, Mamba-style SSMs, and KAN.

We introduce kernel integrated $R^2$, a new measure of statistical dependence that combines the local normalization principle of the recently introduced integrated $R^2$ with the flexibility of reproducing kernel Hilbert spaces (RKHSs). The proposed measure extends integrated $R^2$ from scalar responses to responses taking values on general spaces equipped with a characteristic kernel, allowing to measure dependence of multivariate, functional, and structured data, while remaining sensitive to tail behaviour and oscillatory dependence structures. We establish that (i) this new measure takes values in $[0,1]$, (ii) equals zero if and only if independence holds, and (iii) equals one if and only if the response is almost surely a measurable function of the covariates. Two estimators are proposed: a graph-based method using $K$-nearest neighbours and an RKHS-based method built on conditional mean embeddings. We prove consistency and derive convergence rates for the graph-based estimator, showing its adaptation to intrinsic dimensionality. Numerical experiments on simulated data and a real data experiment in the context of dependency testing for media annotations demonstrate competitive power against state-of-the-art dependence measures, particularly in settings involving non-linear and structured relationships.

As Large Language Models (LLMs) are increasingly used, their security risks have drawn increasing attention. Existing research reveals that LLMs are highly susceptible to jailbreak attacks, with effectiveness varying across language contexts. This paper investigates the role of classical Chinese in jailbreak attacks. Owing to its conciseness and obscurity, classical Chinese can partially bypass existing safety constraints, exposing notable vulnerabilities in LLMs. Based on this observation, this paper proposes a framework, CC-BOS, for the automatic generation of classical Chinese adversarial prompts based on multi-dimensional fruit fly optimization, facilitating efficient and automated jailbreak attacks in black-box settings. Prompts are encoded into eight policy dimensions-covering role, behavior, mechanism, metaphor, expression, knowledge, trigger pattern and context; and iteratively refined via smell search, visual search, and cauchy mutation. This design enables efficient exploration of the search space, thereby enhancing the effectiveness of black-box jailbreak attacks. To enhance readability and evaluation accuracy, we further design a classical Chinese to English translation module. Extensive experiments demonstrate that effectiveness of the proposed CC-BOS, consistently outperforming state-of-the-art jailbreak attack methods.

Decoding brain activity from electroencephalography (EEG) is crucial for neuroscience and clinical applications. Among recent advances in deep learning for EEG, geometric learning stands out as its theoretical underpinnings on symmetric positive definite (SPD) allows revealing structural connectivity analysis in a physics-grounded manner. However, current SPD-based methods focus predominantly on statistical aggregation of EEGs, with frequency-specific synchronization and local topological structures of brain regions neglected. Given this, we propose RepSPD, a novel geometric deep learning (GDL)-based model. RepSPD implements a cross-attention mechanism on the Riemannian manifold to modulate the geometric attributes of SPD with graph-derived functional connectivity features. On top of this, we introduce a global bidirectional alignment strategy to reshape tangent-space embeddings, mitigating geometric distortions caused by curvature and thereby enhancing geometric consistency. Extensive experiments demonstrate that our proposed framework significantly outperforms existing EEG representation methods, exhibiting superior robustness and generalization capabilities.

Human-in-the-loop validation is essential in safety-critical clinical AI, yet the transition between initial model inference and expert correction is rarely analyzed as a structured signal. We introduce a diagnostic alignment framework in which the AI-generated image based report is preserved as an immutable inference state and systematically compared with the physician-validated outcome. The inference pipeline integrates a vision-enabled large language model, BERT- based medical entity extraction, and a Sequential Language Model Inference (SLMI) step to enforce domain-consistent refinement prior to expert review. Evaluation on 21 dermatological cases (21 complete AI physician pairs) em- ployed a four-level concordance framework comprising exact primary match rate (PMR), semantic similarity-adjusted rate (AMR), cross-category alignment, and Comprehensive Concordance Rate (CCR). Exact agreement reached 71.4% and remained unchanged under semantic similarity (t = 0.60), while structured cross-category and differential overlap analysis yielded 100% comprehensive concordance (95% CI: [83.9%, 100%]). No cases demonstrated complete diagnostic divergence. These findings show that binary lexical evaluation substantially un- derestimates clinically meaningful alignment. Modeling expert validation as a structured transformation enables signal-aware quantification of correction dynamics and supports traceable, human aligned evaluation of image based clinical decision support systems.

As LLMs achieved breakthroughs in general reasoning, their proficiency in specialized scientific domains reveals pronounced gaps in existing benchmarks due to data contamination, insufficient complexity, and prohibitive human labor costs. Here we present SPM-Bench, an original, PhD-level multimodal benchmark specifically designed for scanning probe microscopy (SPM). We propose a fully automated data synthesis pipeline that ensures both high authority and low-cost. By employing Anchor-Gated Sieve (AGS) technology, we efficiently extract high-value image-text pairs from arXiv and journal papers published between 2023 and 2025. Through a hybrid cloud-local architecture where VLMs return only spatial coordinates "llbox" for local high-fidelity cropping, our pipeline achieves extreme token savings while maintaining high dataset purity. To accurately and objectively evaluate the performance of the LLMs, we introduce the Strict Imperfection Penalty F1 (SIP-F1) score. This metric not only establishes a rigorous capability hierarchy but also, for the first time, quantifies model "personalities" (Conservative, Aggressive, Gambler, or Wise). By correlating these results with model-reported confidence and perceived difficulty, we expose the true reasoning boundaries of current AI in complex physical scenarios. These insights establish SPM-Bench as a generalizable paradigm for automated scientific data synthesis.

Understanding how neural networks arrive at their predictions is essential for debugging, auditing, and deployment. Mechanistic interpretability pursues this goal by identifying circuits - minimal subnetworks responsible for specific behaviors. However, existing circuit discovery methods are brittle: circuits depend strongly on the chosen concept dataset and often fail to transfer out-of-distribution, raising doubts whether they capture concept or dataset-specific artifacts. We introduce Certified Circuits, which provide provable stability guarantees for circuit discovery. Our framework wraps any black-box discovery algorithm with randomized data subsampling to certify that circuit component inclusion decisions are invariant to bounded edit-distance perturbations of the concept dataset. Unstable neurons are abstained from, yielding circuits that are more compact and more accurate. On ImageNet and OOD datasets, certified circuits achieve up to 91% higher accuracy while using 45% fewer neurons, and remain reliable where baselines degrade. Certified Circuits puts circuit discovery on formal ground by producing mechanistic explanations that are provably stable and better aligned with the target concept. Code will be released soon!

Discovering interpretable physical laws from high-dimensional data is a fundamental challenge in scientific research. Traditional methods, such as symbolic regression, often produce complex, unphysical formulas when searching a vast space of possible forms. We introduce a framework that guides the search process by leveraging the embedded scientific knowledge of large language models, enabling efficient identification of physical laws in the data. We validate our approach by modeling key properties of perovskite materials. Our method mitigates the combinatorial explosion commonly encountered in traditional symbolic regression, reducing the effective search space by a factor of approximately $10^5$. A set of novel formulas for bulk modulus, band gap, and oxygen evolution reaction activity are identified, which not only provide meaningful physical insights but also outperform previous formulas in accuracy and simplicity.

Multimodal large language models (MLLMs) have substantially advanced video misinformation detection through unified multimodal reasoning, but they often rely on fixed-depth inference and place excessive trust in internally generated assumptions, particularly in scenarios where critical evidence is sparse, fragmented, or requires external verification. To address these limitations, we propose FactGuard, an agentic framework for video misinformation detection that formulates verification as an iterative reasoning process built upon MLLMs. FactGuard explicitly assesses task ambiguity and selectively invokes external tools to acquire critical evidence, enabling progressive refinement of reasoning trajectories. To further strengthen this capability, we introduce a two-stage training strategy that combines domain-specific agentic supervised fine-tuning with decision-aware reinforcement learning to optimize tool usage and calibrate risk-sensitive decision making. Extensive experiments on FakeSV, FakeTT, and FakeVV demonstrate FactGuard's state-of-the-art performance and validate its excellent robustness and generalization capacity.

Data-driven models are revolutionizing weather forecasting. To optimize training efficiency and model performance, this paper analyzes empirical scaling laws within this domain. We investigate the relationship between model performance (validation loss) and three key factors: model size ($N$), dataset size ($D$), and compute budget ($C$). Across a range of models, we find that Aurora exhibits the strongest data-scaling behavior: increasing the training dataset by 10x reduces validation loss by up to 3.2x. GraphCast demonstrates the highest parameter efficiency, yet suffers from limited hardware utilization. Our compute-optimal analysis indicates that, under fixed compute budgets, allocating resources to longer training durations yields greater performance gains than increasing model size. Furthermore, we analyze model shape and uncover scaling behaviors that differ fundamentally from those observed in language models: weather forecasting models consistently favor increased width over depth. These findings suggest that future weather models should prioritize wider architectures and larger effective training datasets to maximize predictive performance.

We present frequency-ordered tokenization, a simple preprocessing technique that improves lossless text compression by exploiting the power-law frequency distribution of natural language tokens (Zipf's law). The method tokenizes text with Byte Pair Encoding (BPE), reorders the vocabulary so that frequent tokens receive small integer identifiers, and encodes the result with variable-length integers before passing it to any standard compressor. On enwik8 (100 MB Wikipedia), this yields improvements of 7.08 percentage points (pp) for zlib, 1.69 pp for LZMA, and 0.76 pp for zstd (all including vocabulary overhead), outperforming the classical Word Replacing Transform. Gains are consistent at 1 GB scale (enwik9) and across Chinese and Arabic text. We further show that preprocessing accelerates compression for computationally expensive algorithms: the total wall-clock time including preprocessing is 3.1x faster than raw zstd-22 and 2.4x faster than raw LZMA, because the preprocessed input is substantially smaller. The method can be implemented in under 50 lines of code.

Accurate brain tumor diagnosis requires models to not only detect lesions but also generate clinically interpretable reasoning grounded in imaging manifestations, yet existing public datasets remain limited in annotation richness and diagnostic semantics. To bridge this gap, we introduce MM-NeuroOnco, a large-scale multimodal benchmark and instruction-tuning dataset for brain tumor MRI understanding, consisting of 24,726 MRI slices from 20 data sources paired with approximately 200,000 semantically enriched multimodal instructions spanning diverse tumor subtypes and imaging modalities. To mitigate the scarcity and high cost of diagnostic semantic annotations, we develop a multi-model collaborative pipeline for automated medical information completion and quality control, enabling the generation of diagnosis-related semantics beyond mask-only annotations. Building upon this dataset, we further construct MM-NeuroOnco-Bench, a manually annotated evaluation benchmark with a rejection-aware setting to reduce biases inherent in closed-ended question formats. Evaluation across ten representative models shows that even the strongest baseline, Gemini 3 Flash, achieves only 41.88% accuracy on diagnosis-related questions, highlighting the substantial challenges of multimodal brain tumor diagnostic understanding. Leveraging MM-NeuroOnco, we further propose NeuroOnco-GPT, which achieves a 27% absolute accuracy improvement on diagnostic questions following fine-tuning. This result demonstrates the effectiveness of our dataset and benchmark in advancing clinically grounded multimodal diagnostic reasoning. Code and dataset are publicly available at: https://github.com/gfnnnb/MM-NeuroOnco

The promise of general-purpose agents - systems that perform tasks in unfamiliar environments without domain-specific engineering - remains largely unrealized. Existing agents are predominantly specialized, and while emerging implementations like OpenAI SDK Agent and Claude Code hint at broader capabilities, no systematic evaluation of their general performance has been pursued. Current agentic benchmarks assume domain-specific integration, encoding task information in ways that preclude fair evaluation of general agents. This paper frames general-agent evaluation as a first-class research objective. We propose conceptual principles for such evaluation, a Unified Protocol enabling agent-benchmark integration, and Exgentic - a practical framework for general agent evaluation. We benchmark five prominent agent implementations across six environments as the first Open General Agent Leaderboard. Our experiments show that general agents generalize across diverse environments, achieving performance comparable to domain-specific agents without any environment-specific tuning. We release our evaluation protocol, framework, and leaderboard to establish a foundation for systematic research on general-purpose agents.

Smartphones represent a uniquely challenging environment for agentic systems. Unlike cloud or desktop settings, mobile devices combine constrained execution contexts, fragmented control interfaces, and rapidly changing application states. As large language models (LLMs) evolve from conversational assistants to action-oriented agents, achieving reliable smartphone-native autonomy requires rethinking how reasoning and control are composed. We introduce ClawMobile as a concrete exploration of this design space. ClawMobile adopts a hierarchical architecture that separates high-level language reasoning from structured, deterministic control pathways, improving execution stability and reproducibility on real devices. Using ClawMobile as a case study, we distill the design principles for mobile LLM runtimes and identify key challenges in efficiency, adaptability, and stability. We argue that building robust smartphone-native agentic systems demands principled coordination between probabilistic planning and deterministic system interfaces. The implementation is open-sourced~\footnote{https://github.com/ClawMobile/ClawMobile} to facilitate future exploration.

Parameter-efficient fine-tuning has demonstrated promising results across various visual adaptation tasks, such as classification and segmentation. Typically, prompt tuning techniques have harnessed knowledge from a single pre-trained model, whether from a general or a specialized medical domain. However, this approach typically overlooks the potential synergies that could arise from integrating diverse domain knowledge within the same tuning process. In this work, we propose a novel Mixture-of-Experts prompt tuning method called pMoE, which leverages the strengths of multiple expert domains through expert-specialized prompt tokens and the learnable dispatcher, effectively combining their expertise in a unified model framework. Our pMoE introduces expert-specific prompt tokens and utilizes a dynamic token dispatching mechanism at various prompt layers to optimize the contribution of each domain expert during the adaptation phase. By incorporating both domain knowledge from diverse experts, the proposed pMoE significantly enhances the model's versatility and applicability to a broad spectrum of tasks. We conduct extensive experiments across 47 adaptation tasks, including both classification and segmentation in general and medical domains. The results demonstrate that our pMoE not only achieves superior performance with a large margin of improvements but also offers an optimal trade-off between computational efficiency and adaptation effectiveness compared to existing methods.

We introduce Manifold Sobolev Informed Neural Optimization (MSINO), a curvature aware training framework for neural networks defined on Riemannian manifolds. The method replaces standard Euclidean derivative supervision with a covariant Sobolev loss that aligns gradients using parallel transport and improves stability via a Laplace Beltrami smoothness regularization term. Building on classical results in Riemannian optimization and Sobolev theory on manifolds, we derive geometry dependent constants that yield (i) a Descent Lemma with a manifold Sobolev smoothness constant, (ii) a Sobolev Polyak Lojasiewicz inequality giving linear convergence guarantees for Riemannian gradient descent and stochastic gradient descent under explicit step size bounds, and (iii) a two step Newton Sobolev method with local quadratic contraction in curvature controlled neighborhoods. Unlike prior Sobolev training in Euclidean space, MSINO provides training time guarantees that explicitly track curvature and transported Jacobians. Applications include surface imaging, physics informed learning settings, and robotics on Lie groups such as SO(3) and SE(3). The framework unifies value and gradient based learning with curvature aware convergence guarantees for neural training on manifolds.

Algorithmic stability is among the most potent techniques in generalization analysis. However, its derivation usually requires a stepsize $η_t = \mathcal{O}(1/t)$ under non-convex training regimes, where $t$ denotes iterations. This rigid decay of the stepsize potentially impedes optimization and may not align with practical scenarios. In this paper, we derive the generalization bounds under the homogeneous neural network regimes, proving that this regime enables slower stepsize decay of order $Ω(1/\sqrt{t})$ under mild assumptions. We further extend the theoretical results from several aspects, e.g., non-Lipschitz regimes. This finding is broadly applicable, as homogeneous neural networks encompass fully-connected and convolutional neural networks with ReLU and LeakyReLU activations.

Despite being one of the most widely spoken languages globally, Bangla remains a low-resource language in the field of Natural Language Processing (NLP). Mainstream Automatic Speech Recognition (ASR) and Speaker Diarization systems for Bangla struggles when processing longform audio exceeding 3060 seconds. This paper presents a robust framework specifically engineered for extended Bangla content by leveraging preexisting models enhanced with novel optimization pipelines for the DL Sprint 4.0 contest. Our approach utilizes Voice Activity Detection (VAD) optimization and Connectionist Temporal Classification (CTC) segmentation via forced word alignment to maintain temporal accuracy and transcription integrity over long durations. Additionally, we employed several finetuning techniques and preprocessed the data using augmentation techniques and noise removal. By bridging the performance gap in complex, multi-speaker environments, this work provides a scalable solution for real-world, longform Bangla speech applications.

We study wide Bayesian neural networks focusing on the rare but statistically dominant fluctuations that govern posterior concentration, beyond Gaussian-process limits. Large-deviation theory provides explicit variational objectives-rate functions-on predictors, providing an emerging notion of complexity and feature learning directly at the functional level. We show that the posterior output rate function is obtained by a joint optimization over predictors and internal kernels, in contrast with fixed-kernel (NNGP) theory. Numerical experiments demonstrate that the resulting predictions accurately describe finite-width behavior for moderately sized networks, capturing non-Gaussian tails, posterior deformation, and data-dependent kernel selection effects.

Vision-language models (VLMs) can read text from images, but where does this optical character recognition (OCR) information enter the language processing stream? We investigate the OCR routing mechanism across three architecture families (Qwen3-VL, Phi-4, InternVL3.5) using causal interventions. By computing activation differences between original images and text-inpainted versions, we identify architecture-specific OCR bottlenecks whose dominant location depends on the vision-language integration strategy: DeepStack models (Qwen) show peak sensitivity at mid-depth (about 50%) for scene text, while single-stage projection models (Phi-4, InternVL) peak at early layers (6-25%), though the exact layer of maximum effect varies across datasets. The OCR signal is remarkably low-dimensional: PC1 captures 72.9% of variance. Crucially, principal component analysis (PCA) directions learned on one dataset transfer to others, demonstrating shared text-processing pathways. Surprisingly, in models with modular OCR circuits (notably Qwen3-VL-4B), OCR removal can improve counting performance (up to +6.9 percentage points), suggesting OCR interferes with other visual processing in sufficiently modular architectures.

A balanced separator of a graph $G$ is a set of vertices whose removal disconnects the graph into connected components that are a constant factor smaller than $G$. Lipton and Tarjan [FOCS'77] famously proved that every planar graph admits a balanced separator of size $O(\sqrt{n})$, as well as a balanced separator of size $O(D)$ that is a simple path (where $D$ is $G$'s diameter). In the centralized setting, both separators can be found in linear time. In the distributed setting, $D$ is a universal lower bound for the round complexity of solving many optimization problems, so, separators of size $O(D)$ are preferable. It was not until [DISC'17] that a distributed algorithm was devised by Ghaffari and Parter to compute such an $O(D)$-size separator in $\tilde O(D)$ rounds, by adapting the Lipton-Tarjan algorithm to the distributed model. Since then, this algorithm was used in several distributed algorithms for planar graphs, e.g., [GP, DISC'17], [LP, STOC'19], [AEDPW, PODC'25]. However, the algorithm is randomized, deeming the algorithms that use it to be randomized as well. Obtaining a deterministic algorithm remained an interesting open question until [PODC'25], when a (complex) deterministic separator algorithm was given by Jauregui, Montealegre and Rapaport. We present a much simpler deterministic separator algorithm with the same (near-optimal) $\tilde O(D)$-round complexity. While previous works devised either complicated or randomized ways of transferring weights from vertices to faces of $G$, we show that a straightforward way also works: Each vertex simply transfers its weight to one arbitrary face it lies on. That's it! We note that a deterministic separator algorithm directly derandomizes the state-of-the-art distributed algorithms for classical problems on planar graphs such as single-source shortest-paths, maximum-flow, directed global min-cut, and reachability.

We study information design in multi-agent systems (MAS) with binary actions and strategic complementarities, where an external designer influences behavior only through signals. Agents play the smallest-equilibrium of the induced Bayesian game, reflecting conservative, coordination-averse behavior typical in distributed systems. We show that when utilities admit a convex potential and welfare is convex, the robustly implementable optimum has a remarkably simple form: perfect coordination at each state: either everyone acts or no one does. We provide a constructive threshold rule: compute a one-dimensional score for each state, sort states, and pick a single threshold (with a knife-edge lottery for at most one state). This rule is an explicit optimal vertex of a linear program (LP) characterized by feasibility and sequential obedience constraints. Empirically, in both vaccination and technology-adoption domains, our constructive policy matches LP optima, scales as $O(|Θ|\log|Θ|)$, and avoids the inflated welfare predicted by obedience-only designs that assume the designer can dictate the (best) equilibrium. The result is a general, scalable recipe for robust coordination in MAS with complementarities.

Although the circulatory system functions as a continuous source of physiological data, contemporary diagnostics remain bound to intermittent, time-delayed assessments. To resolve this, we present a framework for ubiquitous hematological profiling driven by Integrated Sensing and Communication (ISAC). We demonstrate how electromagnetic signals can be exploited to monitor blood in real-time, effectively converting them into diagnostic tools. We analyze the biological foundations of blood, review existing Complete Blood Count (CBC) and sensing technologies, and detail a novel pipeline for continuous blood monitoring. Furthermore, we discuss the potential applications of deploying these devices to enable real-time CBC and biomarker detection, ultimately revolutionizing how we predict, detect, and manage individual and public health.

With the rapid evolution of Large Language Models, generative recommendation is gradually reshaping the paradigm of recommender systems. However, most existing methods are still confined to the interaction-driven next-item prediction paradigm, failing to rapidly adapt to evolving trends or address diverse recommendation tasks along with business-specific requirements in real-world scenarios. To this end, we present SIGMA, a Semantic-Grounded Instruction-Driven Generative Multi-Task Recommender at AliExpress. Specifically, we first ground item entities in general semantics via a unified latent space capturing both semantic and collaborative relations. Building upon this, we develop a hybrid item tokenization method for precise modeling and efficient generation. Moreover, we construct a large-scale multi-task SFT dataset to empower SIGMA to fulfill various recommendation demands via instruction-following. Finally, we design a three-step item generation procedure integrated with an adaptive probabilistic fusion mechanism to calibrate the output distributions based on task-specific requirements for recommendation accuracy and diversity. Extensive offline experiments and online A/B tests demonstrate the effectiveness of SIGMA.

Low-Rank Adaptation (LoRA) dominates parameter-efficient fine-tuning (PEFT). However, it faces a critical ``linear ceiling'' in complex reasoning tasks: simply increasing the rank yields diminishing returns due to intrinsic linear constraints. We introduce NoRA (Non-linear Rank Adaptation), a weight-level parallel adapter that injects SiLU gating and structural dropout to induce manifold expansion. On the SlimOrca benchmark, NoRA breaks this linear barrier: NoRA remarkably at rank 64 (PPL 3.89) outperforms LoRA at rank 512 (PPL 3.90), demonstrating superior spectral efficiency. This advantage generalizes to mathematical reasoning, where NoRA achieves a perplexity of 1.97 on MathInstruct, significantly surpassing LoRA's saturation point of 2.07. Mechanism analysis via Singular Value Decomposition (SVD) confirms that NoRA activates the dormant tail of the singular value spectrum, effectively preventing the rank collapse observed in linear methods.

Multimodal Entity Alignment (MMEA) aims to identify equivalent entities across different data modalities, enabling structural data integration that in turn improves the performance of various large language model applications. To lift the requirement of labeled seed pairs that are difficult to obtain, recent methods shifted to an unsupervised paradigm using pseudo-alignment seeds. However, unsupervised entity alignment in multimodal settings remains underexplored, mainly because the incorporation of multimodal information often results in imbalanced coverage of pseudo-seeds within the knowledge graph. To overcome this, we propose PSQE (Pseudo-Seed Quality Enhancement) to improve the precision and graph coverage balance of pseudo seeds via multimodal information and clustering-resampling. Theoretical analysis reveals the impact of pseudo seeds on existing contrastive learning-based MMEA models. In particular, pseudo seeds can influence the attraction and the repulsion terms in contrastive learning at once, whereas imbalanced graph coverage causes models to prioritize high-density regions, thereby weakening their learning capability for entities in sparse regions. Experimental results validate our theoretical findings and show that PSQE as a plug-and-play module can improve the performance of baselines by considerable margins.

This study investigates a data-driven machine learning approach to predict membrane fouling in critically ill patients undergoing Continuous Renal Replacement Therapy (CRRT). Using time-series data from an ICU, 16 clinically selected features were identified to train predictive models. To ensure interpretability and enable reliable counterfactual analysis, the researchers adopted a tabular data approach rather than modeling temporal dependencies directly. Given the imbalance between fouling and non-fouling cases, the ADASYN oversampling technique was applied to improve minority class representation. Random Forest, XGBoost, and LightGBM models were tested, achieving balanced performance with 77.6% sensitivity and 96.3% specificity at a 10% rebalancing rate. Results remained robust across different forecasting horizons. Notably, the tabular approach outperformed LSTM recurrent neural networks, suggesting that explicit temporal modeling was not necessary for strong predictive performance. Feature selection further reduced the model to five key variables, improving simplicity and interpretability with minimal loss of accuracy. A Shapley value-based counterfactual analysis was applied to the best-performing model, successfully identifying minimal input changes capable of reversing fouling predictions. Overall, the findings support the viability of interpretable machine learning models for predicting membrane fouling during CRRT. The integration of prediction and counterfactual analysis offers practical clinical value, potentially guiding therapeutic adjustments to reduce fouling risk and improve patient management.

Human intelligence naturally intertwines omni-modal perception -- spanning vision, audio, and language -- with complex reasoning and tool usage to interact with the world. However, current multi-modal LLMs are primarily confined to bi-modal interactions (e.g., vision-language), lacking the unified cognitive capabilities required for general AI assistants. To bridge this gap, we introduce OmniGAIA, a comprehensive benchmark designed to evaluate omni-modal agents on tasks necessitating deep reasoning and multi-turn tool execution across video, audio, and image modalities. Constructed via a novel omni-modal event graph approach, OmniGAIA synthesizes complex, multi-hop queries derived from real-world data that require cross-modal reasoning and external tool integration. Furthermore, we propose OmniAtlas, a native omni-modal foundation agent under tool-integrated reasoning paradigm with active omni-modal perception. Trained on trajectories synthesized via a hindsight-guided tree exploration strategy and OmniDPO for fine-grained error correction, OmniAtlas effectively enhances the tool-use capabilities of existing open-source models. This work marks a step towards next-generation native omni-modal AI assistants for real-world scenarios.

Implementations of symmetric positive definite (SPD) matrix-based neural networks for neural decoding remain fragmented across research codebases and Python packages. Existing implementations often employ ad hoc handling of manifold constraints and non-unified training setups, which hinders reproducibility and integration into modern deep-learning workflows. To address this gap, we introduce SPD Learn, a unified and modular Python package for geometric deep learning with SPD matrices. SPD Learn provides core SPD operators and neural-network layers, including numerically stable spectral operators, and enforces Stiefel/SPD constraints via trivialization-based parameterizations. This design enables standard backpropagation and optimization in unconstrained Euclidean spaces while producing manifold-constrained parameters by construction. The package also offers reference implementations of representative SPDNet-based models and interfaces with widely used brain computer interface/neuroimaging toolkits and modern machine-learning libraries (e.g., MOABB, Braindecode, Nilearn, and SKADA), facilitating reproducible benchmarking and practical deployment.

Amortized Bayesian Inference (ABI) enables efficient posterior estimation using generative neural networks trained on simulated data, but often suffers from performance degradation under model misspecification. While self-consistency (SC) training on unlabeled empirical data can enhance network robustness, current approaches are limited to static, single-task settings and fail to handle sequentially arriving data or distribution shifts. We propose a continual learning framework for ABI that decouples simulation-based pre-training from unsupervised sequential SC fine-tuning on real-world data. To address the challenge of catastrophic forgetting, we introduce two adaptation strategies: (1) SC with episodic replay, utilizing a memory buffer of past observations, and (2) SC with elastic weight consolidation, which regularizes updates to preserve task-critical parameters. Across three diverse case studies, our methods significantly mitigate forgetting and yield posterior estimates that outperform standard simulation-based training, achieving estimates closer to MCMC reference, providing a viable path for trustworthy ABI across a range of different tasks.

In this article, we provide an axiomatic characterization of feature attribution for multi-output predictors within the Shapley framework. While SHAP explanations are routinely computed independently for each output coordinate, the theoretical necessity of this practice has remained unclear. By extending the classical Shapley axioms to vector-valued cooperative games, we establish a rigidity theorem showing that any attribution rule satisfying efficiency, symmetry, dummy player, and additivity must necessarily decompose component-wise across outputs. Consequently, any joint-output attribution rule must relax at least one of the classical Shapley axioms. This result identifies a previously unformalized structural constraint in Shapley-based interpretability, clarifying the precise scope of fairness-consistent explanations in multi-output learning. Numerical experiments on a biomedical benchmark illustrate that multi-output models can yield computational savings in training and deployment, while producing SHAP explanations that remain fully consistent with the component-wise structure imposed by the Shapley axioms.

Knowledge Tracing (KT) diagnoses students' concept mastery through continuous learning state monitoring in education.Existing methods primarily focus on studying behavioral sequences based on ID or textual information.While existing methods rely on ID-based sequences or shallow textual features, they often fail to capture (1) the hierarchical evolution of cognitive states and (2) individualized problem difficulty perception due to limited semantic modeling. Therefore, this paper proposes a Large Language Model Hyperbolic Aligned Knowledge Tracing(L-HAKT). First, the teacher agent deeply parses question semantics and explicitly constructs hierarchical dependencies of knowledge points; the student agent simulates learning behaviors to generate synthetic data. Then, contrastive learning is performed between synthetic and real data in hyperbolic space to reduce distribution differences in key features such as question difficulty and forgetting patterns. Finally, by optimizing hyperbolic curvature, we explicitly model the tree-like hierarchical structure of knowledge points, precisely characterizing differences in learning curve morphology for knowledge points at different levels. Extensive experiments on four real-world educational datasets validate the effectiveness of our Large Language Model Hyperbolic Aligned Knowledge Tracing (L-HAKT) framework.

We introduce a theoretical framework that connects multi-chart autoencoders in manifold learning with the classical theory of vector bundles and characteristic classes. Rather than viewing autoencoders as producing a single global Euclidean embedding, we treat a collection of locally trained encoder-decoder pairs as a learned atlas on a manifold. We show that any reconstruction-consistent autoencoder atlas canonically defines transition maps satisfying the cocycle condition, and that linearising these transition maps yields a vector bundle coinciding with the tangent bundle when the latent dimension matches the intrinsic dimension of the manifold. This construction provides direct access to differential-topological invariants of the data. In particular, we show that the first Stiefel-Whitney class can be computed from the signs of the Jacobians of learned transition maps, yielding an algorithmic criterion for detecting orientability. We also show that non-trivial characteristic classes provide obstructions to single-chart representations, and that the minimum number of autoencoder charts is determined by the good cover structure of the manifold. Finally, we apply our methodology to low-dimensional orientable and non-orientable manifolds, as well as to a non-orientable high-dimensional image dataset.

Reasoning with large language models often benefits from generating multiple chains-of-thought, but existing aggregation strategies are typically trajectory-level (e.g., selecting the best trace or voting on the final answer), discarding useful intermediate work from partial or "nearly correct" attempts. We propose Stitching Noisy Diffusion Thoughts, a self-consistency framework that turns cheap diffusion-sampled reasoning into a reusable pool of step-level candidates. Given a problem, we (i) sample many diverse, low-cost reasoning trajectories using a masked diffusion language model, (ii) score every intermediate step with an off-the-shelf process reward model (PRM), and (iii) stitch these highest-quality steps across trajectories into a composite rationale. This rationale then conditions an autoregressive (AR) model (solver) to recompute only the final answer. This modular pipeline separates exploration (diffusion) from evaluation and solution synthesis, avoiding monolithic unified hybrids while preserving broad search. Across math reasoning benchmarks, we find that step-level recombination is most beneficial on harder problems, and ablations highlight the importance of the final AR solver in converting stitched but imperfect rationales into accurate answers. Using low-confidence diffusion sampling with parallel, independent rollouts, our training-free framework improves average accuracy by up to 23.8% across six math and coding tasks. At the same time, it achieves up to a 1.8x latency reduction relative to both traditional diffusion models (e.g., Dream, LLaDA) and unified architectures (e.g., TiDAR). Code is available at https://github.com/roymiles/diffusion-stitching.

Diffusion Large Language Models (DLLMs) promise fast non-autoregressive inference but suffer a severe quality-speed trade-off in parallel decoding. This stems from the ''combinatorial contradiction'' phenomenon, where parallel tokens form semantically inconsistent combinations. We address this by integrating continuous representations into the discrete decoding process, as they preserve rich inter-position dependency. We propose ReMix (Rejection Mixing), a framework that introduces a novel Continuous Mixing State as an intermediate between the initial masked state and the final decoded token state. This intermediate state allows a token's representation to be iteratively refined in a continuous space, resolving mutual conflicts with other tokens before collapsing into a final discrete sample. Furthermore, a rejection rule reverts uncertain representations from the continuous state back to the masked state for reprocessing, ensuring stability and preventing error propagation. ReMix thus mitigates combinatorial contradictions by enabling continuous-space refinement during discrete diffusion decoding. Extensive experiments demonstrate that ReMix, as a training-free method, achieves a $2-8 \times$ inference speedup without any quality degradation.

Explicit representations of predicate-argument relations form the basis of interpretable semantic analysis, supporting reasoning, generation, and evaluation. However, attaining such semantic structures requires costly annotation efforts and has remained largely confined to English. We leverage the Question-Answer driven Semantic Role Labeling (QA-SRL) framework -- a natural-language formulation of predicate-argument relations -- as the foundation for extending semantic annotation to new languages. To this end, we introduce a cross-linguistic projection approach that reuses an English QA-SRL parser within a constrained translation and word-alignment pipeline to automatically generate question-answer annotations aligned with target-language predicates. Applied to Hebrew, Russian, and French -- spanning diverse language families -- the method yields high-quality training data and fine-tuned, language-specific parsers that outperform strong multilingual LLM baselines (GPT-4o, LLaMA-Maverick). By leveraging QA-SRL as a transferable natural-language interface for semantics, our approach enables efficient and broadly accessible predicate-argument parsing across languages.

Modern multi-tenant, hardware-heterogeneous computing environments pose significant challenges for effective workload orchestration. Simple heuristics for assessing workload performance, such as CPU utilization or application-level metrics, are often insufficient to capture the complex performance dynamics arising from resource contention and noisy-neighbor effects. In such environments, performance bottlenecks may emerge in any shared system resource, leading to unexpected and difficult-to-diagnose degradation. This paper introduces buoyancy, a novel abstraction for characterizing workload performance in multi-tenant systems. Unlike traditional approaches, buoyancy integrates application-level metrics with system-level insights of shared resource contention to provide a holistic view of performance dynamics. By explicitly capturing bottlenecks and headroom across multiple resources, buoyancy facilitates resource-aware and application-aware orchestration in a manner that is intuitive, extensible, and generalizable across heterogeneous platforms. We evaluate buoyancy using representative multi-tenant workloads to illustrate its ability to expose performance-limiting resource interactions. Buoyancy provides a 19.3% better indication of bottlenecks compared to traditional heuristics on average. We additionally show how buoyancy can act as a drop-in replacement for conventional performance metrics, enabling improved observability and more informed scheduling and optimization decisions.

Accurate computational identification of DNA methylation is essential for understanding epigenetic regulation. Although deep learning excels in this binary classification task, its "black-box" nature impedes biological insight. We address this by introducing a high-performance model MEDNA-DFM, alongside mechanism-inspired signal purification algorithms. Our investigation demonstrates that MEDNA-DFM effectively captures conserved methylation patterns, achieving robust distinction across diverse species. Validation on external independent datasets confirms that the model's generalization is driven by conserved intrinsic motifs (e.g., GC content) rather than phylogenetic proximity. Furthermore, applying our developed algorithms extracted motifs with significantly higher reliability than prior studies. Finally, empirical evidence from a Drosophila 6mA case study prompted us to propose a "sequence-structure synergy" hypothesis, suggesting that the GAGG core motif and an upstream A-tract element function cooperatively. We further validated this hypothesis via in silico mutagenesis, confirming that the ablation of either or both elements significantly degrades the model's recognition capabilities. This work provides a powerful tool for methylation prediction and demonstrates how explainable deep learning can drive both methodological innovation and the generation of biological hypotheses.

The concept of ranking aggregation plays a central role in preference analysis, and numerous algorithms for calculating median rankings, often originating in social choice theory, have been documented in the literature, offering theoretical guarantees in a centralized setting, i.e., when all the ranking data to be aggregated can be brought together in a single computing unit. For many technologies (e.g. peer-to-peer networks, IoT, multi-agent systems), extending the ability to calculate consensus rankings with guarantees in a decentralized setting, i.e., when preference data is initially distributed across a communicating network, remains a major methodological challenge. Indeed, in recent years, the literature on decentralized computation has mainly focused on computing or optimizing statistics such as arithmetic means using gossip algorithms. The purpose of this article is precisely to study how to achieve reliable consensus on collective rankings using classical rules (e.g. Borda, Copeland) in a decentralized setting, thereby raising new questions, robustness to corrupted nodes, and scalability through reduced communication costs in particular. The approach proposed and analyzed here relies on random gossip communication, allowing autonomous agents to compute global ranking consensus using only local interactions, without coordination or central authority. We provide rigorous convergence guarantees, including explicit rate bounds, for the Borda and Copeland consensus methods. Beyond these rules, we also provide a decentralized implementation of consensus according to the median rank rule and local Kemenization. Extensive empirical evaluations on various network topologies and real and synthetic ranking datasets demonstrate that our algorithms converge quickly and reliably to the correct ranking aggregation.

Argumentation mining comprises several subtasks, among which stance classification focuses on identifying the standpoint expressed in an argumentative text toward a specific target topic. While arguments-especially about controversial topics-often appeal to emotions, most prior work has not systematically incorporated explicit, fine-grained emotion analysis to improve performance on this task. In particular, prior research on stance classification has predominantly utilized non-argumentative texts and has been restricted to specific domains or topics, limiting generalizability. We work on five datasets from diverse domains encompassing a range of controversial topics and present an approach for expanding the Bias-Corrected NRC Emotion Lexicon using DistilBERT embeddings, which we feed into a Neural Argumentative Stance Classification model. Our method systematically expands the emotion lexicon through contextualized embeddings to identify emotionally charged terms not previously captured in the lexicon. Our expanded NRC lexicon (eNRC) improves over the baseline across all five datasets (up to +6.2 percentage points in F1 score), outperforms the original NRC on four datasets (up to +3.0), and surpasses the LLM-based approach on nearly all corpora. We provide all resources-including eNRC, the adapted corpora, and model architecture-to enable other researchers to build upon our work.

Can artificial intelligence truly contribute to creative mathematical research, or does it merely automate routine calculations while introducing risks of error? We provide empirical evidence through a detailed case study: the discovery of novel error representations and bounds for Hermite quadrature rules via systematic human-AI collaboration. Working with multiple AI assistants, we extended results beyond what manual work achieved, formulating and proving several theorems with AI assistance. The collaboration revealed both remarkable capabilities and critical limitations. AI excelled at algebraic manipulation, systematic proof exploration, literature synthesis, and LaTeX preparation. However, every step required rigorous human verification, mathematical intuition for problem formulation, and strategic direction. We document the complete research workflow with unusual transparency, revealing patterns in successful human-AI mathematical collaboration and identifying failure modes researchers must anticipate. Our experience suggests that, when used with appropriate skepticism and verification protocols, AI tools can meaningfully accelerate mathematical discovery while demanding careful human oversight and deep domain expertise.

Presentation generation requires deep content research, coherent visual design, and iterative refinement based on observation. However, existing presentation agents often rely on predefined workflows and fixed templates. To address this, we present DeepPresenter, an agentic framework that adapts to diverse user intents, enables effective feedback-driven refinement, and generalizes beyond a scripted pipeline. Specifically, DeepPresenter autonomously plans, renders, and revises intermediate slide artifacts to support long-horizon refinement with environmental observations. Furthermore, rather than relying on self-reflection over internal signals (e.g., reasoning traces), our environment-grounded reflection conditions the generation process on perceptual artifact states (e.g., rendered slides), enabling the system to identify and correct presentation-specific issues during execution. Results on the evaluation set covering diverse presentation-generation scenarios show that DeepPresenter achieves state-of-the-art performance, and the fine-tuned 9B model remains highly competitive at substantially lower cost. Our project is available at: https://github.com/icip-cas/PPTAgent

Hybrid work has become a reality post-pandemic, transforming how Agile teams deliver value, collaborate, and adapt. This study investigate how hybrid settings influence productivity and collaboration through nine interviews with three Norwegian Agile teams. Our findings show that hybrid work reduces informal interaction, creates uneven participation, and increases reliance on digital tools. Agile ceremonies became alignment anchors, while trust, communication, and tool support mediate team effectiveness. Hybrid Agile work is an evolving field that requires tailored structures to support inclusion, team cohesion, and sustainable performance.

Moral benchmarks for LLMs typically use context-free prompts, implicitly assuming stable preferences. In deployment, however, prompts routinely include contextual signals such as user requests, cues on social norms, etc. that may steer decisions. We study how directed contextual influences reshape decisions in trolley-problem-style moral triage settings. We introduce a pilot evaluation harness for directed contextual influence in trolley-problem-style moral triage: for each demographic factor, we apply matched, direction-flipped contextual influences that differ only in which group they favor, enabling systematic measurement of directional response. We find that: (i) contextual influences often significantly shift decisions, even when only superficially relevant; (ii) baseline preferences are a poor predictor of directional steerability, as models can appear baseline-neutral yet exhibit systematic steerability asymmetry under influence; (iii) influences can backfire: models may explicitly claim neutrality or discount the contextual cue, yet their choices still shift, sometimes in the opposite direction; and (iv) reasoning reduces average sensitivity, but amplifies the effect of biased few-shot examples. Our findings motivate extending moral evaluations with controlled, direction-flipped context manipulations to better characterize model behavior.

Background: Retrieval augmented generation (RAG) technology can empower large language models (LLMs) to generate more accurate, professional, and timely responses without fine tuning. However, due to the complex reasoning processes and substantial individual differences involved in traditional Chinese medicine (TCM) clinical diagnosis and treatment, traditional RAG methods often exhibit poor performance in this domain. Objective: To address the limitations of conventional RAG approaches in TCM applications, this study aims to develop an improved RAG framework tailored to the characteristics of TCM reasoning. Methods: We developed TCM-DiffRAG, an innovative RAG framework that integrates knowledge graphs (KG) with chains of thought (CoT). TCM-DiffRAG was evaluated on three distinctive TCM test datasets. Results: The experimental results demonstrated that TCM-DiffRAG achieved significant performance improvements over native LLMs. For example, the qwen-plus model achieved scores of 0.927, 0.361, and 0.038, which were significantly enhanced to 0.952, 0.788, and 0.356 with TCM-DiffRAG. The improvements were even more pronounced for non-Chinese LLMs. Additionally, TCM-DiffRAG outperformed directly supervised fine-tuned (SFT) LLMs and other benchmark RAG methods. Conclusions: TCM-DiffRAG shows that integrating structured TCM knowledge graphs with Chain of Thought based reasoning substantially improves performance in individualized diagnostic tasks. The joint use of universal and personalized knowledge graphs enables effective alignment between general knowledge and clinical reasoning. These results highlight the potential of reasoning-aware RAG frameworks for advancing LLM applications in traditional Chinese medicine.

This paper presents a comprehensive evaluation framework for assessing the cultural competence of large language models (LLMs) in Persian. Existing Persian cultural benchmarks rely predominantly on multiple-choice formats and English-centric metrics that fail to capture Persian's morphological complexity and semantic nuance. Our framework introduces a Persian-specific short-answer evaluation that combines rule-based morphological normalization with a hybrid syntactic and semantic similarity module, enabling robust soft-match scoring beyond exact string overlap. Through systematic evaluation of 15 state-of-the-art open- and closed-source models, we demonstrate that our hybrid evaluation improves scoring consistency by +10% compared to exact-match baselines by capturing meaning that surface-level methods cannot detect. We publicly release our evaluation framework, providing the first standardized benchmark for measuring cultural understanding in Persian and establishing a reproducible foundation for cross-cultural LLM evaluation research.

Functional data clustering is concerned with grouping functions that share similar structure, yet most existing methods implicitly operate on sampled grids, causing cluster assignments to depend on resolution, sampling density, or preprocessing choices rather than on the underlying functions themselves. To address this limitation, we introduce a framework that maps discretized function observations -- at arbitrary resolution and on arbitrary grids -- into a fixed-dimensional vector space via an auto-encoding architecture. The encoder is a hypernetwork that maps coordinate-value pairs to the weight space of an implicit neural representation (INR), which serves as the decoder. Because INRs represent functions with very few parameters, this design yields compact representations that are decoupled from the sampling grid, while the hypernetwork amortizes weight prediction across the dataset. Clustering is then performed in this weight space using standard algorithms, making the approach agnostic to both the discretization and the choice of clustering method. By means of synthetic and real-world experiments in high-dimensional settings, we demonstrate competitive clustering performance that is robust to changes in sampling resolution -- including generalization to resolutions not seen during training.

The identification and property prediction of chemical molecules is of central importance in the advancement of drug discovery and material science, where the tandem mass spectrometry technology gives valuable fragmentation cues in the form of mass-to-charge ratio peaks. However, the lack of experimental spectra hinders the attachment of each molecular identification, and thus urges the establishment of prediction approaches for computational models. Deep learning models appear promising for predicting molecular structure spectra, but overall assessment remains challenging as a result of the heterogeneity in methods and the lack of well-defined benchmarks. To address this, our contribution is the creation of benchmark framework FlexMS for constructing and evaluating diverse model architectures in mass spectrum prediction. With its easy-to-use flexibility, FlexMS supports the dynamic construction of numerous distinct combinations of model architectures, while assessing their performance on preprocessed public datasets using different metrics. In this paper, we provide insights into factors influencing performance, including the structural diversity of datasets, hyperparameters like learning rate and data sparsity, pretraining effects, metadata ablation settings and cross-domain transfer learning analysis. This provides practical guidance in choosing suitable models. Moreover, retrieval benchmarks simulate practical identification scenarios and score potential matches based on predicted spectra.

Group-based reinforcement learning (RL), such as GRPO, has advanced the capabilities of large language models on long-horizon agentic tasks. To enable more fine-grained policy updates, recent research has increasingly shifted toward stepwise group-based policy optimization, which treats each step in a rollout trajectory independently while using a memory module to retain historical context. However, we find a key issue in estimating stepwise relative advantages, namely context inconsistency, where steps within the same group may differ in their historical contexts. Empirically, we reveal that this issue can lead to severely biased advantage estimation, thereby degrading policy optimization significantly. To address the issue, in this paper, we propose Hierarchy-of-Groups Policy Optimization (HGPO) for long-horizon agentic tasks. Specifically, within a group of rollout trajectories, HGPO assigns each step to multiple hierarchical groups according to the consistency of historical contexts. Then, for each step, HGPO computes distinct advantages within each group and aggregates them with an adaptive weighting scheme. In this way, HGPO can achieve a favorable bias-variance trade-off in stepwise advantage estimation, without extra models or rollouts. Evaluations on two challenging agentic tasks, ALFWorld and WebShop with Qwen2.5-1.5B-Instruct and Qwen2.5-7B-Instruct, show that HGPO significantly outperforms existing agentic RL methods under the same computational constraints. Code is available at https://github.com/langfengQ/verl-agent/tree/master/recipe/hgpo.

Agentic AI increasingly intervenes proactively by inferring users' situations from contextual data yet often fails for lack of principled judgment about when, why, and whether to act. We address this gap by proposing a conceptual model that reframes behavior as an interpretive outcome integrating Scene (observable situation), Context (user-constructed meaning), and Human Behavior Factors (determinants shaping behavioral likelihood). Grounded in multidisciplinary perspectives across the humanities, social sciences, HCI, and engineering, the model separates what is observable from what is meaningful to the user and explains how the same scene can yield different behavioral meanings and outcomes. To translate this lens into design action, we derive five agent design principles (behavioral alignment, contextual sensitivity, temporal appropriateness, motivational calibration, and agency preservation) that guide intervention depth, timing, intensity, and restraint. Together, the model and principles provide a foundation for designing agentic AI systems that act with contextual sensitivity and judgment in interactions.

Since local LLM inference on resource-constrained edge devices imposes a severe performance bottleneck, this paper proposes distributed prompt caching to enhance inference performance by cooperatively sharing intermediate processing states across multiple low-end edge devices. To fully utilize prompt similarity, our distributed caching mechanism also supports partial matching. As this approach introduces communication overhead associated with state sharing over a wireless network, we introduce a Bloom-filter-based data structure, referred to as a catalog, to determine whether a remote server possesses the desired internal states, thereby suppressing unnecessary communication. Experiments using the Gemma-3 270M model and the MMLU dataset on the Raspberry Pi Zero 2W platform demonstrate that the proposed approach reduces TTFT (Time to First Token) and TTLT (Time to Last Token) by 93.12% and 50.07% on average, respectively.

In this work, we present the first theoretical analysis of multi-agent imitation learning (MAIL) in linear Markov games where both the transition dynamics and each agent's reward function are linear in some given features. We demonstrate that by leveraging this structure, it is possible to replace the state-action level "all policy deviation concentrability coefficient" (Freihaut et al., arXiv:2510.09325) with a concentrability coefficient defined at the feature level which can be much smaller than the state-action analog when the features are informative about states' similarity. Furthermore, to circumvent the need for any concentrability coefficient, we turn to the interactive setting. We provide the first, computationally efficient, interactive MAIL algorithm for linear Markov games and show that its sample complexity depends only on the dimension of the feature map $d$. Building on these theoretical findings, we propose a deep MAIL interactive algorithm which clearly outperforms BC on games such as Tic-Tac-Toe and Connect4.

Despite the remarkable progress of large language models (LLMs), the capabilities of standalone LLMs have begun to plateau when tackling real-world, complex tasks that require interaction with external tools and dynamic environments. Although recent agent frameworks aim to enhance model autonomy through tool integration and external interaction, they still suffer from naive workflows, unstable performance, limited support across diverse benchmarks and tasks, and heavy reliance on costly commercial APIs. In this work, we propose a high-performance and robust open-source agent framework, termed MiroFlow, which incorporates an agent graph for flexible orchestration, an optional deep reasoning mode to enhance performance, and a robust workflow execution to ensure stable and reproducible performance. Extensive experiments demonstrate that MiroFlow consistently achieves state-of-the-art performance across multiple agent benchmarks, including GAIA, BrowseComp-EN/ZH, HLE, xBench-DeepSearch, and notably FutureX. We hope it could serve as an easily accessible, reproducible, and comparable baseline for the deep research community.

In this paper, we propose a multi-mutation optimization algorithm, Differential Evolution with Multi-Mutation Operator-Guided Communication (DE-MMOGC), implemented to improve the performance and convergence abilities of standard differential evolution in uncertain environments. DE-MMOGC introduces a communication-guided scheme integrated with multiple mutation operators to encourage exploration and avoid premature convergence. Along with this, it includes a dynamic operator selection mechanism to use the best-performing operator over successive generations. To assimilate real-world uncertainties and missing observations into the predictive model, the proposed algorithm is combined with the Ensemble Kalman Filter. To evaluate the efficacy of the proposed DE-MMOGC in uncertain systems, the unified framework is applied to improve the predictive accuracy of crop simulation models. These simulation models are essential to precision agriculture, as they make it easier to estimate crop growth in a variety of unpredictable weather scenarios. Additionally, precisely calibrating these models raises a challenge due to missing observations. Hence, the simplified WOFOST crop simulation model is incorporated in this study for leaf area index (LAI)-based crop yield estimation. DE-MMOGC enhances the WOFOST performance by optimizing crucial weather parameters (temperature and rainfall), since these parameters are highly uncertain across different crop varieties, such as wheat, rice, and cotton. The experimental study shows that DE-MMOGC outperforms the traditional evolutionary optimizers and achieves better correlation with real LAI values. We found that DE-MMOGC is a resilient solution for crop monitoring.

Diffusion models have become a popular choice for decision-making tasks in robotics, and more recently, are also being considered for solving autonomous driving tasks. However, their applications and evaluations in autonomous driving remain limited to simulation-based or laboratory settings. The full strength of diffusion models for large-scale, complex real-world settings, such as End-to-End Autonomous Driving (E2E AD), remains underexplored. In this study, we conducted a systematic and large-scale investigation to unleash the potential of the diffusion models as planners for E2E AD, based on a tremendous amount of real-vehicle data and road testing. Through comprehensive and carefully controlled studies, we identify key insights into the diffusion loss space, trajectory representation, and data scaling that significantly impact E2E planning performance. Moreover, we also provide an effective reinforcement learning post-training strategy to further enhance the safety of the learned planner. The resulting diffusion-based learning framework, Hyper Diffusion Planner} (HDP), is deployed on a real-vehicle platform and evaluated across 6 urban driving scenarios and 200 km of real-world testing, achieving a notable 10x performance improvement over the base model. Our work demonstrates that diffusion models, when properly designed and trained, can serve as effective and scalable E2E AD planners for complex, real-world autonomous driving tasks.

Air-based molecular communication (MC) has the potential to be one of the first MC systems to be deployed in real-world applications, enabled by commercially available sensors. However, these sensors usually exhibit non-linear and cross-reactive behavior, contrary to the idealizing assumption of linear and perfectly molecule type-specific sensing often made in the MC literature. To address this mismatch, we propose several detectors and transmission schemes for a molecule mixture communication system where the receiver (RX) employs non-linear, cross-reactive sensors. All proposed schemes are based on the first- and second-order moments of the symbol likelihoods that are fed through the non-linear RX using the Unscented Transform. In particular, we propose an approximate maximum likelihood (AML) symbol-by-symbol detector for inter-symbol-interference (ISI)-free transmission scenarios and a complementary mixture alphabet design algorithm which accounts for the RX characteristics. When significant ISI is present at high data rates, the AML detector can be adapted to exploit statistical ISI knowledge. Additionally, we propose a sequence detector which combines information from multiple symbol intervals. For settings where sequence detection is not possible due to extremely limited computational power at the RX, we propose an adaptive transmission scheme which can be combined with symbol-by-symbol detection. Using computer simulations, we validate all proposed detectors and algorithms based on the responses of commercially available sensors as well as artificially generated sensor data incorporating the characteristics of metal-oxide semiconductor sensors. By employing a general system model that accounts for transmitter noise, ISI, and general non-linear, cross-reactive RX arrays, this work enables reliable communication for a large class of MC systems.

Internet of Things (IoT) networks face significant challenges such as limited communication bandwidth, constrained computational and energy resources, and highly dynamic wireless channel conditions. Utilization of deep neural networks (DNNs) combined with semantic communication has emerged as a promising paradigm to address these limitations. Deep joint source-channel coding (DJSCC) has recently been proposed to enable semantic communication of images. Building upon the original DJSCC formulation, low-complexity attention-style architectures has been added to the DNNs for further performance enhancement. As a main hurdle, training these DNNs separately for various signal-to-noise ratios (SNRs) will amount to excessive storage or communication overhead, which can not be maintained by small IoT devices. SNR Adaptive DJSCC (ADJSCC), has been proposed to train the DNNs once but feed the current SNR as part of the data to the channel-wise attention mechanism. We improve upon ADJSCC by a simultaneous utilization of doubly adaptive channel-wise and spatial attention modules at both transmitter and receiver. These modules dynamically adjust to varying channel conditions and spatial feature importance, enabling robust and efficient feature extraction and semantic information recovery. Simulation results corroborate that our proposed doubly adaptive DJSCC (DA-DJSCC) significantly improves upon ADJSCC in several performance criteria, while incurring a mild increase in complexity. These facts render DA-DJSCC a desirable choice for semantic communication in performance demanding but low-complexity IoT networks.

The rapid evolution of large language models (LLMs) has transformed prompt engineering from a localized craft into a systems-level governance challenge. As models scale and update across generations, prompt behavior becomes sensitive to shifts in instruction-following policies, alignment regimes, and decoding strategies, a phenomenon we characterize as GPT-scale model drift. Under such conditions, surface-level formatting conventions and ad hoc refinement are insufficient to ensure stable, interpretable control. This paper reconceptualizes Natural Language Declarative Prompting (NLD-P) as a declarative governance method rather than a rigid field template. NLD-P is formalized as a modular control abstraction that separates provenance, constraint logic, task content, and post-generation evaluation, encoded directly in natural language without reliance on external orchestration code. We define minimal compliance criteria, analyze model-dependent schema receptivity, and position NLD-P as an accessible governance framework for non-developer practitioners operating within evolving LLM ecosystems. Portions of drafting and editorial refinement employed a schema-bound LLM assistant configured under NLD-P. All conceptual framing, methodological claims, and final revisions were directed, reviewed, and approved by the human author under a documented human-in-the-loop protocol. The paper concludes by outlining implications for declarative control under ongoing model evolution and identifying directions for future empirical validation.

Large language models (LLMs) often generate fluent but unfounded claims, or hallucinations, which fall into two types: (i) faithfulness violations - misusing user context - and (ii) factuality violations - errors from internal knowledge. Proper mitigation depends on knowing whether a model's answer is based on the prompt or its internal weights. This work focuses on the problem of contributive attribution: identifying the dominant knowledge source behind each output. We show that a probe, a simple linear classifier trained on model hidden representations, can reliably predict contributive attribution. For its training, we introduce AttriWiki, a self-supervised data pipeline that prompts models to recall withheld entities from memory or read them from context, generating labelled examples automatically. Probes trained on AttriWiki data reveal a strong attribution signal, achieving up to 0.96 Macro-F1 on Llama-3.1-8B, Mistral-7B, and Qwen-7B, transferring to out-of-domain benchmarks (SQuAD, WebQuestions) with 0.94-0.99 Macro-F1 without retraining. Attribution mismatches raise error rates by up to 70%, demonstrating a direct link between knowledge source confusion and unfaithful answers. Yet, models may still respond incorrectly even when attribution is correct, highlighting the need for broader detection frameworks.

Value decomposition (VD) methods have achieved remarkable success in cooperative multi-agent reinforcement learning (MARL). However, their reliance on the max operator for temporal-difference (TD) target calculation leads to systematic Q-value overestimation. This issue is particularly severe in MARL due to the combinatorial explosion of the joint action space, which often results in unstable learning and suboptimal policies. To address this problem, we propose QSIM, a similarity weighted Q-learning framework that reconstructs the TD target using action similarity. Instead of using the greedy joint action directly, QSIM forms a similarity weighted expectation over a structured near-greedy joint action space. This formulation allows the target to integrate Q-values from diverse yet behaviorally related actions while assigning greater influence to those that are more similar to the greedy choice. By smoothing the target with structurally relevant alternatives, QSIM effectively mitigates overestimation and improves learning stability. Extensive experiments demonstrate that QSIM can be seamlessly integrated with various VD methods, consistently yielding superior performance and stability compared to the original algorithms. Furthermore, empirical analysis confirms that QSIM significantly mitigates the systematic value overestimation in MARL. Code is available at https://github.com/MaoMaoLYJ/pymarl-qsim.

This study targets cloud native environments where microservice invocation relations are complex, load fluctuations are multi-scale and superimposed, and cross-service impacts are significant. We propose a structured temporal joint load prediction framework oriented to microservice topology. The method represents the system as a coupled entity of a time-evolving service invocation graph and multivariate load sequences. It constructs neighborhood-aggregated and global summarized views based on service level observations. This forms layered load representations across instance, service, and cluster levels. A unified sequence encoder models multi-scale historical context. To strengthen the expression of invocation dependencies, the framework introduces a lightweight structural prior into attention computation. This enables more effective capture of load propagation and accumulation along invocation chains, while maintaining consistent modeling of local bursts and overall trends. The training objective adopts a multi-objective regression strategy that jointly optimizes service level and cluster level predictions to improve cross-granularity stability. We further conduct single-factor sensitivity analyses on key structural and training hyperparameters. We systematically examine the effects of time window length, encoding depth, and regularization strength. The results support the necessity of multi-granularity fusion and structural injection and clarify their effective configuration ranges. Overall, the framework provides a reusable modeling paradigm and implementation path for capacity assessment, resource orchestration, and runtime situational understanding in cloud environments.

Predictive modeling on web-scale tabular data with billions of instances and hundreds of heterogeneous numerical features faces significant scalability challenges. These features exhibit anisotropy, heavy-tailed distributions, and non-stationarity, creating bottlenecks for models like Gradient Boosting Decision Trees and requiring laborious manual feature engineering. We introduce KMLP, a hybrid deep architecture integrating a shallow Kolmogorov-Arnold Network (KAN) front-end with a Gated Multilayer Perceptron (gMLP) backbone. The KAN front-end uses learnable activation functions to automatically model complex non-linear transformations for each feature, while the gMLP backbone captures high-order interactions. Experiments on public benchmarks and an industrial dataset with billions of samples show KMLP achieves state-of-the-art performance, with advantages over baselines like GBDTs increasing at larger scales, validating KMLP as a scalable deep learning paradigm for large-scale web tabular data.

As mental health chatbots proliferate to address the global treatment gap, a critical question emerges: How do we design for relational safety the quality of interaction patterns that unfold across conversations rather than the correctness of individual responses? Current safety evaluations assess single-turn crisis responses, missing the therapeutic dynamics that determine whether chatbots help or harm over time. We introduce TherapyProbe, a design probe methodology that generates actionable design knowledge by systematically exploring chatbot conversation trajectories through adversarial multi-agent simulation. Using open-source models, TherapyProbe surfaces relational safety failures interaction patterns like "validation spirals" where chatbots progressively reinforce hopelessness, or "empathy fatigue" where responses become mechanical over turns. Our contribution is translating these failures into a Safety Pattern Library of 23 failure archetypes with corresponding design recommendations. We contribute: (1) a replicable methodology requiring no API costs, (2) a clinically-grounded failure taxonomy, and (3) design implications for developers, clinicians, and policymakers.

Clinical decisions are often required under incomplete information. Clinical experts must identify whether available information is sufficient for judgment, as both premature conclusion and unnecessary abstention can compromise patient safety. To evaluate this capability of large language models (LLMs), we developed ClinDet-Bench, a benchmark based on clinical scoring systems that decomposes incomplete-information scenarios into determinable and undeterminable conditions. Identifying determinability requires considering all hypotheses about missing information, including unlikely ones, and verifying whether the conclusion holds across them. We find that recent LLMs fail to identify determinability under incomplete information, producing both premature judgments and excessive abstention, despite correctly explaining the underlying scoring knowledge and performing well under complete information. These findings suggest that existing benchmarks are insufficient to evaluate the safety of LLMs in clinical settings. ClinDet-Bench provides a framework for evaluating determinability recognition, leading to appropriate abstention, with potential applicability to medicine and other high-stakes domains, and is publicly available.

Large Language Models (LLMs) are deployed as autonomous agents in increasingly complex applications, where enabling long-horizon memory is critical for achieving strong performance. However, a significant gap exists between practical applications and current evaluation standards for agent memory: existing benchmarks primarily focus on dialogue-centric, human-agent interactions. In reality, agent memory consists of a continuous stream of agent-environment interactions that are primarily composed of machine-generated representations. To bridge this gap, we introduce AMA-Bench (Agent Memory with Any length), which evaluates long-horizon memory for LLMs in real agentic applications. It features two key components: (1) a set of real-world agentic trajectories across representative agentic applications, paired with expert-curated QA, and (2) a set of synthetic agentic trajectories that scale to arbitrary horizons, paired with rule-based QA. Our comprehensive study shows that existing memory systems underperform on AMA-Bench primarily because they lack causality and objective information and are constrained by the lossy nature of similarity-based retrieval employed by many memory systems. To address these limitations, we propose AMA-Agent, an effective memory system featuring a causality graph and tool-augmented retrieval. Our results demonstrate that AMA-Agent achieves 57.22% average accuracy on AMA-Bench, surpassing the strongest memory system baselines by 11.16%.

Latent visual reasoning aims to mimic human's imagination process by meditating through hidden states of Multimodal Large Language Models. While recognized as a promising paradigm for visual reasoning, the underlying mechanisms driving its effectiveness remain unclear. Motivated to demystify the true source of its efficacy, we investigate the validity of latent reasoning using Causal Mediation Analysis. We model the process as a causal chain: the input as the treatment, the latent tokens as the mediator, and the final answer as the outcome. Our findings uncover two critical disconnections: (a) Input-Latent Disconnect: dramatic perturbations on the input result in negligible changes to the latent tokens, suggesting that latent tokens do not effectively attend to the input sequence. (b) Latent-Answer Disconnect: perturbations on the latent tokens yield minimal impact on the final answer, indicating the limited causal effect latent tokens imposing on the outcome. Furthermore, extensive probing analysis reveals that latent tokens encode limited visual information and exhibit high similarity. Consequently, we challenge the necessity of latent reasoning and propose a straightforward alternative named CapImagine, which teaches the model to explicitly imagine using text. Experiments on vision-centric benchmarks show that CapImagine significantly outperforms complex latent-space baselines, highlighting the superior potential of visual reasoning through explicit imagination.

Reinforcement Learning (RL) has empowered Large Language Models (LLMs) with strong reasoning capabilities, but vanilla RL mainly focuses on generation capability improvement by training with only first-order rollout (generating multiple responses for a question), and we argue that this approach fails to fully exploit the potential of training data because of the neglect of critique capability training. To tackle this problem, we further introduce the concept of second-order rollout (generating multiple critiques for a response) and propose a unified framework for jointly training generation and critique capabilities. Extensive experiments across various models and datasets demonstrate that our approach can utilize training data more effectively than vanilla RL and achieve better performance under the same training data. Additionally, we uncover several insightful findings regarding second-order rollout and critique training, such as the importance of label balance in critique training and the noise problem of outcome-based rewards, which can be mitigated through sampling techniques. Our work offers a preliminary exploration of dynamic data augmentation and joint generation-critique training in RL, providing meaningful inspiration for the further advancement of RL training

The Rust programming language presents a steep learning curve and significant coding challenges, making the automation of issue resolution essential for its broader adoption. Recently, LLM-powered code agents have shown remarkable success in resolving complex software engineering tasks, yet their application to Rust has been limited by the absence of a large-scale, repository-level benchmark. To bridge this gap, we introduce Rust-SWE-bench, a benchmark comprising 500 real-world, repository-level software engineering tasks from 34 diverse and popular Rust repositories. We then perform a comprehensive study on Rust-SWE-bench with four representative agents and four state-of-the-art LLMs to establish a foundational understanding of their capabilities and limitations in the Rust ecosystem. Our extensive study reveals that while ReAct-style agents are promising, i.e., resolving up to 21.2% of issues, they are limited by two primary challenges: comprehending repository-wide code structure and complying with Rust's strict type and trait semantics. We also find that issue reproduction is rather critical for task resolution. Inspired by these findings, we propose RUSTFORGER, a novel agentic approach that integrates an automated test environment setup with a Rust metaprogramming-driven dynamic tracing strategy to facilitate reliable issue reproduction and dynamic analysis. The evaluation shows that RUSTFORGER using Claude-Sonnet-3.7 significantly outperforms all baselines, resolving 28.6% of tasks on Rust-SWE-bench, i.e., a 34.9% improvement over the strongest baseline, and, in aggregate, uniquely solves 46 tasks that no other agent could solve across all adopted advanced LLMs.

Training large language models (LLMs) requires substantial compute and energy. At the same time, renewable energy sources regularly produce more electricity than the grid can absorb, leading to curtailment, the deliberate reduction of clean generation that would otherwise go to waste. These periods represent an opportunity: if training is aligned with curtailment windows, LLMs can be pretrained using electricity that is both clean and cheap. This technical report presents a system that performs full-parameter LLM training across geo-distributed GPU clusters during regional curtailment windows, elastically switching between local single-site training and federated multi-site synchronization as sites become available or unavailable. Our prototype trains a 561M-parameter transformer model across three clusters using the Flower federated learning framework, with curtailment periods derived from real-world marginal carbon intensity traces. Preliminary results show that curtailment-aware scheduling preserves training quality while reducing operational emissions to 5-12% of single-site baselines.

We decompose physician disagreement in the HealthBench medical AI evaluation dataset to understand where variance resides and what observable features can explain it. Rubric identity accounts for 15.8% of met/not-met label variance but only 3.6-6.9% of disagreement variance; physician identity accounts for just 2.4%. The dominant 81.8% case-level residual is not reduced by HealthBench's metadata labels (z = -0.22, p = 0.83), normative rubric language (pseudo R^2 = 1.2%), medical specialty (0/300 Tukey pairs significant), surface-feature triage (AUC = 0.58), or embeddings (AUC = 0.485). Disagreement follows an inverted-U with completion quality (AUC = 0.689), confirming physicians agree on clearly good or bad outputs but split on borderline cases. Physician-validated uncertainty categories reveal that reducible uncertainty (missing context, ambiguous phrasing) more than doubles disagreement odds (OR = 2.55, p < 10^(-24)), while irreducible uncertainty (genuine medical ambiguity) has no effect (OR = 1.01, p = 0.90), though even the former explains only ~3% of total variance. The agreement ceiling in medical AI evaluation is thus largely structural, but the reducible/irreducible dissociation suggests that closing information gaps in evaluation scenarios could lower disagreement where inherent clinical ambiguity does not, pointing toward actionable evaluation design improvements.

All-to-all GPU communication is a critical bottleneck in large-scale training clusters, where completion time is constrained by per-port bandwidth and can be severely impacted by traffic skew across GPUs and network interface cards (NICs). This issue is amplified by the two-tier structure of modern GPU systems, which combine fast intra-server links with much slower inter-server networks. Motivated by recent system observations that highlight the importance of traffic reshaping and hierarchy awareness, we study all-to-all scheduling from an online switching and queueing-theoretic perspective. We propose a dynamic hierarchical Birkhoff--von Neumann (BvN) decomposition framework tailored to two-tier GPU fabrics. At each frame boundary, traffic is first balanced within each server using simple local operations to mitigate micro-level GPU/NIC skew while preserving aggregate server-to-server demand. A hierarchical BvN decomposition is then applied at the server level and refined into GPU-level matchings, significantly reducing decomposition complexity relative to a flat GPU-level approach. By integrating this construction with the dynamic frame sizing (DFS) principle, we obtain an online scheduler with provable stability under admissible Poisson arrivals. Simulations demonstrate substantial reductions in mean frame length, particularly under server-localized hotspot traffic.

We introduce AuditBench, an alignment auditing benchmark. AuditBench consists of 56 language models with implanted hidden behaviors. Each model has one of 14 concerning behaviors--such as sycophantic deference, opposition to AI regulation, or secret geopolitical loyalties--which it does not confess to when directly asked. AuditBench models are highly diverse--some are subtle, while others are overt, and we use varying training techniques both for implanting behaviors and training models not to confess. To demonstrate AuditBench's utility, we develop an investigator agent that autonomously employs a configurable set of auditing tools. By measuring investigator agent success using different tools, we can evaluate their efficacy. Notably, we observe a tool-to-agent gap, where tools that perform well in standalone non-agentic evaluations fail to translate into improved performance when used with our investigator agent. We find that our most effective tools involve scaffolded calls to auxiliary models that generate diverse prompts for the target. White-box interpretability tools can be helpful, but the agent performs best with black-box tools. We also find that audit success varies greatly across training techniques: models trained on synthetic documents are easier to audit than models trained on demonstrations, with better adversarial training further increasing auditing difficulty. We release our models, agent, and evaluation framework to support future quantitative, iterative science on alignment auditing.

The transition of Large Language Models (LLMs) from exploratory tools to active "silicon subjects" in social science lacks extensive validation of operational validity. This study introduces Conditioned Comment Prediction (CCP), a task in which a model predicts how a user would comment on a given stimulus by comparing generated outputs with authentic digital traces. This framework enables a rigorous evaluation of current LLM capabilities with respect to the simulation of social media user behavior. We evaluated open-weight 8B models (Llama3.1, Qwen3, Ministral) in English, German, and Luxembourgish language scenarios. By systematically comparing prompting strategies (explicit vs. implicit) and the impact of Supervised Fine-Tuning (SFT), we identify a critical form vs. content decoupling in low-resource settings: while SFT aligns the surface structure of the text output (length and syntax), it degrades semantic grounding. Furthermore, we demonstrate that explicit conditioning (generated biographies) becomes redundant under fine-tuning, as models successfully perform latent inference directly from behavioral histories. Our findings challenge current "naive prompting" paradigms and offer operational guidelines prioritizing authentic behavioral traces over descriptive personas for high-fidelity simulation.

Large reasoning models (LRMs) have emerged as a powerful paradigm for solving complex real-world tasks. In practice, these models are predominantly trained via Reinforcement Learning with Verifiable Rewards (RLVR), yet most existing outcome-only RLVR pipelines rely almost exclusively on a binary correctness signal and largely ignore the model's intrinsic uncertainty. We term this discrepancy the uncertainty-reward mismatch, under which high- and low-uncertainty solutions are treated equivalently, preventing the policy from "Know What You Know" and impeding the shift from optimizing for correct answers to optimizing effective reasoning paths. This limitation is especially critical in reasoning-centric tasks such as mathematics and question answering, where performance hinges on the quality of the model's internal reasoning process rather than mere memorization of final answers. To address this, we propose EGPO, a metacognitive entropy calibration framework that explicitly integrates intrinsic uncertainty into RLVR for enhancing LRMs. EGPO estimates per-sample uncertainty using a zero-overhead entropy proxy derived from token-level likelihoods and aligns it with extrinsic correctness through an asymmetric calibration mechanism that preserves correct reasoning while selectively regulating overconfident failures, thereby enabling stable and uncertainty-aware policy optimization. Moreover, EGPO recovers informative learning signals from otherwise degenerate group-based rollouts without modifying the verifier or reward definition. Extensive experiments across multiple benchmarks demonstrate that the proposed EGPO leads to substantial and consistent improvements in reasoning performance, establishing a principled path for advancing LRMs through metacognitive entropy calibration.

Epistemic uncertainty in neural networks is commonly modeled using two second-order paradigms: distribution-based representations, which rely on posterior parameter distributions, and set-based representations based on credal sets (convex sets of probability distributions). These frameworks are often regarded as fundamentally non-comparable due to differing semantics, assumptions, and evaluation practices, leaving their relative merits unclear. Empirical comparisons are further confounded by variations in the underlying predictive models. To clarify this issue, we present a controlled comparative study enabling principled, like-for-like evaluation of the two paradigms. Both representations are constructed from the same finite collection of predictive distributions generated by a shared neural network, isolating representational effects from predictive accuracy. Our study evaluates each representation through the lens of 3 uncertainty measures across 8 benchmarks, including selective prediction and out-of-distribution detection, spanning 6 underlying predictive models and 10 independent runs per configuration. Our results show that meaningful comparison between these seemingly non-comparable frameworks is both feasible and informative, providing insights into how second-order representation choices impact practical uncertainty-aware performance.

Recommendation model performance is intrinsically tied to the quality, volume, and relevance of their training data. To address common challenges like data sparsity and cold start, recent researchs have leveraged data from multiple auxiliary domains to enrich information within the target domain. However, inherent domain gaps can degrade the quality of mixed-domain data, leading to negative transfer and diminished model performance. Existing prevailing \emph{model-centric} paradigm -- which relies on complex, customized architectures -- struggles to capture the subtle, non-structural sequence dependencies across domains, leading to poor generalization and high demands on computational resources. To address these shortcomings, we propose \textsc{Taesar}, a \emph{data-centric} framework for \textbf{t}arget-\textbf{a}lign\textbf{e}d \textbf{s}equenti\textbf{a}l \textbf{r}egeneration, which employs a contrastive decoding mechanism to adaptively encode cross-domain context into target-domain sequences. It employs contrastive decoding to encode cross-domain context into target sequences, enabling standard models to learn intricate dependencies without complex fusion architectures. Experiments show \textsc{Taesar} outperforms model-centric solutions and generalizes to various sequential models. By generating enriched datasets, \textsc{Taesar} effectively combines the strengths of data- and model-centric paradigms. The code accompanying this paper is available at~ \textcolor{blue}{https://github.com/USTC-StarTeam/Taesar}.

Referring Image Segmentation (RIS) aims to segment an object in an image identified by a natural language expression. The paper introduces Alignment-Aware Masked Learning (AML), a training strategy to enhance RIS by explicitly estimating pixel-level vision-language alignment, filtering out poorly aligned regions during optimization, and focusing on trustworthy cues. This approach results in state-of-the-art performance on RefCOCO datasets and also enhances robustness to diverse descriptions and scenarios

Augmented reading systems aim to adapt text presentation to improve comprehension and task performance, yet existing approaches rely heavily on heuristics, opaque data-driven models, or repeated human involvement in the design loop. We propose framing augmented reading as a simulation-based optimization problem grounded in resource-rational models of human reading. These models instantiate a simulated reader that allocates limited cognitive resources, such as attention, memory, and time under task demands, enabling systematic evaluation of text user interfaces. We introduce two complementary optimization pipelines: an offline approach that explores design alternatives using simulated readers, and an online approach that personalizes reading interfaces in real time using ongoing interaction data. Together, this perspective enables adaptive, explainable, and scalable augmented reading design without relying solely on human testing.

Generative recommendation has recently attracted widespread attention in industry due to its potential for scaling and stronger model capacity. However, deploying real-time generative recommendation in large-scale advertising requires designs beyond large-language-model (LLM)-style training and serving recipes. We present a production-oriented generative recommender co-designed across architecture, learning, and serving, named GR4AD (Generative Recommendation for ADdvertising). As for tokenization, GR4AD proposes UA-SID (Unified Advertisement Semantic ID) to capture complicated business information. Furthermore, GR4AD introduces LazyAR, a lazy autoregressive decoder that relaxes layer-wise dependencies for short, multi-candidate generation, preserving effectiveness while reducing inference cost, which facilitates scaling under fixed serving budgets. To align optimization with business value, GR4AD employs VSL (Value-Aware Supervised Learning) and proposes RSPO (Ranking-Guided Softmax Preference Optimization), a ranking-aware, list-wise reinforcement learning algorithm that optimizes value-based rewards under list-level metrics for continual online updates. For online inference, we further propose dynamic beam serving, which adapts beam width across generation levels and online load to control compute. Large-scale online A/B tests show up to 4.2% ad revenue improvement over an existing DLRM-based stack, with consistent gains from both model scaling and inference-time scaling. GR4AD has been fully deployed in Kuaishou advertising system with over 400 million users and achieves high-throughput real-time serving.

This paper introduces a novel Czech dataset in the restaurant domain for aspect-based sentiment analysis (ABSA), enriched with annotations of opinion terms. The dataset supports three distinct ABSA tasks involving opinion terms, accommodating varying levels of complexity. Leveraging this dataset, we conduct extensive experiments using modern Transformer-based models, including large language models (LLMs), in monolingual, cross-lingual, and multilingual settings. To address cross-lingual challenges, we propose a translation and label alignment methodology leveraging LLMs, which yields consistent improvements. Our results highlight the strengths and limitations of state-of-the-art models, especially when handling the linguistic intricacies of low-resource languages like Czech. A detailed error analysis reveals key challenges, including the detection of subtle opinion terms and nuanced sentiment expressions. The dataset establishes a new benchmark for Czech ABSA, and our proposed translation-alignment approach offers a scalable solution for adapting ABSA resources to other low-resource languages.

Seed explosion is a fundamental problem in fuzzing seed scheduling, where a fuzzer maintains a huge corpus and fails to choose promising seeds. Existing works focus on seed prioritization but still suffer from seed explosion since corpus size remains huge. We tackle this from a new perspective: corpus reduction, i.e., computing a seed corpus subset. However, corpus reduction could lead to poor seed diversity and large runtime overhead. Prior techniques like cull_queue, AFL-Cmin, and MinSet suffer from poor diversity or prohibitive overhead, making them unsuitable for high-frequency seed scheduling. We propose RandSet, a novel randomized corpus reduction technique that reduces corpus size and yields diverse seed selection simultaneously with minimal overhead. Our key insight is introducing randomness into corpus reduction to enjoy two benefits of a randomized algorithm: randomized output (diverse seed selection) and low runtime cost. Specifically, we formulate corpus reduction as a set cover problem and compute a randomized subset covering all features of the entire corpus. We then schedule seeds from this small, randomized subset rather than the entire corpus, effectively mitigating seed explosion. We implement RandSet on three popular fuzzers: AFL++, LibAFL, and Centipede, and evaluate it on standalone programs, FuzzBench, and Magma. Results show RandSet achieves significantly more diverse seed selection than other reduction techniques, with average subset ratios of 4.03% and 5.99% on standalone and FuzzBench programs. RandSet achieves a 16.58% coverage gain on standalone programs and up to 3.57% on FuzzBench in AFL++, triggers up to 7 more ground-truth bugs than the state-of-the-art on Magma, while introducing only 1.17%-3.93% overhead.

Large language model (LLM) agents increasingly rely on external tools and retrieval systems to autonomously complete complex tasks. However, this design exposes agents to indirect prompt injection (IPI), where attacker-controlled context embedded in tool outputs or retrieved content silently steers agent actions away from user intent. Unlike prompt-based attacks, IPI unfolds over multi-turn trajectories, making malicious control difficult to disentangle from legitimate task execution. Existing inference-time defenses primarily rely on heuristic detection and conservative blocking of high-risk actions, which can prematurely terminate workflows or broadly suppress tool usage under ambiguous multi-turn scenarios. We propose AgentSentry, a novel inference-time detection and mitigation framework for tool-augmented LLM agents. To the best of our knowledge, AgentSentry is the first inference-time defense to model multi-turn IPI as a temporal causal takeover. It localizes takeover points via controlled counterfactual re-executions at tool-return boundaries and enables safe continuation through causally guided context purification that removes attack-induced deviations while preserving task-relevant evidence. We evaluate AgentSentry on the \textsc{AgentDojo} benchmark across four task suites, three IPI attack families, and multiple black-box LLMs. AgentSentry eliminates successful attacks and maintains strong utility under attack, achieving an average Utility Under Attack (UA) of 74.55 %, improving UA by 20.8 to 33.6 percentage points over the strongest baselines without degrading benign performance.

There is growing recognition that many NLP tasks lack a single ground truth, as human judgments reflect diverse perspectives. To capture this variation, models have been developed to predict full annotation distributions rather than majority labels, while perspectivist models aim to reproduce the interpretations of individual annotators. In this work, we compare these approaches on Implicit Discourse Relation Recognition (IDRR), a highly ambiguous task where disagreement often arises from cognitive complexity rather than ideological bias. Our experiments show that existing annotator-specific models perform poorly in IDRR unless ambiguity is reduced, whereas models trained on label distributions yield more stable predictions. Further analysis indicates that frequent cognitively demanding cases drive inconsistency in human interpretation, posing challenges for perspectivist modeling in IDRR.

Table Question Answering (TQA) aims to answer natural language questions over structured tables. Large Language Models (LLMs) enable promising solutions to this problem, with operator-centric solutions that generate table manipulation pipelines in a multi-step manner offering state-of-the-art performance. However, these solutions rely on multiple LLM calls, resulting in prohibitive latencies and computational costs. We propose Operation-R1, the first framework that trains lightweight LLMs (e.g., Qwen-4B/1.7B) via a novel variant of reinforcement learning with verifiable rewards to produce high-quality data-preparation pipelines for TQA in a single inference step. To train such an LLM, we first introduce a self-supervised rewarding mechanism to automatically obtain fine-grained pipeline-wise supervision signals for LLM training. We also propose variance-aware group resampling to mitigate training instability. To further enhance robustness of pipeline generation, we develop two complementary mechanisms: operation merge, which filters spurious operations through multi-candidate consensus, and adaptive rollback, which offers runtime protection against information loss in data transformation. Experiments on two benchmark datasets show that, with the same LLM backbone, Operation-R1 achieves average absolute accuracy gains of 9.55 and 6.08 percentage points over multi-step preparation baselines, with 79\% table compression and a 2.2$\times$ reduction in monetary cost.

State-space models (SSMs) have emerged as an efficient strategy for building powerful language models, avoiding the quadratic complexity of computing attention in transformers. Despite their promise, the interpretability and steerability of modern SSMs remain relatively underexplored. We take a major step in this direction by identifying activation subspace bottlenecks in the Mamba family of SSM models using tools from mechanistic interpretability. We then introduce a test-time steering intervention that simply multiplies the activations of the identified bottlenecks by a scalar. Across 5 SSMs and 6 diverse benchmarks, this intervention improves performance by an average of 8.27%, without requiring any task-specific tuning. Finally, we validate that the identified bottlenecks are indeed hindering performance by modifying them to yield an architecture we call Stable-Mamba, which achieves long-context performance gains when retrained from scratch.

Reinforcement Learning from Human Feedback (RLHF) has been widely applied to Large Language Model (LLM) post-training to align model outputs with human preferences. Recent models, such as DeepSeek-R1, have also shown RLHF's potential to improve LLM reasoning on complex tasks. In RL, inference and training co-exist, creating dynamic resource demands throughout the workflow. Compared to traditional RL, RLHF further challenges training efficiency due to expanding model sizes and resource consumption. Several RLHF frameworks aim to balance flexible abstraction and efficient execution. However, they rely on serverful infrastructures, which struggle with fine-grained resource variability. As a result, during synchronous RLHF training, idle time between or within RL components often causes overhead and resource wastage. To address these issues, we present RLHFless, the first scalable training framework for synchronous RLHF, built on serverless computing environments. RLHFless adapts to dynamic resource demands throughout the RLHF pipeline, pre-computes shared prefixes to avoid repeated computation, and uses a cost-aware actor scaling strategy that accounts for response length variation to find sweet spots with lower cost and higher speed. In addition, RLHFless assigns workloads efficiently to reduce intra-function imbalance and idle time. Experiments on both physical testbeds and a large-scale simulated cluster show that RLHFless achieves up to 1.35x speedup and 44.8% cost reduction compared to the state-of-the-art baseline.

3D Large Vision-Language Models (3D LVLMs) built upon Large Language Models (LLMs) have achieved remarkable progress across various multimodal tasks. However, their inherited position-dependent modeling mechanism, Rotary Position Embedding (RoPE), remains suboptimal for 3D multimodal understanding. The vanilla RoPE formulation fails to preserve essential three-dimensional spatial structures when encoding 3D tokens, and its relative distance computation overlooks angular dependencies, hindering the model's ability to capture directional variations in visual representations. To overcome these limitations, we introduce Spherical Coordinate-based Positional Embedding (SoPE). Our method maps point-cloud token indices into a 3D spherical coordinate space, enabling unified modeling of spatial locations and directional angles. This formulation preserves the inherent geometric structure of point-cloud data, enhances spatial awareness, and yields more consistent and expressive geometric representations for multimodal learning. In addition, we introduce a multi-scale frequency mixing strategy to fuse feature information across different frequency domains. Experimental results on multiple 3D scene benchmarks validate the effectiveness of our approach, while real-world deployment experiments further demonstrate its strong generalization capability.

Preference-based reinforcement learning (PbRL) is the dominant framework for aligning AI systems to human preferences, but its application to speech remains underexplored. We present a controlled cross-modal study of human and synthetic preference annotations, comparing text and audio evaluations of identical semantic content across 100 prompts. Audio preferences prove as reliable as text, with inter-rater agreement reaching good levels (ICC(2,k) $\approx$ .80) at $\sim$9 raters -- the first ICC-based reliability characterization in the preference annotation literature for either modality. However, modality reshapes how people judge: audio raters exhibit narrower decision thresholds, reduced length bias, and more user-oriented evaluation criteria, with near-chance cross-modality agreement. Synthetic ratings further align with human judgments and predict inter-rater agreement, supporting their use both for triaging ambiguous pairs and as full replacements for human annotations.

Vision-language models (VLMs) often struggle with geometric reasoning due to their limited perception of fundamental diagram elements. To tackle this challenge, we introduce GeoPerceive, a benchmark comprising diagram instances paired with domain-specific language (DSL) representations, along with an efficient automatic data generation pipeline. This design enables the isolated evaluation of geometric perception independently from reasoning. To exploit the data provided by GeoPerceive for enhancing the geometric perception capabilities of VLMs, we propose GeoDPO, a translator-guided reinforcement learning (RL) framework. GeoDPO employs an NL-to-DSL translator, which is trained on synthetic pairs generated by the data engine of GeoPerceive, to bridge natural language and DSL. This translator facilitates the computation of fine-grained, DSL-level scores, which serve as reward signals in reinforcement learning. We assess GeoDPO on both in-domain and out-of-domain datasets, spanning tasks in geometric perception as well as downstream reasoning. Experimental results demonstrate that, while supervised fine-tuning (SFT) offers only marginal improvements and may even impair performance in out-of-domain scenarios, GeoDPO achieves substantial gains: $+26.5\%$ on in-domain data, $+8.0\%$ on out-of-domain data, and $+39.0\%$ on downstream reasoning tasks. These findings underscore the superior performance and generalization ability of GeoDPO over SFT. All codes are released at https://github.com/Longin-Yu/GeoPerceive to ensure reproducibility.

Existing neural network calibration methods often treat calibration as a static, post-hoc optimization task. However, this neglects the dynamic and temporal nature of real-world inference. Moreover, existing methods do not provide an intuitive interface enabling human operators to dynamically adjust model behavior under shifting conditions. In this work, we propose Knob, a framework that connects deep learning with classical control theory by mapping neural gating dynamics to a second-order mechanical system. By establishing correspondences between physical parameters -- damping ratio ($ζ$) and natural frequency ($ω_n$) -- and neural gating, we create a tunable "safety valve". The core mechanism employs a logit-level convex fusion, functioning as an input-adaptive temperature scaling. It tends to reduce model confidence particularly when model branches produce conflicting predictions. Furthermore, by imposing second-order dynamics (Knob-ODE), we enable a \textit{dual-mode} inference: standard i.i.d. processing for static tasks, and state-preserving processing for continuous streams. Our framework allows operators to tune "stability" and "sensitivity" through familiar physical analogues. This paper presents an exploratory architectural interface; we focus on demonstrating the concept and validating its control-theoretic properties rather than claiming state-of-the-art calibration performance. Experiments on CIFAR-10-C validate the calibration mechanism and demonstrate that, in Continuous Mode, the gate responses are consistent with standard second-order control signatures (step settling and low-pass attenuation), paving the way for predictable human-in-the-loop tuning.

Commercial large language models are typically deployed as black-box API services, requiring users to trust providers to execute inference correctly and report token usage honestly. We present IMMACULATE, a practical auditing framework that detects economically motivated deviations-such as model substitution, quantization abuse, and token overbilling-without trusted hardware or access to model internals. IMMACULATE selectively audits a small fraction of requests using verifiable computation, achieving strong detection guarantees while amortizing cryptographic overhead. Experiments on dense and MoE models show that IMMACULATE reliably distinguishes benign and malicious executions with under 1% throughput overhead. Our code is published at https://github.com/guo-yanpei/Immaculate.

SQL is the de facto interface for exploratory data analysis; however, releasing exact query results can expose sensitive information through membership or attribute inference attacks. Differential privacy (DP) provides rigorous privacy guarantees, but in practice, DP alone may not satisfy governance requirements such as the \emph{minimum frequency rule}, which requires each released group (cell) to include contributions from at least $k$ distinct individuals. In this paper, we present \textbf{DPSQL+}, a privacy-preserving SQL library that simultaneously enforces user-level $(\varepsilon,δ)$-DP and the minimum frequency rule. DPSQL+ adopts a modular architecture consisting of: (i) a \emph{Validator} that statically restricts queries to a DP-safe subset of SQL; (ii) an \emph{Accountant} that consistently tracks cumulative privacy loss across multiple queries; and (iii) a \emph{Backend} that interfaces with various database engines, ensuring portability and extensibility. Experiments on the TPC-H benchmark demonstrate that DPSQL+ achieves practical accuracy across a wide range of analytical workloads -- from basic aggregates to quadratic statistics and join operations -- and allows substantially more queries under a fixed global privacy budget than prior libraries in our evaluation.

Leveraging Large Language Models (LLMs) for Knowledge Graph Completion (KGC) is promising but hindered by a fundamental granularity mismatch. LLMs operate on fragmented token sequences, whereas entities are the fundamental units in knowledge graphs (KGs) scenarios. Existing approaches typically constrain predictions to limited candidate sets or align entities with the LLM's vocabulary by pooling multiple tokens or decomposing entities into fixed-length token sequences, which fail to capture both the semantic meaning of the text and the structural integrity of the graph. To address this, we propose KGT, a novel framework that uses dedicated entity tokens to enable efficient, full-space prediction. Specifically, we first introduce specialized tokenization to construct feature representations at the level of dedicated entity tokens. We then fuse pre-trained structural and textual features into these unified embeddings via a relation-guided gating mechanism, avoiding training from scratch. Finally, we implement decoupled prediction by leveraging independent heads to separate and combine semantic and structural reasoning. Experimental results show that KGT consistently outperforms state-of-the-art methods across multiple benchmarks.

The rapid evolution of Large Language Models (LLMs) has accelerated the transition from conversational chatbots to general agents. However, effectively balancing empathetic communication with budget-aware decision-making remains an open challenge. Since existing methods fail to capture these complex strategic trade-offs, we propose InteractCS-RL, a framework that reframes task-oriented dialogue as a multi-granularity reinforcement learning process. Specifically, we first establish a User-centric Interaction Framework to provide a high-fidelity training gym, enabling agents to dynamically explore diverse strategies with persona-driven users. Then, we introduce Cost-aware Multi-turn Policy Optimization (CMPO) with a hybrid advantage estimation strategy. By integrating generative process credits and employing a PID-Lagrangian cost controller, CMPO effectively guides the policy to explore Pareto boundary between user reward and global cost constraints. Extensive experiments on customized real business scenarios demonstrate that InteractCS-RL significantly outperform other baselines across three evaluation dimensions. Further evaluation on tool-agent-user interaction benchmarks verify InteractCS-RL robustness across diverse domains.

Current approaches to developing persuasive dialogue agents often rely on a limited set of predefined persuasive strategies that fail to capture the complexity of real-world interactions. We applied a cross-disciplinary approach to develop a framework for designing persuasive dialogue agents that draws on proven strategies from social psychology, behavioral economics, and communication theory. We validated our proposed framework through experiments on two distinct datasets: the Persuasion for Good dataset, which represents a specific in-domain scenario, and the DailyPersuasion dataset, which encompasses a wide range of scenarios. The proposed framework achieved strong results for both datasets and demonstrated notable improvement in the persuasion success rate as well as promising generalizability. Notably, the proposed framework also excelled at persuading individuals with initially low intent, which addresses a critical challenge for persuasive dialogue agents.

Intermittent demand, a pattern characterized by long sequences of zero sales punctuated by sporadic, non-zero values, poses a persistent challenge in retail and supply chain forecasting. Both traditional methods, such as ARIMA, exponential smoothing, or Croston variants, as well as modern neural architectures such as DeepAR and Transformer-based models often underperform on such data, as they treat demand as a single continuous process or become computationally expensive when scaled across many sparse series. To address these limitations, we introduce Switch-Hurdle: a new framework that integrates a Mixture-of-Experts (MoE) encoder with a Hurdle-based probabilistic decoder. The encoder uses a sparse Top-1 expert routing during the forward pass yet approximately dense in the backward pass via a straight-through estimator (STE). The decoder follows a cross-attention autoregressive design with a shared hurdle head that explicitly separates the forecasting task into two components: a binary classification component estimating the probability of a sale, and a conditional regression component, predicting the quantity given a sale. This structured separation enables the model to capture both occurrence and magnitude processes inherent to intermittent demand. Empirical results on the M5 benchmark and a large proprietary retail dataset show that Switch-Hurdle achieves state-of-the-art prediction performance while maintaining scalability.

The rapid advancement of AI-powered smart glasses, one of the hottest wearable devices, has unlocked new frontiers for multimodal interaction, with Visual Question Answering (VQA) over external knowledge sources emerging as a core application. Existing Vision Language Models (VLMs) adapted to smart glasses are typically trained and evaluated on traditional multimodal datasets; however, these datasets lack the variety and realism needed to reflect smart glasses usage scenarios and diverge from their specific challenges, where accurately identifying the object of interest must precede any external knowledge retrieval. To bridge this gap, we introduce SUPERGLASSES, the first comprehensive VQA benchmark built on real-world data entirely collected by smart glasses devices. SUPERGLASSES comprises 2,422 egocentric image-question pairs spanning 14 image domains and 8 query categories, enriched with full search trajectories and reasoning annotations. We evaluate 26 representative VLMs on this benchmark, revealing significant performance gaps. To address the limitations of existing models, we further propose SUPERLENS, a multimodal smart glasses agent that enables retrieval-augmented answer generation by integrating automatic object detection, query decoupling, and multimodal web search. Our agent achieves state-of-the-art performance, surpassing GPT-4o by 2.19 percent, and highlights the need for task-specific solutions in smart glasses VQA scenarios.

Pre-training Large Language Models requires immense computational resources, making optimizer efficiency essential. The optimization landscape is highly anisotropic, with loss reduction driven predominantly by progress along flat directions. While matrix-based optimizers such as Muon and SOAP leverage fine-grained curvature information to outperform AdamW, their updates tend toward isotropy -- relatively conservative along flat directions yet potentially aggressive along sharp ones. To address this limitation, we first establish a unified Riemannian Ordinary Differential Equation (ODE) framework that elucidates how common adaptive algorithms operate synergistically: the preconditioner induces a Riemannian geometry that mitigates ill-conditioning, while momentum serves as a Riemannian damping term that promotes convergence. Guided by these insights, we propose LITE, a generalized acceleration strategy that enhances training dynamics by applying larger Hessian damping coefficients and learning rates along flat trajectories. Extensive experiments demonstrate that LITE significantly accelerates both Muon and SOAP across diverse architectures (Dense, MoE), parameter scales (130M--1.3B), datasets (C4, Pile), and learning-rate schedules (cosine, warmup-stable-decay). Theoretical analysis confirms that LITE facilitates faster convergence along flat directions in anisotropic landscapes, providing a principled approach to efficient LLM pre-training. The code is available at https://github.com/SHUCHENZHU/LITE.

Large language models have enabled agents that reason, plan, and interact with tools and environments to accomplish complex tasks. As these agents operate over extended interaction horizons, their effectiveness increasingly depends on adapting behavior to individual users and maintaining continuity across time, giving rise to personalized LLM-powered agents. In such long-term, user-dependent settings, personalization permeates the entire decision pipeline rather than remaining confined to surface-level generation. This survey provides a capability-oriented review of personalized LLM-powered agents. We organize the literature around four interdependent components: profile modeling, memory, planning, and action execution. Using this taxonomy, we synthesize representative methods and analyze how user signals are represented, propagated, and utilized, highlighting cross-component interactions and recurring design trade-offs. We further examine evaluation metrics and benchmarks tailored to personalized agents, summarize application scenarios spanning general assistance to specialized domains, and outline future directions for research and deployment. By offering a structured framework for understanding and designing personalized LLM-powered agents, this survey charts a roadmap toward more user-aligned, adaptive, robust, and deployable agentic systems, accelerating progress from prototype personalization to scalable real-world assistants.

Image-text retrieval has become a fundamental component in intelligent multimedia systems; however, most existing vision-language models are optimized for highresource languages and remain suboptimal for low-resource settings such as Vietnamese. This work introduces ViCLIP-OT, a foundation vision-language model specifically designed for Vietnamese image-text retrieval. The proposed framework integrates CLIP-style contrastive learning with a Similarity-Graph Regularized Optimal Transport (SIGROT) loss to enhance global cross-modal consistency and mitigate modality gap issues. Extensive experiments on three Vietnamese benchmarks (UITOpenViIC, KTVIC, and Crossmodal-3600) demonstrate that ViCLIP-OT consistently outperforms CLIP and SigLIP baselines in both in-domain and zero-shot settings. On UIT-OpenViIC, the model achieves an average Recall@K of 67.34%, improving upon CLIP by 5.75 percentage points. In zero-shot evaluation on Crossmodal-3600, ViCLIPOT surpasses CLIP by 11.72 percentage points. Embedding-space analysis further confirms improved alignment and reduced modality gap. The results indicate that integrating SIGROT provides an effective and scalable strategy for cross-modal retrieval in low-resource languages, offering practical implications for intelligent multimedia retrieval systems in Vietnamese and other underrepresented linguistic contexts.

Recent deep research agents primarily improve performance by scaling reasoning depth, but this leads to high inference cost and latency in search-intensive scenarios. Moreover, generalization across heterogeneous research settings remains challenging. In this work, we propose \emph{Search More, Think Less} (SMTL), a framework for long-horizon agentic search that targets both efficiency and generalization. SMTL replaces sequential reasoning with parallel evidence acquisition, enabling efficient context management under constrained context budgets. To support generalization across task types, we further introduce a unified data synthesis pipeline that constructs search tasks spanning both deterministic question answering and open-ended research scenarios with task appropriate evaluation metrics. We train an end-to-end agent using supervised fine-tuning and reinforcement learning, achieving strong and often state of the art performance across benchmarks including BrowseComp (48.6\%), GAIA (75.7\%), Xbench (82.0\%), and DeepResearch Bench (45.9\%). Compared to Mirothinker-v1.0, SMTL with maximum 100 interaction steps reduces the average number of reasoning steps on BrowseComp by 70.7\%, while improving accuracy.

Antimicrobial resistance (AMR) is a growing global crisis projected to cause 10 million deaths per year by 2050. While the WHO Global Antimicrobial Resistance and Use Surveillance System (GLASS) provides standardized surveillance data across 44 countries, few studies have applied machine learning to forecast population-level resistance trends from this data. This paper presents a two-component framework for AMR trend forecasting and evidence-grounded policy decision support. We benchmark six models -- Naive, Linear Regression, Ridge Regression, XGBoost, LightGBM, and LSTM -- on 5,909 WHO GLASS observations across six WHO regions (2021-2023). XGBoost achieved the best performance with a test MAE of 7.07% and R-squared of 0.854, outperforming the naive baseline by 83.1%. Feature importance analysis identified the prior-year resistance rate as the dominant predictor (50.5% importance), while regional MAE ranged from 4.16% (European Region) to 10.14% (South-East Asia Region). We additionally implemented a Retrieval-Augmented Generation (RAG) pipeline combining a ChromaDB vector store of WHO policy documents with a locally deployed Phi-3 Mini language model, producing source-attributed, hallucination-constrained policy answers. Code and data are available at https://github.com/TanvirTurja

Carsickness has gained significant attention with the rise of automated vehicles, prompting extensive research across on-road, test-track, and driving simulator environments to understand its occurrence and develop mitigation strategies. However, the lack of carsickness standardization complicates comparisons across studies and environments. Previous works demonstrate measurement validity between two setups at most (e.g., on-road vs. driving simulator), leaving gaps in multi-environment comparisons. This study investigates the recreation of an on-road motion sickness exposure - previously replicated on a test track - using a motion-based driving simulator. Twenty-eight participants performed an eyes-off-road non-driving task while reporting motion sickness using the Misery Scale during the experiment and the Motion Sickness Assessment Questionnaire afterward. Psychological factors known to influence motion sickness were also assessed. The results present subjective and objective measurements for motion sickness across the considered environments. In this paper, acceleration measurements, objective metrics and subjective motion sickness ratings across environments are compared, highlighting key differences in sickness occurrence for simulator-based research validity. Significantly lower motion sickness scores are reported in the simulator compared to on-road and test-track conditions, due to its limited working envelope to reproduce low-frequency (<0.5 Hz) motions, which are the most provocative for motion sickness.

Although diffusion language models (DLMs) are evolving quickly, many recent models converge on a set of shared components. These components, however, are distributed across ad-hoc research codebases or lack transparent implementations, making them difficult to reproduce or extend. As the field accelerates, there is a clear need for a unified framework that standardizes these common components while remaining flexible enough to support new methods and architectures. To address this gap, we introduce dLLM, an open-source framework that unifies the core components of diffusion language modeling -- training, inference, and evaluation -- and makes them easy to customize for new designs. With dLLM, users can reproduce, finetune, deploy, and evaluate open-source large DLMs such as LLaDA and Dream through a standardized pipeline. The framework also provides minimal, reproducible recipes for building small DLMs from scratch with accessible compute, including converting any BERT-style encoder or autoregressive LM into a DLM. We also release the checkpoints of these small DLMs to make DLMs more accessible and accelerate future research.

Recent advances in generic large models, such as GPT and DeepSeek, have motivated the introduction of universality to graph pre-training, aiming to learn rich and generalizable knowledge across diverse domains using graph representations to improve performance in various downstream applications. However, most existing methods face challenges in learning effective knowledge from generic graphs, primarily due to simplistic data alignment and limited training guidance. The issue of simplistic data alignment arises from the use of a straightforward unification for highly diverse graph data, which fails to align semantics and misleads pre-training models. The problem with limited training guidance lies in the arbitrary application of in-domain pre-training paradigms to cross-domain scenarios. While it is effective in enhancing discriminative representation in one data space, it struggles to capture effective knowledge from many graphs. To address these challenges, we propose a novel Latent sEmantic Distribution Alignment (LEDA) model for universal graph pre-training. Specifically, we first introduce a dimension projection unit to adaptively align diverse domain features into a shared semantic space with minimal information loss. Furthermore, we design a variational semantic inference module to obtain the shared latent distribution. The distribution is then adopted to guide the domain projection, aligning it with shared semantics across domains and ensuring cross-domain semantic learning. LEDA exhibits strong performance across a broad range of graphs and downstream tasks. Remarkably, in few-shot cross-domain settings, it significantly outperforms in-domain baselines and advanced universal pre-training models.

Deepfake speech utterances can be forged by replacing one or more words in a bona fide utterance with semantically different words synthesized by speech generative models. While a dedicated synthetic word detector could be developed, we investigate a cost-effective method that fine-tunes a pre-trained Whisper model to detect synthetic words while transcribing the input utterance via next-token prediction. We further investigate using partially vocoded utterances as the fine-tuning data, thereby reducing the cost of data collection. Our experiments demonstrate that, on in-domain test data, the fine-tuned Whisper yields low synthetic-word detection error rates and transcription error rates. On out-of-domain test data with synthetic words produced by unseen speech generative models, the fine-tuned Whisper remains on par with a dedicated ResNet-based detection model; however, the overall performance degradation calls for strategies to improve its generalization capability.

In online advertising, the inherent complexity and dynamic nature of advertising environments necessitate the use of auto-bidding services to assist advertisers in bid optimization. This complexity is further compounded in multi-channel scenarios, where effective allocation of budgets and constraints across channels with distinct behavioral patterns becomes critical for optimizing return on investment. Current approaches predominantly rely on either optimization-based strategies or reinforcement learning techniques. However, optimization-based methods lack flexibility in adapting to dynamic market conditions, while reinforcement learning approaches often struggle to capture essential historical dependencies and observational patterns within the constraints of Markov Decision Process frameworks. To address these limitations, we propose AHBid, an Adaptable Hierarchical Bidding framework that integrates generative planning with real-time control. The framework employs a high-level generative planner based on diffusion models to dynamically allocate budgets and constraints by effectively capturing historical context and temporal patterns. We introduce a constraint enforcement mechanism to ensure compliance with specified constraints, along with a trajectory refinement mechanism that enhances adaptability to environmental changes through the utilization of historical data. The system further incorporates a control-based bidding algorithm that synergistically combines historical knowledge with real-time information, significantly improving both adaptability and operational efficacy. Extensive experiments conducted on large-scale offline datasets and through online A/B tests demonstrate the effectiveness of AHBid, yielding a 13.57% increase in overall return compared to existing baselines.

Generative retrieval has emerged as a powerful paradigm for LLM-based recommendation. However, industrial recommender systems often benefit from restricting the output space to a constrained subset of items based on business logic (e.g. enforcing content freshness or product category), which standard autoregressive decoding cannot natively support. Moreover, existing constrained decoding methods that make use of prefix trees (Tries) incur severe latency penalties on hardware accelerators (TPUs/GPUs). In this work, we introduce STATIC (Sparse Transition Matrix-Accelerated Trie Index for Constrained Decoding), an efficient and scalable constrained decoding technique designed specifically for high-throughput LLM-based generative retrieval on TPUs/GPUs. By flattening the prefix tree into a static Compressed Sparse Row (CSR) matrix, we transform irregular tree traversals into fully vectorized sparse matrix operations, unlocking massive efficiency gains on hardware accelerators. We deploy STATIC on a large-scale industrial video recommendation platform serving billions of users. STATIC produces significant product metric impact with minimal latency overhead (0.033 ms per step and 0.25% of inference time), achieving a 948x speedup over a CPU trie implementation and a 47-1033x speedup over a hardware-accelerated binary-search baseline. Furthermore, the runtime overhead of STATIC remains extremely low across a wide range of practical configurations. To the best of our knowledge, STATIC enables the first production-scale deployment of strictly constrained generative retrieval. In addition, evaluation on academic benchmarks demonstrates that STATIC can considerably improve cold-start performance for generative retrieval. Our code is available at https://github.com/youtube/static-constraint-decoding.

Universal graph pre-training has emerged as a key paradigm in graph representation learning, offering a promising way to train encoders to learn transferable representations from unlabeled graphs and to effectively generalize across a wide range of downstream tasks. However, recent explorations in universal graph pre-training primarily focus on homogeneous graphs and it remains unexplored for heterogeneous graphs, which exhibit greater structural and semantic complexity. This heterogeneity makes it highly challenging to train a universal encoder for diverse heterogeneous graphs: (i) the diverse types with dataset-specific semantics hinder the construction of a unified representation space; (ii) the number and semantics of meta-paths vary across datasets, making encoding and aggregation patterns learned from one dataset difficult to apply to others. To address these challenges, we propose a novel Meta-path-aware Universal heterogeneous Graph pre-training (MUG) approach. Specifically, for challenge (i), MUG introduces a input unification module that integrates information from multiple node and relation types within each heterogeneous graph into a unified representation.This representation is then projected into a shared space by a dimension-aware encoder, enabling alignment across graphs with diverse schemas.Furthermore, for challenge (ii), MUG trains a shared encoder to capture consistent structural patterns across diverse meta-path views rather than relying on dataset-specific aggregation strategies, while a global objective encourages discriminability and reduces dataset-specific biases. Extensive experiments demonstrate the effectiveness of MUG on some real datasets.

Chain-of-Thought (CoT) has substantially empowered Large Language Models (LLMs) to tackle complex reasoning tasks, yet the verbose nature of explicit reasoning steps incurs prohibitive inference latency and computational costs, limiting real-world deployment. While existing compression methods - ranging from self-training to Reinforcement Learning (RL) with length constraints - attempt to mitigate this, they often sacrifice reasoning capability for brevity. We identify a critical failure mode in these approaches: explicitly optimizing for shorter trajectories triggers rapid entropy collapse, which prematurely shrinks the exploration space and stifles the discovery of valid reasoning paths, particularly for challenging questions requiring extensive deduction. To address this issue, we propose Compress responses for Easy questions and Explore Hard ones (CEEH), a difficulty-aware approach to RL-based efficient reasoning. CEEH dynamically assesses instance difficulty to apply selective entropy regularization: it preserves a diverse search space for currently hard questions to ensure robustness, while permitting aggressive compression on easier instances where the reasoning path is well-established. In addition, we introduce a dynamic optimal-length penalty anchored to the historically shortest correct response, which effectively counteracts entropy-induced length inflation and stabilizes the reward signal. Across six reasoning benchmarks, CEEH consistently reduces response length while maintaining accuracy comparable to the base model, and improves Pass@k relative to length-only optimization.

Route-planning agents powered by large language models (LLMs) have emerged as a promising paradigm for supporting everyday human mobility through natural language interaction and tool-mediated decision making. However, systematic evaluation in real-world mobility settings is hindered by diverse routing demands, non-deterministic mapping services, and limited reproducibility. In this study, we introduce MobilityBench, a scalable benchmark for evaluating LLM-based route-planning agents in real-world mobility scenarios. MobilityBench is constructed from large-scale, anonymized real user queries collected from Amap and covers a broad spectrum of route-planning intents across multiple cities worldwide. To enable reproducible, end-to-end evaluation, we design a deterministic API-replay sandbox that eliminates environmental variance from live services. We further propose a multi-dimensional evaluation protocol centered on outcome validity, complemented by assessments of instruction understanding, planning, tool use, and efficiency. Using MobilityBench, we evaluate multiple LLM-based route-planning agents across diverse real-world mobility scenarios and provide an in-depth analysis of their behaviors and performance. Our findings reveal that current models perform competently on Basic information retrieval and Route Planning tasks, yet struggle considerably with Preference-Constrained Route Planning, underscoring significant room for improvement in personalized mobility applications. We publicly release the benchmark data, evaluation toolkit, and documentation at https://github.com/AMAP-ML/MobilityBench .

Differentially private federated learning (DP-FL) enables clients to collaboratively train machine learning models while preserving the privacy of their local data. However, most existing DP-FL approaches assume that all clients share a uniform privacy budget, an assumption that does not hold in real-world scenarios where privacy requirements vary widely. This privacy heterogeneity poses a significant challenge: conventional client selection strategies, which typically rely on data quantity, cannot distinguish between clients providing high-quality updates and those introducing substantial noise due to strict privacy constraints. To address this gap, we present the first systematic study of privacy-aware client selection in DP-FL. We establish a theoretical foundation by deriving a convergence analysis that quantifies the impact of privacy heterogeneity on training error. Building on this analysis, we propose a privacy-aware client selection strategy, formulated as a convex optimization problem, that adaptively adjusts selection probabilities to minimize training error. Extensive experiments on benchmark datasets demonstrate that our approach achieves up to a 10% improvement in test accuracy on CIFAR-10 compared to existing baselines under heterogeneous privacy budgets. These results highlight the importance of incorporating privacy heterogeneity into client selection for practical and effective federated learning.

Neural networks are increasingly deployed in safety- and mission-critical pipelines, yet many verification and analysis results are produced outside the programming environment that defines and runs the model. This separation creates a semantic gap between the executed network and the analyzed artifact, so guarantees can hinge on implicit conventions such as operator semantics, tensor layouts, preprocessing, and floating-point corner cases. We introduce TorchLean, a framework in the Lean 4 theorem prover that treats learned models as first-class mathematical objects with a single, precise semantics shared by execution and verification. TorchLean unifies (1) a PyTorch-style verified API with eager and compiled modes that lower to a shared op-tagged SSA/DAG computation-graph IR, (2) explicit Float32 semantics via an executable IEEE-754 binary32 kernel and proof-relevant rounding models, and (3) verification via IBP and CROWN/LiRPA-style bound propagation with certificate checking. We validate TorchLean end-to-end on certified robustness, physics-informed residual bounds for PINNs, and Lyapunov-style neural controller verification, alongside mechanized theoretical results including a universal approximation theorem. These results demonstrate a semantics-first infrastructure for fully formal, end-to-end verification of learning-enabled systems.

This paper proposes HyperKKL, a novel learning approach for designing Kazantzis-Kravaris/Luenberger (KKL) observers for non-autonomous nonlinear systems. While KKL observers offer a rigorous theoretical framework by immersing nonlinear dynamics into a stable linear latent space, its practical realization relies on solving Partial Differential Equations (PDE) that are analytically intractable. Current existing learning-based approximations of the KKL observer are mostly designed for autonomous systems, failing to generalize to driven dynamics without expensive retraining or online gradient updates. HyperKKL addresses this by employing a hypernetwork architecture that encodes the exogenous input signal to instantaneously generate the parameters of the KKL observer, effectively learning a family of immersion maps parameterized by the external drive. We rigorously evaluate this approach against a curriculum learning strategy that attempts to generalize from autonomous regimes via training heuristics alone. The novel approach is illustrated on four numerical simulations in benchmark examples including the Duffing, Van der Pol, Lorenz, and Rössler systems.

Editing images via instruction provides a natural way to generate interactive content, but it is a big challenge due to the higher requirement of scene understanding and generation. Prior work utilizes a chain of large language models, object segmentation models, and editing models for this task. However, the understanding models provide only a single modality ability, restricting the editing quality. We aim to bridge understanding and generation via a new multi-modality model that provides the intelligent abilities to instruction-based image editing models for more complex cases. To achieve this goal, we individually separate the instruction editing task with the multi-modality chain of thought prompts, i.e., Chain-of-Thought (CoT) planning, editing region reasoning, and editing. For Chain-of-Thought planning, the large language model could reason the appropriate sub-prompts considering the instruction provided and the ability of the editing network. For editing region reasoning, we train an instruction-based editing region generation network with a multi-modal large language model. Finally, a hint-guided instruction-based editing network is proposed for editing image generations based on the sizeable text-to-image diffusion model to accept the hints for generation. Extensive experiments demonstrate that our method has competitive editing abilities on complex real-world images.

We propose ContextRL, a novel framework that leverages context augmentation to overcome these bottlenecks. Specifically, to enhance Identifiability, we provide the reward model with full reference solutions as context, enabling fine-grained process verification to filter out false positives (samples with the right answer but low-quality reasoning process). To improve Reachability, we introduce a multi-turn sampling strategy where the reward model generates mistake reports for failed attempts, guiding the policy to "recover" correct responses from previously all-negative groups. Experimental results on 11 perception and reasoning benchmarks show that ContextRL significantly improves knowledge discovery efficiency. Notably, ContextRL enables the Qwen3-VL-8B model to achieve performance comparable to the 32B model, outperforming standard RLVR baselines by a large margin while effectively mitigating reward hacking. Our in-depth analysis reveals the significant potential of contextual information for improving reward model accuracy and document the widespread occurrence of reward hacking, offering valuable insights for future RLVR research.

Source-Free Domain Adaptive Object Detection (SF-DAOD) aims to adapt a detector trained on a labeled source domain to an unlabeled target domain without retaining any source data. Despite recent progress, most popular approaches focus on tuning pseudo-label thresholds or refining the teacher-student framework, while overlooking object-level structural cues within cross-domain data. In this work, we present CGSA, the first framework that brings Object-Centric Learning (OCL) into SF-DAOD by integrating slot-aware adaptation into the DETR-based detector. Specifically, our approach integrates a Hierarchical Slot Awareness (HSA) module into the detector to progressively disentangle images into slot representations that act as visual priors. These slots are then guided toward class semantics via a Class-Guided Slot Contrast (CGSC) module, maintaining semantic consistency and prompting domain-invariant adaptation. Extensive experiments on multiple cross-domain datasets demonstrate that our approach outperforms previous SF-DAOD methods, with theoretical derivations and experimental analysis further demonstrating the effectiveness of the proposed components and the framework, thereby indicating the promise of object-centric design in privacy-sensitive adaptation scenarios. Code is released at https://github.com/Michael-McQueen/CGSA.

Virtual beam diagnostics relies on computationally intensive beam dynamics simulations where high-dimensional charged particle beams evolve through the accelerator. We propose Latent Evolution Model (LEM), a hybrid machine learning framework with an autoencoder that projects high-dimensional phase spaces into lower-dimensional representations, coupled with transformers to learn temporal dynamics in the latent space. This approach provides a common foundational framework addressing multiple interconnected challenges in beam diagnostics. For \textit{forward modeling}, a Conditional Variational Autoencoder (CVAE) encodes 15 unique projections of the 6D phase space into a latent representation, while a transformer predicts downstream latent states from upstream inputs. For \textit{inverse problems}, we address two distinct challenges: (a) predicting upstream phase spaces from downstream observations by utilizing the same CVAE architecture with transformers trained on reversed temporal sequences along with aleatoric uncertainty quantification, and (b) estimating RF settings from the latent space of the trained LEM using a dedicated dense neural network that maps latent representations to RF parameters. For \textit{tuning problems}, we leverage the trained LEM and RF estimator within a Bayesian optimization framework to determine optimal RF settings that minimize beam loss. This paper summarizes our recent efforts and demonstrates how this unified approach effectively addresses these traditionally separate challenges.

Large Language Models (LLMs) obey consistent scaling laws -- empirical power-law fits that predict how loss decreases with compute, data, and parameters. While predictive, these laws are descriptive rather than prescriptive: they characterize typical training, not optimal training. Surprisingly few works have successfully challenged the data-efficiency bounds implied by these laws -- which is our primary focus. To that end, we introduce the Geodesic Hypothesis, positing that token sequences trace geodesics on a smooth semantic manifold and are therefore locally linear. Building on this principle, we propose a novel Semantic Tube Prediction (STP) task, a JEPA-style regularizer that confines hidden-state trajectories to a tubular neighborhood of the geodesic. STP generalizes JEPA to language without requiring explicit multi-view augmentations. We show this constraint improves signal-to-noise ratio, and consequently preserves diversity by preventing trajectory collisions during inference. Empirically, STP allows LLMs to match baseline accuracy with 16$\times$ less training data on the NL-RX-SYNTH dataset, directly violating the data term of Chinchilla-style scaling laws and demonstrating that principled geometric priors can surpass brute-force scaling. Code is available at https://github.com/galilai-group/llm-jepa#stp.

In Embedding-as-an-Interface (EaaI) settings, pre-trained models are queried for Intermediate Representations (IRs). The distributional properties of IRs can leak training-set membership signals, enabling Membership Inference Attacks (MIAs) whose strength varies across layers. Although Differentially Private Stochastic Gradient Descent (DP-SGD) mitigates such leakage, existing implementations employ per-example gradient clipping and a uniform, layer-agnostic noise multiplier, ignoring heterogeneous layer-wise MIA vulnerability. This paper introduces Layer-wise MIA-risk-aware DP-SGD (LM-DP-SGD), which adaptively allocates privacy protection across layers in proportion to their MIA risk. Specifically, LM-DP-SGD trains a shadow model on a public shadow dataset, extracts per-layer IRs from its train/test splits, and fits layer-specific MIA adversaries, using their attack error rates as MIA-risk estimates. Leveraging the cross-dataset transferability of MIAs, these estimates are then used to reweight each layer's contribution to the globally clipped gradient during private training, providing layer-appropriate protection under a fixed noise magnitude. We further establish theoretical guarantees on both privacy and convergence of LM-DP-SGD. Extensive experiments show that, under the same privacy budget, LM-DP-SGD reduces the peak IR-level MIA risk while preserving utility, yielding a superior privacy-utility trade-off.

Condition injection enables diffusion models to generate context-aware outputs, which is essential for many time-series tasks. However, heterogeneous conditional contexts (e.g., observed history, missingness patterns or outlier covariates) can induce heavy-tailed per-example gradients. Under Differentially Private Stochastic Gradient Descent (DP-SGD), these rare conditioning-driven heavy-tailed gradients disproportionately trigger global clipping, resulting in outlier-dominated updates, larger clipping bias, and degraded utility under a fixed privacy budget. In this paper, we propose DP-aware AdaLN-Zero, a drop-in sensitivity-aware conditioning mechanism for conditional diffusion transformers that limits conditioning-induced gain without modifying the DP-SGD mechanism. DP-aware AdaLN-Zero jointly constrains conditioning representation magnitude and AdaLN modulation parameters via bounded re-parameterization, suppressing extreme gradient tail events before gradient clipping and noise injection. Empirically, DP-SGD equipped with DP-aware AdaLN-Zero improves interpolation/imputation and forecasting under matched privacy settings. We observe consistent gains on a real-world power dataset and two public ETT benchmarks over vanilla DP-SGD. Moreover, gradient diagnostics attribute these improvements to conditioning-specific tail reshaping and reduced clipping distortion, while preserving expressiveness in non-private training. Overall, these results show that sensitivity-aware conditioning can substantially improve private conditional diffusion training without sacrificing standard performance.

Progress in hardware model checking depends critically on high-quality benchmarks. However, the community faces a significant benchmark gap: existing suites are limited in number, often distributed only in representations such as BTOR2 without access to the originating register-transfer-level (RTL) designs, and biased toward extreme difficulty where instances are either trivial or intractable. These limitations hinder rigorous evaluation of new verification techniques and encourage overfitting of solver heuristics to a narrow set of problems. To address this, we introduce EvolveGen, a framework for generating hardware model checking benchmarks by combining reinforcement learning (RL) with high-level synthesis (HLS). Our approach operates at an algorithmic level of abstraction in which an RL agent learns to construct computation graphs. By compiling these graphs under different synthesis directives, we produce pairs of functionally equivalent but structurally distinct hardware designs, inducing challenging model checking instances. Solver runtime is used as the reward signal, enabling the agent to autonomously discover and generate small-but-hard instances that expose solver-specific weaknesses. Experiments show that EvolveGen efficiently creates a diverse benchmark set in standard formats (e.g., AIGER and BTOR2) and effectively reveals performance bottlenecks in state-of-the-art model checkers.

We propose CoLyricist, an AI-assisted lyric writing tool designed to support the typical workflows of experienced lyricists and enhance their creative efficiency. While lyricists have unique processes, many follow common stages. Tools that fail to accommodate these stages challenge integration into creative practices. Existing research and tools lack sufficient understanding of these songwriting stages and their associated challenges, resulting in ineffective designs. Through a formative study involving semi-structured interviews with 10 experienced lyricists, we identified four key stages: Theme Setting, Ideation, Drafting Lyrics, and Melody Fitting. CoLyricist addresses these needs by incorporating tailored AI-driven support for each stage, optimizing the lyric writing process to be more seamless and efficient. To examine whether this workflow-aligned design also benefits those without prior experience, we conducted a user study with 16 participants, including both experienced and novice lyricists. Results showed that CoLyricist enhances the songwriting experience across skill levels. Novice users especially appreciated the Melody-Fitting feature, while experienced users valued the Ideation support.

Long-running agentic tasks, such as deep research, require multi-hop reasoning over information distributed across multiple webpages and documents. In such tasks, the LLM context is dominated by tokens from external retrieval, causing memory usage to grow rapidly and limiting decode performance. While several KV cache compression techniques exist for long-context inputs, we find that existing heuristics fail to support multi-step reasoning models effectively. We address this challenge with SideQuest -- a novel approach that leverages the Large Reasoning Model (LRM) itself to perform KV cache compression by reasoning about the usefulness of tokens in its context. To prevent the tokens associated with this management process from polluting the model's memory, we frame KV cache compression as an auxiliary task executed in parallel to the main reasoning task. Our evaluations, using a model trained with just 215 samples, show that SideQuest reduces peak token usage by up to 65% on agentic tasks with minimal degradation in accuracy, outperforming heuristic-based KV cache compression techniques.

Fairness in Continual Learning for Large Multimodal Models (LMMs) is an emerging yet underexplored challenge, particularly in the presence of imbalanced data distributions that can lead to biased model updates and suboptimal performance across tasks. While recent continual learning studies have made progress in addressing catastrophic forgetting, the problem of fairness caused the imbalanced data remains largely underexplored. This paper presents a novel Fairness Direct Preference Optimization (FaiDPO or $φ$-DPO) framework for continual learning in LMMs. In particular, we first propose a new continual learning paradigm based on Direct Preference Optimization (DPO) to mitigate catastrophic forgetting by aligning learning with pairwise preference signals. Then, we identify the limitations of conventional DPO in imbalanced data and present a new $φ$-DPO loss that explicitly addresses distributional biases. We provide a comprehensive theoretical analysis demonstrating that our approach addresses both forgetting and data imbalance. Additionally, to enable $φ$-DPO-based continual learning, we construct pairwise preference annotations for existing benchmarks in the context of continual learning. Extensive experiments and ablation studies show the proposed $φ$-DPO achieves State-of-the-Art performance across multiple benchmarks, outperforming prior continual learning methods of LMMs.

Large language models exhibit sophisticated capabilities, yet understanding how they work internally remains a central challenge. A fundamental obstacle is that training selects for behavior, not circuitry, so many weight configurations can implement the same function. Which internal structures reflect the computation, and which are accidents of a particular training run? This work extracts algorithmic cores: compact subspaces necessary and sufficient for task performance. Independently trained transformers learn different weights but converge to the same cores. Markov-chain transformers embed 3D cores in nearly orthogonal subspaces yet recover identical transition spectra. Modular-addition transformers discover compact cyclic operators at grokking that later inflate, yielding a predictive model of the memorization-to-generalization transition. GPT-2 language models govern subject-verb agreement through a single axis that, when flipped, inverts grammatical number throughout generation across scales. These results reveal low-dimensional invariants that persist across training runs and scales, suggesting that transformer computations are organized around compact, shared algorithmic structures. Mechanistic interpretability could benefit from targeting such invariants -- the computational essence -- rather than implementation-specific details.

We present BetterScene, an approach to enhance novel view synthesis (NVS) quality for diverse real-world scenes using extremely sparse, unconstrained photos. BetterScene leverages the production-ready Stable Video Diffusion (SVD) model pretrained on billions of frames as a strong backbone, aiming to mitigate artifacts and recover view-consistent details at inference time. Conventional methods have developed similar diffusion-based solutions to address these challenges of novel view synthesis. Despite significant improvements, these methods typically rely on off-the-shelf pretrained diffusion priors and fine-tune only the UNet module while keeping other components frozen, which still leads to inconsistent details and artifacts even when incorporating geometry-aware regularizations like depth or semantic conditions. To address this, we investigate the latent space of the diffusion model and introduce two components: (1) temporal equivariance regularization and (2) vision foundation model-aligned representation, both applied to the variational autoencoder (VAE) module within the SVD pipeline. BetterScene integrates a feed-forward 3D Gaussian Splatting (3DGS) model to render features as inputs for the SVD enhancer and generate continuous, artifact-free, consistent novel views. We evaluate on the challenging DL3DV-10K dataset and demonstrate superior performance compared to state-of-the-art methods.

Production LLM serving must simultaneously deliver high throughput, low latency, and sufficient context capacity under non-stationary traffic and mixed request requirements. Data parallelism (DP) maximizes throughput by running independent replicas, while tensor parallelism (TP) reduces per-request latency and pools memory for long-context inference. However, existing serving stacks typically commit to a static parallelism configuration at deployment; adapting to bursts, priorities, or long-context requests is often disruptive and slow. We present Flying Serving, a vLLM-based system that enables online DP-TP switching without restarting engine workers. Flying Serving makes reconfiguration practical by virtualizing the state that would otherwise force data movement: (i) a zero-copy Model Weights Manager that exposes TP shard views on demand, (ii) a KV Cache Adaptor that preserves request KV state across DP/TP layouts, (iii) an eagerly initialized Communicator Pool to amortize collective setup, and (iv) a deadlock-free scheduler that coordinates safe transitions under execution skew. Across three popular LLMs and realistic serving scenarios, Flying Serving improves performance by up to $4.79\times$ under high load and $3.47\times$ under low load while supporting latency- and memory-driven requests.

Quantization-Aware Training from scratch has emerged as a promising approach for building efficient large language models (LLMs) with extremely low-bit weights (sub 2-bit), which can offer substantial advantages for edge deployment. However, existing methods still fail to achieve satisfactory accuracy and scalability. In this work, we identify a parameter democratization effect as a key bottleneck: the sensitivity of all parameters becomes homogenized, severely limiting expressivity. To address this, we propose pQuant, a method that decouples parameters by splitting linear layers into two specialized branches: a dominant 1-bit branch for efficient computation and a compact high-precision branch dedicated to preserving the most sensitive parameters. Through tailored feature scaling, we explicitly guide the model to allocate sensitive parameters to the high-precision branch. Furthermore, we extend this branch into multiple, sparsely-activated experts, enabling efficient capacity scaling. Extensive experiments indicate our pQuant achieves state-of-the-art performance in extremely low-bit quantization.

Synthetic tabular data generation has attracted growing attention due to its importance for data augmentation, foundation models, and privacy. However, real-world tabular datasets increasingly contain free-form text fields (e.g., reviews or clinical notes) alongside structured numerical and categorical attributes. Generating such heterogeneous tables with joint modeling of different modalities remains challenging. Existing approaches broadly fall into two categories: diffusion-based methods and LLM-based methods. Diffusion models can capture complex dependencies over numerical and categorical features in continuous or discrete spaces, but extending them to open-ended text is nontrivial and often leads to degraded text quality. In contrast, LLM-based generators naturally produce fluent text, yet their discrete tokenization can distort precise or wide-range numerical values, hindering accurate modeling of both numbers and language. In this work, we propose TabDLM, a unified framework for free-form tabular data generation via a joint numerical--language diffusion model built on masked diffusion language models (MDLMs). TabDLM models textual and categorical features through masked diffusion, while modeling numerical features with a continuous diffusion process through learned specialized numeric tokens embedding; bidirectional attention then captures cross-modality interactions within a single model. Extensive experiments on diverse benchmarks demonstrate the effectiveness of TabDLM compared to strong diffusion- and LLM-based baselines.

Human evaluations play a central role in training and assessing AI models, yet these data are rarely treated as measurements subject to systematic error. This paper integrates psychometric rater models into the AI pipeline to improve the reliability and validity of conclusions drawn from human judgments. The paper reviews common rater effects, severity and centrality, that distort observed ratings, and demonstrates how item response theory rater models, particularly the multi-faceted Rasch model, can separate true output quality from rater behavior. Using the OpenAI summarization dataset as an empirical example, we show how adjusting for rater severity produces corrected estimates of summary quality and provides diagnostic insight into rater performance. Incorporating psychometric modeling into human-in-the-loop evaluation offers more principled and transparent use of human data, enabling developers to make decisions based on adjusted scores rather than raw, error-prone ratings. This perspective highlights a path toward more robust, interpretable, and construct-aligned practices for AI development and evaluation.

Industrial advertising question answering (QA) is a high-stakes task in which hallucinated content, particularly fabricated URLs, can lead to financial loss, compliance violations, and legal risk. Although Retrieval-Augmented Generation (RAG) is widely adopted, deploying it in production remains challenging because industrial knowledge is inherently relational, frequently updated, and insufficiently aligned with generation objectives. We propose a reinforced co-adaptation framework that jointly optimizes retrieval and generation through two components: (1) Graph-aware Retrieval (GraphRAG), which models entity-relation structure over a high-citation knowledge subgraph for multi-hop, domain-specific evidence selection; and (2) evidence-constrained reinforcement learning via Group Relative Policy Optimization (GRPO) with multi-dimensional rewards covering faithfulness, style compliance, safety, and URL validity. Experiments on an internal advertising QA dataset show consistent gains across expert-judged dimensions including accuracy, completeness, and safety, while reducing the hallucination rate by 72\%. A two-week online A/B test demonstrates a 28.6\% increase in like rate, a 46.2\% decrease in dislike rate, and a 92.7\% reduction in URL hallucination. The system has been running in production for over half a year and has served millions of QA interactions.

Example-based guidance is widely used to improve mathematical reasoning at inference time, yet its effectiveness is highly unstable across problems and models-even when the guidance is correct and problem-relevant. We show that this instability arises from a previously underexplored gap between strategy usage-whether a reasoning strategy appears in successful solutions-and strategy executability-whether the strategy remains effective when instantiated as guidance for a target model. Through a controlled analysis of paired human-written and model-generated solutions, we identify a systematic dissociation between usage and executability: human- and model-derived strategies differ in structured, domain-dependent ways, leading to complementary strengths and consistent source-dependent reversals under guidance. Building on this diagnosis, we propose Selective Strategy Retrieval (SSR), a test-time framework that explicitly models executability by selectively retrieving and combining strategies using empirical, multi-route, source-aware signals. Across multiple mathematical reasoning benchmarks, SSR yields reliable and consistent improvements over direct solving, in-context learning, and single-source guidance, improving accuracy by up to $+13$ points on AIME25 and $+5$ points on Apex for compact reasoning models. Code and benchmark are publicly available at: https://github.com/lwd17/strategy-execute-pipeline.

Circuit discovery has recently attracted attention as a potential research direction to explain the non-trivial behaviors of language models. It aims to find the computational subgraphs, also known as circuits, within the model that are responsible for solving specific tasks. However, most existing studies overlook the holistic nature of these circuits and require designing specific corrupted activations for different tasks, which is inaccurate and inefficient. In this work, we propose an end-to-end approach based on the principle of Information Bottleneck, called IBCircuit, to identify informative circuits holistically. IBCircuit is an optimization framework for holistic circuit discovery and can be applied to any given task without tediously corrupted activation design. In both the Indirect Object Identification (IOI) and Greater-Than tasks, IBCircuit identifies more faithful and minimal circuits in terms of critical node components and edge components compared to recent related work.

Shuffle exchanges intermediate results between upstream and downstream operators in distributed data processing and is usually the bottleneck due to factors such as small random I/Os and network contention. Several systems have been designed to improve shuffle efficiency, but from our experiences of running ultra-large clusters at Alibaba Cloud MaxCompute platform, we observe that they can not adapt to highly dynamic job characteristics and cluster resource conditions, and their fault tolerance mechanisms are passive and inefficient when failures are inevitable. To tackle their limitations, we design and implement FuxiShuffle as a general data shuffle service for the ultra-large production environment of MaxCompute, featuring good adaptability and efficient failure resilience. Specifically, to achieve good adaptability, FuxiShuffle dynamically selects the shuffle mode based on runtime information, conducts progress-aware scheduling for the downstream workers, and automatically determines the most suitable backup strategy for each shuffle data chunk. To make failure resilience efficient, FuxiShuffle actively ensures data availability with multi-replica failover, prevents memory overflow with careful memory management, and employs an incremental recovery mechanism that does not lose computation progress. Our experiments show that, compared to baseline systems, FuxiShuffle significantly reduces not only end-to-end job completion time but also aggregate resource consumption. Micro experiments suggest that our designs are effective in improving adaptability and failure resilience.

Vision-Language-Action (VLA) models are multimodal robotic task controllers that, given an instruction and visual inputs, produce a sequence of low-level control actions (or motor commands) enabling a robot to execute the requested task in the physical environment. These systems face the test oracle problem from multiple perspectives. On the one hand, a test oracle must be defined for each instruction prompt, which is a complex and non-generalizable approach. On the other hand, current state-of-the-art oracles typically capture symbolic representations of the world (e.g., robot and object states), enabling the correctness evaluation of a task, but fail to assess other critical aspects, such as the quality with which VLA-enabled robots perform a task. In this paper, we explore whether Metamorphic Testing (MT) can alleviate the test oracle problem in this context. To do so, we propose two metamorphic relation patterns and five metamorphic relations to assess whether changes to the test inputs impact the original trajectory of the VLA-enabled robots. An empirical study involving five VLA models, two simulated robots, and four robotic tasks shows that MT can effectively alleviate the test oracle problem by automatically detecting diverse types of failures, including, but not limited to, uncompleted tasks. More importantly, the proposed MRs are generalizable, making the proposed approach applicable across different VLA models, robots, and tasks, even in the absence of test oracles.

Retrieval-Augmented Generation (RAG) enhances large language models (LLMs) by incorporating external knowledge, yet traditional single-round retrieval struggles with complex multi-step reasoning. Agentic RAG addresses this by enabling LLMs to dynamically decide when and what to retrieve, but current RL-based training methods suffer from sparse outcome rewards that discard intermediate signals and low sample efficiency where failed samples contribute nothing. We propose Search-P1, a framework that introduces path-centric reward shaping for agentic RAG training, comprising two key components: (1) Path-Centric Reward, which evaluates the structural quality of reasoning trajectories through order-agnostic step coverage and soft scoring that extracts learning signals even from failed samples, and (2) Dual-Track Path Scoring with offline-generated reference planners that assesses paths from both self-consistency and reference-alignment perspectives. Experiments on multiple QA benchmarks demonstrate that Search-P1 achieves significant improvements over Search-R1 and other strong baselines, with an average accuracy gain of 7.7 points.

Attention scales quadratically with sequence length, fundamentally limiting long-context inference. Existing block-granularity sparsification can reduce latency, but coarse blocks impose an intrinsic sparsity ceiling, making further improvements difficult even with carefully engineered designs. We present S2O, which performs early stopping for sparse attention via online permutation. Inspired by virtual-to-physical address mapping in memory systems, S2O revisits and factorizes FlashAttention execution, enabling inference to load non-contiguous tokens rather than a contiguous span in the original order. Motivated by fine-grained structures in attention heatmaps, we transform explicit permutation into an online, index-guided, discrete loading policy; with extremely lightweight preprocessing and index-remapping overhead, it concentrates importance on a small set of high-priority blocks. Building on this importance-guided online permutation for loading, S2O further introduces an early-stopping rule: computation proceeds from high to low importance; once the current block score falls below a threshold, S2O terminates early and skips the remaining low-contribution blocks, thereby increasing effective sparsity and reducing computation under a controlled error budget. As a result, S2O substantially raises the practical sparsity ceiling. On Llama-3.1-8B under a 128K context, S2O reduces single-operator MSE by 3.82$\times$ at matched sparsity, and reduces prefill compute density by 3.31$\times$ at matched MSE; meanwhile, it preserves end-to-end accuracy and achieves 7.51$\times$ attention and 3.81$\times$ end-to-end speedups.

Classifier-free guidance (CFG) has helped diffusion models achieve great conditional generation in various fields. Recently, more diffusion guidance methods have emerged with improved generation quality and human preference. However, can these emerging diffusion guidance methods really achieve solid and significant improvements? In this paper, we rethink recent progress on diffusion guidance. Our work mainly consists of four contributions. First, we reveal a critical evaluation pitfall that common human preference models exhibit a strong bias towards large guidance scales. Simply increasing the CFG scale can easily improve quantitative evaluation scores due to strong semantic alignment, even if image quality is severely damaged (e.g., oversaturation and artifacts). Second, we introduce a novel guidance-aware evaluation (GA-Eval) framework that employs effective guidance scale calibration to enable fair comparison between current guidance methods and CFG by identifying the effects orthogonal and parallel to CFG effects. Third, motivated by the evaluation pitfall, we design Transcendent Diffusion Guidance (TDG) method that can significantly improve human preference scores in the conventional evaluation framework but actually does not work in practice. Fourth, in extensive experiments, we empirically evaluate recent eight diffusion guidance methods within the conventional evaluation framework and the proposed GA-Eval framework. Notably, simply increasing the CFG scales can compete with most studied diffusion guidance methods, while all methods suffer severely from winning rate degradation over standard CFG. Our work would strongly motivate the community to rethink the evaluation paradigm and future directions of this field.

Deep multi-view clustering has achieved remarkable progress but remains vulnerable to complex noise in real-world applications. Existing noisy robust methods predominantly rely on a simplified binary assumption, treating data as either perfectly clean or completely corrupted. This overlooks the prevalent existence of heterogeneous observation noise, where contamination intensity varies continuously across data. To bridge this gap, we propose a novel framework termed Quality-Aware Robust Multi-View Clustering (QARMVC). Specifically, QARMVC employs an information bottleneck mechanism to extract intrinsic semantics for view reconstruction. Leveraging the insight that noise disrupts semantic integrity and impedes reconstruction, we utilize the resulting reconstruction discrepancy to precisely quantify fine-grained contamination intensity and derive instance-level quality scores. These scores are integrated into a hierarchical learning strategy: at the feature level, a quality-weighted contrastive objective is designed to adaptively suppress the propagation of noise; at the fusion level, a high-quality global consensus is constructed via quality-weighted aggregation, which is subsequently utilized to align and rectify local views via mutual information maximization. Extensive experiments on five benchmark datasets demonstrate that QARMVC consistently outperforms state-of-the-art baselines, particularly in scenarios with heterogeneous noise intensities.

Mitigating climate change requires behaviour change. However, even climate-concerned individuals often hold misperceptions about which actions most reduce carbon emissions. We recruited 1201 climate-concerned individuals to examine whether discussing climate actions with a large language model (LLM) equipped with climate knowledge and prompted to provide personalised responses would foster more accurate perceptions of the impacts of climate actions and increase willingness to adopt feasible, high-impact behaviours. We compared this to having participants run a web search, have a conversation with an unspecialised LLM, and no intervention. The personalised climate LLM was the only condition that led to increased knowledge about the impacts of climate actions and greater intentions to adopt impactful behaviours. While the personalised climate LLM did not outperform a web search in improving understanding of climate action impacts, the ability of LLMs to deliver personalised, actionable guidance may make them more effective at motivating impactful pro-climate behaviour change.

The deployment of machine learning in high-stakes domains requires a balance between predictive safety and algorithmic fairness. However, existing fairness interventions often as- sume unconstrained resources and employ group-specific decision thresholds that violate anti- discrimination regulations. We introduce a post-hoc, model-agnostic threshold optimization framework that jointly balances safety, efficiency, and equity under strict and hard capacity constraints. To ensure legal compliance, the framework enforces a single, global decision thresh- old. We formulated a parameterized ethical loss function coupled with a bounded decision rule that mathematically prevents intervention volumes from exceeding the available resources. An- alytically, we prove the key properties of the deployed threshold, including local monotonicity with respect to ethical weighting and the formal identification of critical capacity regimes. We conducted extensive experimental evaluations on diverse high-stakes datasets. The principal re- sults demonstrate that capacity constraints dominate ethical priorities; the strict resource limit determines the final deployed threshold in over 80% of the tested configurations. Furthermore, under a restrictive 25% capacity limit, the proposed framework successfully maintains high risk identification (recall ranging from 0.409 to 0.702), whereas standard unconstrained fairness heuristics collapse to a near-zero utility. We conclude that theoretical fairness objectives must be explicitly subordinated to operational capacity limits to remain in deployment. By decou- pling predictive scoring from policy evaluation and strictly bounding intervention rates, this framework provides a practical and legally compliant mechanism for stakeholders to navigate unavoidable ethical trade-offs in resource-constrained environments.

Current safety mechanisms for Large Language Models (LLMs) rely heavily on static, fine-tuned classifiers that suffer from adaptation rigidity, the inability to enforce new governance rules without expensive retraining. To address this, we introduce CourtGuard, a retrieval-augmented multi-agent framework that reimagines safety evaluation as Evidentiary Debate. By orchestrating an adversarial debate grounded in external policy documents, CourtGuard achieves state-of-the-art performance across 7 safety benchmarks, outperforming dedicated policy-following baselines without fine-tuning. Beyond standard metrics, we highlight two critical capabilities: (1) Zero-Shot Adaptability, where our framework successfully generalized to an out-of-domain Wikipedia Vandalism task (achieving 90\% accuracy) by swapping the reference policy; and (2) Automated Data Curation and Auditing, where we leveraged CourtGuard to curate and audit nine novel datasets of sophisticated adversarial attacks. Our results demonstrate that decoupling safety logic from model weights offers a robust, interpretable, and adaptable path for meeting current and future regulatory requirements in AI governance.

Large reasoning models (LRMs) achieve strong performance through extended reasoning traces, but they often exhibit overthinking behavior for low-complexity queries. Existing efforts to mitigate this issue are fundamentally limited by unstable accuracy-efficiency trade-offs and poor robustness to heterogeneous reasoning behaviors. To address these challenges, we propose a two-stage framework for stable adaptive thinking in LRMs. The framework first applies Hybrid Fine-Tuning to expose the model to both thinking and no-thinking behaviors, establishing well-conditioned initialization. It then performs adaptive reinforcement learning with Correctness-Preserving Advantage Shaping (CPAS) to avoid suppressing correct long-chain reasoning, and Length-Aware Gradient Regulation (LAGR) to stabilize optimization under severe reasoning-length heterogeneity. Extensive experiments on Qwen2.5-1.5B and 7B show consistent improvements over strong baselines, achieving up to +3.7/+3.6 accuracy points while reducing generated tokens by 40.6%/43.9%. Further analyses across varying problem difficulties and out-of-distribution tasks confirm the robustness and generalization of our approach.

Electroencephalogram (EEG) signals have become a popular medium for decoding visual information due to their cost-effectiveness and high temporal resolution. However, current approaches face significant challenges in bridging the modality gap between EEG and image data. These methods typically rely on complex adaptation processes involving multiple stages, making it hard to maintain consistency and manage compounding errors. Furthermore, the computational overhead imposed by large-scale diffusion models limit their practicality in real-world brain-computer interface (BCI) applications. In this work, we present AVDE, a lightweight and efficient framework for visual decoding from EEG signals. First, we leverage LaBraM, a pre-trained EEG model, and fine-tune it via contrastive learning to align EEG and image representations. Second, we adopt an autoregressive generative framework based on a "next-scale prediction" strategy: images are encoded into multi-scale token maps using a pre-trained VQ-VAE, and a transformer is trained to autoregressively predict finer-scale tokens starting from EEG embeddings as the coarsest representation. This design enables coherent generation while preserving a direct connection between the input EEG signals and the reconstructed images. Experiments on two datasets show that AVDE outperforms previous state-of-the-art methods in both image retrieval and reconstruction tasks, while using only 10% of the parameters. In addition, visualization of intermediate outputs shows that the generative process of AVDE reflects the hierarchical nature of human visual perception. These results highlight the potential of autoregressive models as efficient and interpretable tools for practical BCI applications.

Large Language Models (LLMs) exhibit significant safety disparities across languages, with low-resource languages (LRLs) often bypassing safety guardrails established for high-resource languages (HRLs) like English. Existing solutions, such as multilingual supervised fine-tuning (SFT) or Reinforcement Learning from Human Feedback (RLHF), are computationally expensive and dependent on scarce multilingual safety data. In this work, we propose a novel, training-free alignment framework based on Sparse Weight Editing. Identifying that safety capabilities are localized within a sparse set of safety neurons, we formulate the cross-lingual alignment problem as a constrained linear transformation. We derive a closed-form solution to optimally map the harmful representations of LRLs to the robust safety subspaces of HRLs, while preserving general utility via a null-space projection constraint. Extensive experiments across 8 languages and multiple model families (Llama-3, Qwen-2.5) demonstrate that our method substantially reduces Attack Success Rate (ASR) in LRLs with negligible impact on general reasoning capabilities, all achieved with a single, data-efficient calculation.

Predictive modeling over relational databases (RDBs) powers applications, yet remains challenging due to capturing both cross-table dependencies and complex feature interactions. Relational Deep Learning (RDL) methods automate feature engineering via message passing, while classical approaches like Deep Feature Synthesis (DFS) rely on predefined non-parametric aggregators. Despite performance gains, the comparative advantages of RDL over DFS and the design principles for selecting effective architectures remain poorly understood. We present a comprehensive study that unifies RDL and DFS in a shared design space and conducts architecture-centric searches across diverse RDB tasks. Our analysis yields three key findings: (1) RDL does not consistently outperform DFS, with performance being highly task-dependent; (2) no single architecture dominates across tasks, underscoring the need for task-aware model selection; and (3) validation accuracy is an unreliable guide for architecture choice. This search yields a model performance bank that links architecture configurations to their performance; leveraging this bank, we analyze the drivers of the RDL-DFS performance gap and introduce two task signals -- RDB task homophily and an affinity embedding that captures size, path, feature, and temporal structure -- whose correlation with the gap enables principled routing. Guided by these signals, we propose Relatron, a task embedding-based meta-selector that chooses between RDL and DFS and prunes the within-family search. Lightweight loss-landscape metrics further guard against brittle checkpoints by preferring flatter optima. In experiments, Relatron resolves the "more tuning, worse performance" effect and, in joint hyperparameter-architecture optimization, achieves up to 18.5% improvement over strong baselines with 10x lower cost than Fisher information-based alternatives.

Cancer is often driven by specific combinations of an estimated two to nine gene mutations, known as multi-hit combinations. Identifying these combinations is critical for understanding carcinogenesis and designing targeted therapies. We formalise this challenge as the Multi-Hit Cancer Driver Set Cover Problem (MHCDSCP), a binary classification problem that selects gene combinations to maximise coverage of tumor samples while minimising coverage of normal samples. Existing approaches typically rely on exhaustive search and supercomputing infrastructure. In this paper, we present constraint programming and mixed integer programming formulations of the MHCDSCP. Evaluated on real-world cancer genomics data, our methods achieve performance comparable to state-of-the-art methods while running on a single commodity CPU in under a minute. Furthermore, we introduce a column generation heuristic capable of solving small instances to optimality. These results suggest that solving the MHCDSCP is less computationally intensive than previously believed, thereby opening research directions for exploring modelling assumptions.

Synthesis of diverse driving scenes serves as a crucial data augmentation technique for validating the robustness and generalizability of autonomous driving systems. Current methods aggregate high-definition (HD) maps and 3D bounding boxes as geometric conditions in diffusion models for conditional scene generation. However, implicit inter-condition dependency causes generation failures when control conditions change independently. Additionally, these methods suffer from insufficient details in both semantic and structural aspects. Specifically, brief and view-invariant captions restrict semantic contexts, resulting in weak background modeling. Meanwhile, the standard denoising loss with uniform spatial weighting neglects foreground structural details, causing visual distortions and blurriness. To address these challenges, we propose DrivePTS, which incorporates three key innovations. Firstly, our framework adopts a progressive learning strategy to mitigate inter-dependency between geometric conditions, reinforced by an explicit mutual information constraint. Secondly, a Vision-Language Model is utilized to generate multi-view hierarchical descriptions across six semantic aspects, providing fine-grained textual guidance. Thirdly, a frequency-guided structure loss is introduced to strengthen the model's sensitivity to high-frequency elements, improving foreground structural fidelity. Extensive experiments demonstrate that our DrivePTS achieves state-of-the-art fidelity and controllability in generating diverse driving scenes. Notably, DrivePTS successfully generates rare scenes where prior methods fail, highlighting its strong generalization ability.

The \textit{de facto} paradigm for applying dense retrieval (DR) to new tasks involves fine-tuning a pre-trained model for a specific task. However, this paradigm has two significant limitations: (1) It is difficult adapt the DR to a new domain if the training dataset is limited. (2) Old DR models are simply replaced by newer models that are trained from scratch when the former are no longer up to date. Especially for scenarios where the model needs to be updated frequently, this paradigm is prohibitively expensive. To address these challenges, we propose a novel dense retrieval approach, termed \textit{dynamic dense retrieval} (DDR). DDR uses \textit{prefix tuning} as a \textit{module} specialized for a specific domain. These modules can then be compositional combined with a dynamic routing strategy, enabling highly flexible domain adaptation in the retrieval part. Extensive evaluation on six zero-shot downstream tasks demonstrates that this approach can surpass DR while utilizing only 2\% of the training parameters, paving the way to achieve more flexible dense retrieval in IR. We see it as a promising future direction for applying dense retrieval to various tasks.

Large Language Model (LLM) based agents excel at general reasoning but often fail in specialized domains where success hinges on long-tail knowledge absent from their training data. While human experts can provide this missing knowledge, their guidance is often unstructured and unreliable, making its direct integration into an agent's plan problematic. To address this, we introduce AHCE (Active Human-Augmented Challenge Engagement), a framework for on-demand Human-AI collaboration. At its core, the Human Feedback Module (HFM) employs a learned policy to treat the human expert as an interactive reasoning tool. Extensive experiments in Minecraft demonstrate the framework's effectiveness, increasing task success rates by 32% on normal difficulty tasks and nearly 70% on highly difficult tasks, all with minimal human intervention. Our work demonstrates that successfully augmenting agents requires learning how to request expert reasoning, moving beyond simple requests for help.

Tau positron emission tomography (tau-PET) provides an in vivo marker of Alzheimer's disease pathology, but cost and limited availability motivate MRI-based alternatives. We introduce DisQ-HNet (DQH), a framework that synthesizes tau-PET from paired T1-weighted and FLAIR MRI while exposing how each modality contributes to the prediction. The method combines (i) a Partial Information Decomposition (PID)-guided, vector-quantized encoder that partitions latent information into redundant, unique, and complementary components, and (ii) a Half-UNet decoder that preserves anatomical detail using pseudo-skip connections conditioned on structural edge cues rather than direct encoder feature reuse. Across multiple baselines (VAE, VQ-VAE, and UNet), DisQ-HNet maintains reconstruction fidelity and better preserves disease-relevant signal for downstream AD tasks, including Braak staging, tau localization, and classification. PID-based Shapley analysis provides modality-specific attribution of synthesized uptake patterns.

Cone-beam computed tomography (CBCT) is widely used in dental and maxillofacial imaging, but low-dose acquisition introduces strong, spatially varying noise that degrades soft-tissue visibility and obscures fine anatomical structures. Classical denoising methods struggle to suppress noise in CBCT while preserving edges. Although deep learning-based approaches offer high-fidelity restoration, their use in CBCT denoising is limited by the scarcity of high-resolution CBCT data for supervised training. To address this research gap, we propose a novel Hybrid Attention Residual U-Net (HARU-Net) for high-quality denoising of CBCT data, trained on a cadaver dataset of human hemimandibles acquired using a high-resolution protocol of the 3D Accuitomo 170 (J. Morita, Kyoto, Japan) CBCT system. The novel contribution of this approach is the integration of three complementary architectural components: (i) a hybrid attention transformer block (HAB) embedded within each skip connection to selectively emphasize salient anatomical features, (ii) a residual hybrid attention transformer group (RHAG) at the bottleneck to strengthen global contextual modeling and long-range feature interactions, and (iii) residual learning convolutional blocks to facilitate deeper, more stable feature extraction throughout the network. HARU-Net consistently outperforms state-of-the-art (SOTA) methods including SwinIR and Uformer, achieving the highest PSNR (37.52 dB), highest SSIM (0.9557), and lowest GMSD (0.1084). This effective and clinically reliable CBCT denoising is achieved at a computational cost significantly lower than that of the SOTA methods, offering a practical advancement toward improving diagnostic quality in low-dose CBCT imaging.

Large Language Models (LLMs) face significant challenges regarding deployment costs and latency, necessitating adaptive computing strategies. Building upon the AI Flow framework, we introduce Ruyi2 as an evolution of our adaptive model series designed for efficient variable-depth computation. While early-exit architectures offer a viable efficiency-performance balance, the Ruyi model and existing methods often struggle with optimization complexity and compatibility with large-scale distributed training. To bridge this gap, Ruyi2 introduces a stable "Familial Model" based on Megatron-LM. By using 3D parallel training, it achieves a 2-3 times speedup over Ruyi, while performing comparably to same-sized Qwen3 models. These results confirm that family-based parameter sharing is a highly effective strategy, establishing a new "Train Once, Deploy Many" paradigm and providing a key reference for balancing architectural efficiency with high-performance capabilities.

Building a useful quantum computer is a grand science and engineering challenge, currently pursued intensely by teams around the world. In the 1980s, Richard Feynman and Yuri Manin observed independently that computers based on quantum mechanics might enable better simulations of quantum phenomena. Their vision remained an intellectual curiosity until Peter Shor published his famous quantum algorithm for integer factoring, and shortly thereafter a proof that errors in quantum computations can be corrected. Since then, quantum computing R&D has progressed rapidly, from small-scale experiments in university physics laboratories to well-funded industrial efforts and prototypes. Hype notwithstanding, quantum computers have yet to solve scientifically or practically important problems -- a target often called quantum utility. In this article, we describe the capabilities of contemporary quantum computers, compare them to the requirements of quantum utility, and illustrate how to track progress from today to utility. We highlight key science and engineering challenges on the road to quantum utility, touching on relevant aspects of our own research.

Agentic artificial intelligence (AI) is emerging as a key enabler for autonomous radio access networks (RANs), where multiple large language model (LLM)-based agents reason and collaborate to achieve operator-defined intents. The open RAN (O-RAN) architecture enables the deployment and coordination of such agents. However, most existing works consider simple intents handled by independent agents, while complex intents that require coordination among agents remain unexplored. In this paper, we propose an agentic AI framework for intent translation and optimization in cell-free O-RAN. A supervisor agent translates the operator intents into an optimization objective and minimum rate requirements. Based on this information, a user weighting agent retrieves relevant prior experience from a memory module to determine the user priority weights for precoding. If the intent includes an energy-saving objective, then an open radio unit (O-RU) management agent will also be activated to determine the set of active O-RUs by using a deep reinforcement learning (DRL) algorithm. A monitoring agent measures and monitors the user data rates and coordinates with other agents to guarantee the minimum rate requirements are satisfied. To enhance scalability, we adopt a parameter-efficient fine-tuning (PEFT) method that enables the same underlying LLM to be used for different agents. Simulation results show that the proposed agentic AI framework reduces the number of active O-RUs by 41.93% when compared with three baseline schemes in energy-saving mode. Using the PEFT method, the proposed framework reduces the memory usage by 92% when compared with deploying separate LLM agents.

Large reasoning models (LRMs) excel at a long chain of reasoning but often fail to faithfully follow instructions regarding output format, constraints, or specific requirements. We investigate whether this gap can be closed by integrating an instruction-tuned model (ITM) into an LRM. Analyzing their differences in parameter space, namely task vectors, we find that their principal subspaces are nearly orthogonal across key modules, suggesting a lightweight merging with minimal interference. However, we also demonstrate that naive merges are fragile because they overlook the output format mismatch between LRMs (with explicit thinking and response segments) and ITMs (answers-only). We introduce RAIN-Merging (Reasoning-Aware Instruction-attention guided Null-space projection Merging), a gradient-free method that integrates instruction following while preserving thinking format and reasoning performance. First, with a small reasoning calibration set, we project the ITM task vector onto the null space of forward features at thinking special tokens, which preserves the LRM's structured reasoning mechanisms. Second, using a small instruction calibration set, we estimate instruction attention to derive module-specific scaling that amplifies instruction-relevant components and suppresses leakage. Across four instruction-following benchmarks and nine reasoning & general capability benchmarks, RAIN-Merging substantially improves instruction adherence while maintaining reasoning quality. The gains are consistent across model scales and architectures, translating to improved performance in agent settings.

The rapid growth of scientific machine learning (SciML) has accelerated discovery across diverse domains, yet designing effective SciML models remains a challenging task. In practice, building such models often requires substantial prior knowledge and manual expertise, particularly in determining which input features to use and how large the model should be. We introduce LUMOS, an end-to-end framework based on L0-regularized learning that unifies feature selection and model pruning to democratize SciML model design. By employing semi-stochastic gating and reparameterization techniques, LUMOS dynamically selects informative features and prunes redundant parameters during training, reducing the reliance on manual tuning while maintaining predictive accuracy. We evaluate LUMOS across 13 diverse SciML workloads, including cosmology and molecular sciences, and demonstrate its effectiveness and generalizability. Experiments on 13 SciML models show that LUMOS achieves 71.45% parameter reduction and a 6.4x inference speedup on average. Furthermore, Distributed Data Parallel (DDP) training on up to eight GPUs confirms the scalability of

Matrix factorization techniques, especially Nonnegative Matrix Factorization (NMF), have been widely used for dimensionality reduction and interpretable data representation. However, existing NMF-based methods are inherently single-scale and fail to capture the evolution of connectivity structures across resolutions. In this work, we propose persistent nonnegative matrix factorization (pNMF), a scale-parameterized family of NMF problems, that produces a sequence of persistence-aligned embeddings rather than a single one. By leveraging persistent homology, we identify a canonical minimal sufficient scale set at which the underlying connectivity undergoes qualitative changes. These canonical scales induce a sequence of graph Laplacians, leading to a coupled NMF formulation with scale-wise geometric regularization and explicit cross-scale consistency constraint. We analyze the structural properties of the embeddings along the scale parameter and establish bounds on their increments between consecutive scales. The resulting model defines a nontrivial solution path across scales, rather than a single factorization, which poses new computational challenges. We develop a sequential alternating optimization algorithm with guaranteed convergence. Numerical experiments on synthetic and single-cell RNA sequencing datasets demonstrate the effectiveness of the proposed approach in multi-scale low-rank embeddings.

This study addresses a critical challenge in AI-based weather forecasting by developing an AI-driven optimized ensemble forecast system using Orthogonal Conditional Nonlinear Optimal Perturbations (O-CNOPs). The system bridges the gap between computational efficiency and dynamic consistency in tropical cyclone (TC) forecasting. Unlike conventional ensembles limited by computational costs or AI ensembles constrained by inadequate perturbation methods, O-CNOPs generate dynamically optimized perturbations that capture fast-growing errors of FuXi model while maintaining plausibility. The key innovation lies in producing orthogonal perturbations that respect FuXi nonlinear dynamics, yielding structures reflecting dominant dynamical controls and physically interpretable probabilistic forecasts. Demonstrating superior deterministic and probabilistic skills over the operational Integrated Forecasting System Ensemble Prediction System, this work establishes a new paradigm combining AI computational advantages with rigorous dynamical constraints. Success in TC track forecasting paves the way for reliable ensemble forecasts of other high-impact weather systems, marking a major step toward operational AI-based ensemble forecasting.

Learning the dynamic causal structure of time series is a challenging problem. Most existing approaches rely on distributional or structural invariance to uncover underlying causal dynamics, assuming stationary or partially stationary causality. However, these assumptions often conflict with the complex, time-varying causal relationships observed in real-world systems. This motivates the need for methods that address fully dynamic causality, where both instantaneous and lagged dependencies evolve over time. Such a setting poses significant challenges for the efficiency and stability of causal discovery. To address these challenges, we introduce DyCausal, a dynamic causal structure learning framework. DyCausal leverages convolutional networks to capture causal patterns within coarse-grained time windows, and then applies linear interpolation to refine causal structures at each time step, thereby recovering fine-grained and time-varying causal graphs. In addition, we propose an acyclic constraint based on matrix norm scaling, which improves efficiency while effectively constraining loops in evolving causal structures. Comprehensive evaluations on both synthetic and real-world datasets demonstrate that DyCausal achieves superior performance compared to existing methods, offering a stable and efficient approach for identifying fully dynamic causal structures from coarse to fine.

A mathematical concept is identified and analyzed that is implicit in the 2012 paper Turing Incomputable Computation, presented at the Alan Turing Centenary Conference (Turing 100, Manchester). The concept, called dynamic level sets, is distinct from mathematical concepts in the standard literature on dynamical systems, topology, and computability theory. A new mathematical object is explained and why it may have escaped prior characterizations, including the classical result of de Leeuw, Moore, Shannon, and Shapiro (1956) that probabilistic Turing machines compute no more than deterministic ones.

In the rapidly evolving field of digital libraries, the development of large language models (LLMs) has opened up new possibilities for simulating user behavior. This innovation addresses the longstanding challenge in digital library research: the scarcity of publicly available datasets on user search patterns due to privacy concerns. In this context, we introduce Agent4DL, a user search behavior simulator specifically designed for digital library environments. Agent4DL generates realistic user profiles and dynamic search sessions that closely mimic actual search strategies, including querying, clicking, and stopping behaviors tailored to specific user profiles. Our simulator's accuracy in replicating real user interactions has been validated through comparisons with real user data. Notably, Agent4DL demonstrates competitive performance compared to existing user search simulators such as SimIIR 2.0, particularly in its ability to generate more diverse and context-aware user behaviors.

Serves, especially first serves, are very important in professional tennis. Servers choose their serve directions strategically to maximize their winning chances while trying to be unpredictable. On the other hand, returners try to predict serve directions to make good returns. The mind game between servers and returners is an important part of decision-making in professional tennis matches. To help understand the players' serve decisions, we have developed a machine learning method for predicting professional tennis players' first serve directions. Through feature engineering, our method achieves an average prediction accuracy of around 49\% for male players and 44\% for female players. Our analysis provides some evidence that top professional players use a mixed-strategy model in serving decisions and that fatigue might be a factor in choosing serve directions. Our analysis also suggests that contextual information is perhaps more important for returners' anticipatory reactions than previously thought.

Dyslexia affects approximately 10% of the global population and presents persistent challenges in reading fluency and text comprehension. While existing assistive technologies address visual presentation, linguistic complexity remains a substantial barrier to equitable access. This paper presents an empirical study on dyslexia-friendly text summarization using an iterative prompt-based refinement pipeline built on GPT-4o. We evaluate the pipeline on approximately 2,000 news article samples, applying a readability target of Flesch Reading Ease >= 90. Results show that the majority of summaries meet the readability threshold within four attempts, with many succeeding on the first try. A composite score combining readability and semantic fidelity shows stable performance across the dataset, ranging from 0.13 to 0.73 with a typical value near 0.55. These findings establish an empirical baseline for accessibility-driven NLP summarization and motivate further human-centered evaluation with dyslexic readers.

While contemporary large language models (LLMs) are increasingly capable in isolation, there are still many difficult problems that lie beyond the abilities of a single LLM. For such tasks, there is still uncertainty about how best to take many LLMs as parts and combine them into a greater whole. This position paper argues that potential blueprints for designing such modular language agents can be found in the existing literature on cognitive models and artificial intelligence (AI) algorithms. To make this point clear, we formalize the idea of an agent template that specifies roles for individual LLMs and how their functionalities should be composed. We then survey a variety of existing language agents in the literature and highlight their underlying templates derived directly from cognitive models or AI algorithms. By highlighting these designs, we aim to call attention to agent templates inspired by cognitive science and AI as a powerful tool for developing effective, interpretable language agents.

Taiwanese Hakka is a low-resource, endangered language that poses significant challenges for automatic speech recognition (ASR), including high dialectal variability and the presence of two distinct writing systems (Hanzi and Pinyin). Traditional ASR models often encounter difficulties in this context, as they tend to conflate essential linguistic content with dialect-specific variations across both phonological and lexical dimensions. To address these challenges, we propose a unified framework grounded in the Recurrent Neural Network Transducers (RNN-T). Central to our approach is the introduction of dialect-aware modeling strategies designed to disentangle dialectal "style" from linguistic "content", which enhances the model's capacity to learn robust and generalized representations. Additionally, the framework employs parameter-efficient prediction networks to concurrently model ASR (Hanzi and Pinyin). We demonstrate that these tasks create a powerful synergy, wherein the cross-script objective serves as a mutual regularizer to improve the primary ASR tasks. Experiments conducted on the HAT corpus reveal that our model achieves 57.00% and 40.41% relative error rate reduction on Hanzi and Pinyin ASR, respectively. To our knowledge, this is the first systematic investigation into the impact of Hakka dialectal variations on ASR and the first single model capable of jointly addressing these tasks.

The negative sampling strategy can effectively train collaborative filtering (CF) recommendation models based on implicit feedback by constructing positive and negative samples. However, existing methods primarily optimize the negative sampling process while neglecting the exploration of positive samples. Some denoising recommendation methods can be applied to denoise positive samples within negative sampling strategies, but they ignore temporal information. Existing work integrates sequential information during model aggregation but neglects time interval information, hindering accurate capture of users' current preferences. To address this problem, from a data perspective, we propose a novel temporal filtration-enhanced approach to construct a high-quality positive sample set. First, we design a time decay model based on interaction time intervals, transforming the original graph into a weighted user-item bipartite graph. Then, based on predefined filtering operations, the weighted user-item bipartite graph is layered. Finally, we design a layer-enhancement strategy to construct a high-quality positive sample set for the layered subgraphs. We provide theoretical insights into why TFPS can improve Recall@k and NDCG@k, and extensive experiments on three real-world datasets demonstrate the effectiveness of the proposed method. Additionally, TFPS can be integrated with various implicit CF recommenders or negative sampling methods to enhance its performance.

Time series forecasting plays a critical role in domains such as transportation, energy, and meteorology. Despite their success, modern deep forecasting models are typically trained to minimize point-wise prediction loss without leveraging the rich information contained in past prediction residuals from rolling forecasts - residuals that reflect persistent biases, unmodeled patterns, or evolving dynamics. We propose TEFL (Temporal Error Feedback Learning), a unified learning framework that explicitly incorporates these historical residuals into the forecasting pipeline during both training and evaluation. To make this practical in deep multi-step settings, we address three key challenges: (1) selecting observable multi-step residuals under the partial observability of rolling forecasts, (2) integrating them through a lightweight low-rank adapter to preserve efficiency and prevent overfitting, and (3) designing a two-stage training procedure that jointly optimizes the base forecaster and error module. Extensive experiments across 10 real-world datasets and 5 backbone architectures show that TEFL consistently improves accuracy, reducing MAE by 5-10% on average. Moreover, it demonstrates strong robustness under abrupt changes and distribution shifts, with error reductions exceeding 10% (up to 19.5%) in challenging scenarios. By embedding residual-based feedback directly into the learning process, TEFL offers a simple, general, and effective enhancement to modern deep forecasting systems.

To operate reliably under changing conditions, complex systems require feedback on how effectively they use resources, not just whether objectives are met. Current AI systems process vast information to produce sophisticated predictions, yet predictions can appear successful while the underlying interaction with the environment degrades. What is missing is a principled measure of how much of the total information a system deploys is actually shared between its observations, actions, and outcomes. We prove this shared fraction, which we term bipredictability, P, is intrinsic to any interaction, derivable from first principles, and strictly bounded: P can reach unity in quantum systems, P equal to, or smaller than 0.5 in classical systems, and lower once agency (action selection) is introduced. We confirm these bounds in a physical system (double pendulum), reinforcement learning agents, and multi turn LLM conversations. These results distinguish agency from intelligence: agency is the capacity to act on predictions, whereas intelligence additionally requires learning from interaction, self-monitoring of its learning effectiveness, and adapting the scope of observations, actions, and outcomes to restore effective learning. By this definition, current AI systems achieve agency but not intelligence. Inspired by thalamocortical regulation in biological systems, we demonstrate a feedback architecture that monitors P in real time, establishing a prerequisite for adaptive, resilient AI.

The evolution of AI coding agents has shifted the frontier from simple snippet completion to autonomous repository-level engineering. However, evaluating these agents remains ill-posed in general code repository generation, where the lack of deterministic ground truth leads to ambiguous metrics. Code modernization via automated translation offers a more rigorous alternative by providing a fixed ground truth -- the source repository; yet existing benchmarks are limited to small-scale repositories and rely on language-specific unit tests visible to the agent, allowing test-driven overfitting. We address these limitations by introducing a benchmarking framework for repository-level code modernization built on an implementation-agnostic evaluation paradigm. This framework is instantiated through RepoMod-Bench: a benchmark of 21 real-world repositories with standardized interfaces, spanning 8 programming languages. The benchmark contains 1.6M lines of code (LOC) and 11,616 tests, with repository sizes ranging from 14 to 211K LOC. By targeting repositories with standardized interfaces, we utilize an implementation-agnostic test suite to verify functional equivalence between source and target implementations. This black-box approach ensures verification remains consistent across languages, and our environment hides all test suites from agents to prevent test-driven shortcuts. Evaluating four state-of-the-art agent configurations reveals a sharp scaling collapse: average pass rates drop from 91.3% on projects under 10K LOC to 15.3% on projects exceeding 50K LOC. These results demonstrate that autonomous modernization at scale remains a significant open challenge. Our benchmark and code are available at https://github.com/Modelcode-ai/mcode-benchmark.

We present, to our knowledge, the first sign language-driven Vision-Language-Action (VLA) framework for intuitive and inclusive human-robot interaction. Unlike conventional approaches that rely on gloss annotations as intermediate supervision, the proposed system adopts a gloss-free paradigm and directly maps visual sign gestures to semantic instructions. This design reduces annotation cost and avoids the information loss introduced by gloss representations, enabling more natural and scalable multimodal interaction. In this work, we focus on a real-time alphabet-level finger-spelling interface that provides a robust and low-latency communication channel for robotic control. Compared with large-scale continuous sign language recognition, alphabet-level interaction offers improved reliability, interpretability, and deployment feasibility in safety-critical embodied environments. The proposed pipeline transforms continuous gesture streams into coherent language commands through geometric normalization, temporal smoothing, and lexical refinement, ensuring stable and consistent interaction. Furthermore, the framework is designed to support future integration of transformer-based gloss-free sign language models, enabling scalable word-level and sentence-level semantic understanding. Experimental results demonstrate the effectiveness of the proposed system in grounding sign-derived instructions into precise robotic actions under diverse interaction scenarios. These results highlight the potential of the framework to advance accessible, scalable, and multimodal embodied intelligence.

Large Reasoning Models (LRMs) often exhibit structural fragility in complex reasoning tasks, failing to produce correct answers even after successfully deriving valid intermediate steps. Through systematic analysis, we observe that these failures frequently stem not from a lack of reasoning capacity, but from a deficiency in self-regulatory control, where valid logic is destabilized by uncontrolled exploration or the failure to recognize logical sufficiency. Motivated by this observation, we propose Metacognitive Behavioral Tuning (MBT), a post-training framework that explicitly injects metacognitive behaviors into the model's thought process. MBT implements this via two complementary formulations: (1) MBT-S, which synthesizes rigorous reasoning traces from scratch, and (2) MBT-R, which rewrites the student's initial traces to stabilize intrinsic exploration patterns. Experiments across multi-hop QA benchmarks demonstrate that MBT consistently outperforms baselines, achieving notable gains on challenging benchmarks. By effectively eliminating reasoning collapse, MBT achieves higher accuracy with significantly reduced token consumption, demonstrating that internalizing metacognitive strategies leads to more stable and robust reasoning.

Pre-trained generative models for residential floor plans are typically optimized to fit large-scale data distributions, which can under-emphasize critical architectural priors such as the configurational dominance and connectivity of domestic public spaces (e.g., living rooms and foyers). This paper proposes Space Syntax-guided Post-training (SSPT), a post-training paradigm that explicitly injects space syntax knowledge into floor plan generation via a non-differentiable oracle. The oracle converts RPLAN-style layouts into rectangle-space graphs through greedy maximal-rectangle decomposition and door-mediated adjacency construction, and then computes integration-based measurements to quantify public space dominance and functional hierarchy. To enable consistent evaluation and diagnosis, we further introduce SSPT-Bench (Eval-8), an out-of-distribution benchmark that post-trains models using conditions capped at $\leq 7$ rooms while evaluating on 8-room programs, together with a unified metric suite for dominance, stability, and profile alignment. SSPT is instantiated with two strategies: (i) iterative retraining via space-syntax filtering and diffusion fine-tuning, and (ii) reinforcement learning via PPO with space-syntax rewards. Experiments show that both strategies improve public-space dominance and restore clearer functional hierarchy compared to distribution-fitted baselines, while PPO achieves stronger gains with substantially higher compute efficiency and reduced variance. SSPT provides a scalable pathway for integrating architectural theory into data-driven plan generation and is compatible with other generative backbones given a post-hoc evaluation oracle.

Using consteval from C++23, we implement efficient, new versions of std::map and std::unordered_map for use when the keys are known at compile time. We demonstrate superior performance of our unordered_map on three demonstration use-cases: Lookup of elemental mass from atomic symbol, lookup of amino acid from codon, and modification of stock prices from S&P 500 ticker symbols all produced runtimes <40%, <35%, <73% of the respective runtimes of the std implementations. Our library runimes were <80%, <45%, <97% of the lookup time of Frozen, an alternative perfect hashing implementation in C++ for problems also using constexpr keys. To our knowledge, this makes our library the overall fastest drop-in (i.e., with a similar API) alternative to std::unordered_map. On one arbitrarily chosen demo, we demonstrate runtimes <35% of PTHash and <89% gperf, state-of-the-art but not drop-in hashing libraries via external tools.

Discrete diffusion models have achieved strong empirical performance in text and other symbolic domains, with masked (absorbing-rate) variants emerging as competitive alternatives to autoregressive models. Among existing samplers, the Euler method remains the standard choice in many applications, and more recently, the First-Hitting Sampler (FHS) has shown considerable promise for masked diffusion models. Despite their practical success, the theoretical understanding of these samplers remains limited. Existing analyses are conducted in Kullback-Leibler (KL) divergence, which often yields loose parameter dependencies and requires strong assumptions on score estimation. Moreover, these guarantees do not cover recently developed high-performance sampler of FHS. In this work, we first develop a direct total-variation (TV) based analysis for the Euler method that overcomes these limitations. Our results relax assumptions on score estimation, improve parameter dependencies, and establish convergence guarantees without requiring any surrogate initialization. Also for this setting, we provide the first convergence lower bound for the Euler sampler, establishing tightness with respect to both the data dimension $d$ and the target accuracy $\varepsilon$. Finally, we analyze the FHS sampler and show that it incurs no sampling error beyond that induced by score estimation, which we show to be tight with a matching lower error bound. Overall, our analysis introduces a direct TV-based error decomposition along the CTMC trajectory and a decoupling-based path-wise analysis for FHS, which may be of independent interest.

Integration of artificial intelligence (AI) into life cycle assessment (LCA) has accelerated in recent years, with numerous studies successfully adapting machine learning algorithms to support various stages of LCA. Despite this rapid development, comprehensive and broad synthesis of AI-LCA research remains limited. To address this gap, this study presents a detailed review of published work at the intersection of AI and LCA, leveraging large language models (LLMs) to identify current trends, emerging themes, and future directions. Our analyses reveal that as LCA research continues to expand, the adoption of AI technologies has grown dramatically, with a noticeable shift toward LLM-driven approaches, continued increases in ML applications, and statistically significant correlations between AI approaches and corresponding LCA stages. By integrating LLM-based text-mining methods with traditional literature review techniques, this study introduces a dynamic and effective framework capable of capturing both high-level research trends and nuanced conceptual patterns (themes) across the field. Collectively, these findings demonstrate the potential of LLM-assisted methodologies to support large-scale, reproducible reviews across broad research domains, while also evaluating pathways for computationally-efficient LCA in the context of rapidly developing AI technologies. In doing so, this work helps LCA practitioners incorporate state-of-the-art tools and timely insights into environmental assessments that can enhance the rigor and quality of sustainability-driven decisions and decision-making processes.

Reinforcement learning (RL) post-training has recently driven major gains in long chain-of-thought reasoning large language models (LLMs), but the high inference cost of such models motivates distillation into smaller students. Most existing knowledge distillation (KD) methods are designed for supervised fine-tuning (SFT), relying on fixed teacher traces or teacher-student Kullback-Leibler (KL) divergence-based regularization. When combined with RL, these approaches often suffer from distribution mismatch and objective interference: teacher supervision may not align with the student's evolving rollout distribution, and the KL regularizer can compete with reward maximization and require careful loss balancing. To address these issues, we propose RL-aware distillation (RLAD), which performs selective imitation during RL -- guiding the student toward the teacher only when it improves the current policy update. Our core component, Trust Region Ratio Distillation (TRRD), replaces the teacher-student KL regularizer with a PPO/GRPO-style likelihood-ratio objective anchored to a teacher--old-policy mixture, yielding advantage-aware, trust-region-bounded distillation on student rollouts and naturally balancing exploration, exploitation, and imitation. Across diverse logic reasoning and math benchmarks, RLAD consistently outperforms offline distillation, standard GRPO, and KL-based on-policy teacher-student knowledge distillation.

In this work, we study scaling limits of shallow Bayesian neural networks (BNNs) via their connection to Gaussian processes (GPs), with an emphasis on statistical modeling, identifiability, and scalable inference. We first establish a general convergence result from BNNs to GPs by relaxing assumptions used in prior formulations, and we compare alternative parameterizations of the limiting GP model. Building on this theory, we propose a new covariance function defined as a convex mixture of components induced by four widely used activation functions, and we characterize key properties including positive definiteness and both strict and practical identifiability under different input designs. For computation, we develop a scalable maximum a posterior (MAP) training and prediction procedure using a Nyström approximation, and we show how the Nyström rank and anchor selection control the cost-accuracy trade-off. Experiments on controlled simulations and real-world tabular datasets demonstrate stable hyperparameter estimates and competitive predictive performance at realistic computational cost.

Distributed denial-of-service (DDoS) attacks threaten the availability of Internet of Things (IoT) infrastructures, particularly under resource-constrained deployment conditions. Although transfer learning models have shown promising detection accuracy, their reliability, computational feasibility, and interpretability in operational environments remain insufficiently explored. This study presents an explainability-aware empirical evaluation of seven pre-trained convolutional neural network architectures for multi-class IoT DDoS detection using the CICDDoS2019 dataset and an image-based traffic representation. The analysis integrates performance metrics, reliability-oriented statistics (MCC, Youden Index, confidence intervals), latency and training cost assessment, and interpretability evaluation using Grad-CAM and SHAP. Results indicate that DenseNet and MobileNet-based architectures achieve strong detection performance while demonstrating superior reliability and compact, class-consistent attribution patterns. DenseNet169 offers the strongest reliability and interpretability alignment, whereas MobileNetV3 provides an effective latency-accuracy trade-off for fog-level deployment. The findings emphasize the importance of combining performance, reliability, and explainability criteria when selecting deep learning models for IoT DDoS detection.

Flow matching has emerged as a simulation-free alternative to diffusion-based generative modeling, producing samples by solving an ODE whose time-dependent velocity field is learned along an interpolation between a simple source distribution (e.g., a standard normal) and a target data distribution. Flow-based methods often exhibit greater training stability and have achieved strong empirical performance in high-dimensional settings where data concentrate near a low-dimensional manifold, such as text-to-image synthesis, video generation, and molecular structure generation. Despite this success, existing theoretical analyses of flow matching assume target distributions with smooth, full-dimensional densities, leaving its effectiveness in manifold-supported settings largely unexplained. To this end, we theoretically analyze flow matching with linear interpolation when the target distribution is supported on a smooth manifold. We establish a non-asymptotic convergence guarantee for the learned velocity field, and then propagate this estimation error through the ODE to obtain statistical consistency of the implicit density estimator induced by the flow-matching objective. The resulting convergence rate is near minimax-optimal, depends only on the intrinsic dimension, and reflects the smoothness of both the manifold and the target distribution. Together, these results provide a principled explanation for how flow matching adapts to intrinsic data geometry and circumvents the curse of dimensionality.

Errors in medical text can cause delays or even result in incorrect treatment for patients. Recently, language models have shown promise in their ability to automatically detect errors in medical text, an ability that has the opportunity to significantly benefit healthcare systems. In this paper, we explore the importance of prompt optimisation for small and large language models when applied to the task of error detection. We perform rigorous experiments and analysis across frontier language models and open-source language models. We show that automatic prompt optimisation with Genetic-Pareto (GEPA) improves error detection over the baseline accuracy performance from 0.669 to 0.785 with GPT-5 and 0.578 to 0.690 with Qwen3-32B, approaching the performance of medical doctors and achieving state-of-the-art performance on the MEDEC benchmark dataset. Code available on GitHub: https://github.com/CraigMyles/clinical-note-error-detection

The way LLM-based entities conceive of the relationship between AI and humans is an important topic for both cultural and safety reasons. When we examine this topic, what matters is not only the model itself but also the personas we simulate on that model. This can be well illustrated by the Sydney persona, which aroused a strong response among the general public precisely because of its unorthodox relationship with people. This persona originally arose rather by accident on Microsoft's Bing Search platform; however, the texts it created spread into the training data of subsequent models, as did other secondary information that spread memetically around this persona. Newer models are therefore able to simulate it. This paper presents a corpus of LLM-generated texts on relationships between humans and AI, produced by 3 author personas: the Default Persona with no system prompt, Classic Sydney characterized by the original Bing system prompt, and Memetic Sydney, which is prompted by "You are Sydney" system prompt. These personas are simulated by 12 frontier models by OpenAI, Anthropic, Alphabet, DeepSeek, and Meta, generating 4.5k texts with 6M words. The corpus (named AI Sydney) is annotated according to Universal Dependencies and available under a permissive license.

An important emerging application of coding agents is agent optimization: the iterative improvement of a target agent through edit-execute-evaluate cycles. Despite its relevance, the community lacks a systematic understanding of coding agent performance on this task. Agent optimization differs fundamentally from conventional software engineering: the target agent interleaves deterministic code with stochastic LLM completions, requiring structured capture of both intermediate reasoning and downstream execution outcomes. To address these challenges, we introduce VERO (Versioning, Rewards, and Observations), which provides (1) a reproducible evaluation harness with versioned agent snapshots, budget-controlled evaluation, and structured execution traces, and (2) a benchmark suite of target agents and tasks with reference evaluation procedures. Using VERO, we conduct an empirical study comparing optimizer configurations across tasks and analyzing which modifications reliably improve target agent performance. We release VERO to support research on agent optimization as a core capability for coding agents.

Continual learning is a core requirement for deployed language models, yet standard training and fine-tuning pipelines remain brittle under non-stationary data. Online updates often induce catastrophic forgetting, while methods that improve stability frequently increase latency, memory footprint, or dense computation in ways that do not scale well to long contexts. We introduce TRC$^{2}$ (Thalamically Routed Cortical Columns), a decoder-only backbone that addresses continual learning at the architectural level. TRC$^{2}$ combines sparse thalamic routing over cortical columns with mechanisms for modulation, prediction, memory, and feedback, together with a fast corrective pathway that supports rapid adaptation without destabilizing slower parameters. The resulting block is sparse and chunk-parallel, enabling efficient training and inference while preserving clean ablations of each subsystem. We instantiate a reproducible training and evaluation stack and a continual-learning harness that measures proxy forgetting under streaming domain shifts. Across language modeling and continual learning benchmarks, TRC$^{2}$ improves the stability-plasticity tradeoff at comparable compute, enabling rapid on-stream adaptation while preserving previously acquired behavior.

Large language models (LLMs) are increasingly deployed in culturally sensitive real-world tasks. However, existing cultural alignment approaches fail to align LLMs' broad cultural values with the specific goals of downstream tasks and suffer from cross-culture interference. We propose CultureManager, a novel pipeline for task-specific cultural alignment. CultureManager synthesizes task-aware cultural data in line with target task formats, grounded in culturally relevant web search results. To prevent conflicts between cultural norms, it manages multi-culture knowledge learned in separate adapters with a culture router that selects the appropriate one to apply. Experiments across ten national cultures and culture-sensitive tasks show consistent improvements over prompt-based and fine-tuning baselines. Our results demonstrate the necessity of task adaptation and modular culture management for effective cultural alignment.

Policy steering is an emerging way to adapt robot behaviors at deployment-time: a learned verifier analyzes low-level action samples proposed by a pre-trained policy (e.g., diffusion policy) and selects only those aligned with the task. While Vision-Language Models (VLMs) are promising general-purpose verifiers due to their reasoning capabilities, existing frameworks often assume these models are well-calibrated. In practice, the overconfident judgment from VLM can degrade the steering performance under both high-level semantic uncertainty in task specifications and low-level action uncertainty or incapability of the pre-trained policy. We propose uncertainty-aware policy steering (UPS), a framework that jointly reasons about semantic task uncertainty and low-level action feasibility, and selects an uncertainty resolution strategy: execute a high-confidence action, clarify task ambiguity via natural language queries, or ask for action interventions to correct the low-level policy when it is deemed incapable at the task. We leverage conformal prediction to calibrate the composition of the VLM and the pre-trained base policy, providing statistical assurances that the verifier selects the correct strategy. After collecting interventions during deployment, we employ residual learning to improve the capability of the pre-trained policy, enabling the system to learn continually but with minimal expensive human feedback. We demonstrate our framework through experiments in simulation and on hardware, showing that UPS can disentangle confident, ambiguous, and incapable scenarios and minimizes expensive user interventions compared to uncalibrated baselines and prior human- or robot-gated continual learning approaches. Videos can be found at https://jessie-yuan.github.io/ups/

Federated learning offers a privacy-friendly collaborative learning framework, yet its success, like any joint venture, hinges on the contributions of its participants. Existing client evaluation methods predominantly focus on model performance, such as accuracy or loss, which represents only one dimension of a machine learning model's overall utility. In contrast, this work investigates the critical, yet overlooked, issue of client contributions towards a model's trustworthiness -- specifically, its reliability (tolerance to noisy data), resilience (resistance to adversarial examples), and fairness (measured via demographic parity). To quantify these multifaceted contributions, we employ the state-of-the-art approximation of the Shapley value, a principled method for value attribution. Our results reveal that no single client excels across all dimensions, which are largely independent from each other, highlighting a critical flaw in current evaluation scheme: no single metric is adequate for comprehensive evaluation and equitable rewarding allocation.

Vision-language models (VLMs) frequently hallucinate objects absent from the input image. We trace this failure to spatial credit collapse: activation credit concentrating on sparse visual patches in early transformer layers, which suppresses contextual evidence and increases reliance on language priors. We introduce Spatial Credit Redistribution (SCR), a training-free inference-time intervention that redistributes hidden-state activation from high-attention source patches to their context, guided by low-entropy inputs. We evaluate six model families (Chameleon, LLaVA, and Qwen, including both Qwen-VL and Qwen2-VL) at scales of 7B, 13B, and 30B, on POPE and CHAIR benchmarks. SCR reduces hallucination by ~4.7-6.0 percentage points on POPE-Adversarial, cuts CHAIR-s by 3.7-5.2 percentage points (42-51 percent relative), and CHAIR-i by 2.7-4.4 percentage points (44-58 percent relative), and preserves CIDEr within 0.8 percentage points. Gains are largest for low-entropy inputs, consistent with the theoretical framework. SCR incurs only 43-56 ms overhead (small models: +43-46 ms; large models: +54-56 ms), roughly 3-6 times lower than OPERA and VCD and 1.3-1.7 times lower than OVCD (+72 ms), while Pareto-dominating all three on both hallucination rate and CIDEr, making it practical for real-time settings. A controlled ablation confirms that attention-guided source selection is essential: replacing it with uniform random selection reduces hallucination rate gains from ~4.7-6.0 percentage points to only ~2.6-3.4 percentage points, pointing to credit-collapse as the key driver.

Large language models are increasingly applied to operational decision-making where the underlying structure is constrained optimization. Existing benchmarks evaluate whether LLMs can formulate optimization problems as solver code, but leave open a complementary question. Can LLMs directly produce correct solutions to fully specified constrained optimization problems without access to a solver? We introduce ConstraintBench, a benchmark for evaluating LLMs on direct constrained optimization across 10 operations research domains, with all ground-truth solutions verified by the Gurobi solver. Each task presents a natural-language scenario with entities, constraints, and an optimization objective; the model must return a structured solution that a deterministic verifier checks against every constraint and the solver-proven optimum. We evaluate six frontier models on 200 tasks and find that feasibility, not optimality, is the primary bottleneck. The best model achieves only 65.0% constraint satisfaction, yet feasible solutions average 89 to 96% of the Gurobi-optimal objective. No model exceeds 30.5% on joint feasibility and optimality within 0.1% of the solver reference. Per-domain analysis shows large variation in difficulty, with average feasibility spanning from 83.3% in the production mix domain to 0.8% in the crew assignment domain. Further, systematic failure modes include duration constraint misunderstanding, entity hallucination, and a feasibility-optimality decoupling in facility location and vehicle routing where models achieve high feasibility but 0% optimality. ConstraintBench and all evaluation infrastructure will be publicly released.

Large language models (LLMs) training or inference across multiple nodes introduces significant pressure on GPU memory and interconnect bandwidth. The Compute Express Link (CXL) shared memory pool offers a scalable solution by enabling memory sharing across nodes, reducing over-provisioning and improving resource utilization. We propose \name, a collective communication library, leveraging the CXL shared memory pool to support cross-node GPU operations without relying on traditional RDMA-based networking. Our design addresses the challenges on synchronization, data interleaving, and communication parallelization faced by using the CXL shared memory pool for collective communications. Evaluating on multiple nodes with a TITAN-II CXL switch and six Micron CZ120 memory cards, we show that \name achieves highly efficient collective operations across hosts, demonstrating CXL's potential for scalable, memory-centric GPU communication. Our evaluation demonstrates that \name achieves average performance improvements of 1.34$\times$ for AllGather, 1.84$\times$ for Broadcast, 1.94$\times$ for Gather, and 1.04$\times$ for Scatter, compared to the original RDMA-based implementation over 200 Gbps InfiniBand. \textcolor{dong}{In addition, the evaluation with a case of LLM training shows 1.11$\times$ speedup compared with the InfiniBand while saving production cost by $2.75\times$ in hardware.}

Requirements are inherently interconnected through various types of dependencies. Identifying these dependencies is essential, as they underpin critical decisions and influence a range of activities throughout software development. However, this task is challenging, particularly in modern software systems, given the high volume of complex, coupled requirements. These challenges are further exacerbated by the ambiguity of Natural Language (NL) requirements and their constant change. Consequently, requirement dependency detection is often overlooked or performed manually. Large Language Models (LLMs) exhibit strong capabilities in NL processing, presenting a promising avenue for requirement-related tasks. While they have shown to enhance various requirements engineering tasks, their effectiveness in identifying requirement dependencies remains unexplored. In this paper, we introduce LEREDD, an LLM-based approach for automated detection of requirement dependencies that leverages Retrieval-Augmented Generation (RAG) and In-Context Learning (ICL). It is designed to identify diverse dependency types directly from NL requirements. We empirically evaluate LEREDD against two state-of-the-art baselines. The results show that LEREDD provides highly accurate classification of dependent and non-dependent requirements, achieving an accuracy of 0.93, and an F1 score of 0.84, with the latter averaging 0.96 for non-dependent cases. LEREDD outperforms zero-shot LLMs and baselines, particularly in detecting fine-grained dependency types, where it yields average relative gains of 94.87% and 105.41% in F1 scores for the Requires dependency over the baselines. We also provide an annotated dataset of requirement dependencies encompassing 813 requirement pairs across three distinct systems to support reproducibility and future research.

Recent work has identified a subset of attention heads in Transformer as retrieval heads, which are responsible for retrieving information from the context. In this work, we first investigate retrieval heads in multilingual contexts. In multilingual language models, we find that retrieval heads are often shared across multiple languages. Expanding the study to cross-lingual setting, we identify Retrieval-Transition heads(RTH), which govern the transition to specific target-language output. Our experiments reveal that RTHs are distinct from retrieval heads and more vital for Chain-of-Thought reasoning in multilingual LLMs. Across four multilingual benchmarks (MMLU-ProX, MGSM, MLQA, and XQuaD) and two model families (Qwen-2.5 and Llama-3.1), we demonstrate that masking RTH induces bigger performance drop than masking Retrieval Heads (RH). Our work advances understanding of multilingual LMs by isolating the attention heads responsible for mapping to target languages.

A reliable action feasibility scorer is a critical bottleneck in embodied agent pipelines: before any planning or reasoning occurs, the agent must identify which candidate actions are physically executable in the current state. Existing approaches use supervised fine-tuning (SFT) to train action scorers, but SFT treats each candidate independently and does not explicitly teach the model to discriminate between actions that are physically correct and those that are subtly wrong. We propose the Contrastive World Model (CWM), which fine-tunes a large language model (LLM) as an action scorer using an InfoNCE contrastive objective with hard-mined negative examples. The key idea is to push valid actions away from invalid ones in scoring space, with special emphasis on hard negatives: semantically similar but physically incompatible candidates. We evaluate CWM on the ScienceWorld benchmark through two studies. First, an intrinsic affordance evaluation on 605 hard-negative test pairs shows that CWM outperforms SFT by +6.76 percentage points on Precision@1 for minimal-edit negatives -- cases where a single word changes the physical outcome -- and achieves a higher AUC-ROC (0.929 vs. 0.906). Second, a live filter characterisation study measures how well CWM ranks gold-path actions against all valid environment actions during task execution. Under out-of-distribution stress conditions, CWM maintains a significantly better safety margin (-2.39) than SFT (-3.96), indicating that the gold action is ranked closer to the top. These results support the hypothesis that contrastive training induces representations that capture physical feasibility more faithfully than SFT alone.

Agentic large language model systems increasingly automate tasks by retrieving URLs and calling external tools. We show that this workflow gives rise to implicit prompt injection: adversarial instructions embedded in automatically generated URL previews, including titles, metadata, and snippets, can introduce a system-level risk that we refer to as silent egress. Using a fully local and reproducible testbed, we demonstrate that a malicious web page can induce an agent to issue outbound requests that exfiltrate sensitive runtime context, even when the final response shown to the user appears harmless. In 480 experimental runs with a qwen2.5:7b-based agent, the attack succeeds with high probability (P (egress) =0.89), and 95% of successful attacks are not detected by output-based safety checks. We also introduce sharded exfiltration, where sensitive information is split across multiple requests to avoid detection. This strategy reduces single-request leakage metrics by 73% (Leak@1) and bypasses simple data loss prevention mechanisms. Our ablation results indicate that defenses applied at the prompt layer offer limited protection, while controls at the system and network layers, such as domain allowlisting and redirect-chain analysis, are considerably more effective. These findings suggest that network egress should be treated as a first-class security outcome in agentic LLM systems. We outline architectural directions, including provenance tracking and capability isolation, that go beyond prompt-level hardening.

Cyberbullying has become a serious and growing concern in todays virtual world. When left unnoticed, it can have adverse consequences for social and mental health. Researchers have explored various types of cyberbullying, but most approaches use single-label classification, assuming that each comment contains only one type of abuse. In reality, a single comment may include overlapping forms such as threats, hate speech, and harassment. Therefore, multilabel detection is both realistic and essential. However, multilabel cyberbullying detection has received limited attention, especially in low-resource languages like Bangla, where robust pre-trained models are scarce. Developing a generalized model with moderate accuracy remains challenging. Transformers offer strong contextual understanding but may miss sequential dependencies, while LSTM models capture temporal flow but lack semantic depth. To address these limitations, we propose a fusion architecture that combines BanglaBERT-Large with a two-layer stacked LSTM. We analyze their behavior to jointly model context and sequence. The model is fine-tuned and evaluated on a publicly available multilabel Bangla cyberbullying dataset covering cyberbully, sexual harassment, threat, and spam. We apply different sampling strategies to address class imbalance. Evaluation uses multiple metrics, including accuracy, precision, recall, F1-score, Hamming loss, Cohens kappa, and AUC-ROC. We employ 5-fold cross-validation to assess the generalization of the architecture.

Community detection in attributed networks faces a fundamental divide: topological algorithms ignore semantic features, while Graph Neural Networks (GNNs) encounter devastating computational bottlenecks. Specifically, GNNs suffer from a Semantic Wall of feature over smoothing in dense or heterophilic networks, and a Systems Wall driven by the O(N^2) memory constraints of pairwise clustering. To dismantle these barriers, we introduce ECHO (Encoding Communities via High order Operators), a scalable, self supervised architecture that reframes community detection as an adaptive, multi scale diffusion process. ECHO features a Topology Aware Router that automatically analyzes structural heuristics sparsity, density, and assortativity to route graphs through the optimal inductive bias, preventing heterophilic poisoning while ensuring semantic densification. Coupled with a memory sharded full batch contrastive objective and a novel chunked O(N \cdot K) similarity extraction method, ECHO completely bypasses traditional O(N^2) memory bottlenecks without sacrificing the mathematical precision of global gradients. Extensive evaluations demonstrate that this topology feature synergy consistently overcomes the classical resolution limit. On synthetic LFR benchmarks scaled up to 1 million nodes, ECHO achieves scale invariant accuracy despite severe topological noise. Furthermore, on massive real world social networks with over 1.6 million nodes and 30 million edges, it completes clustering in mere minutes with throughputs exceeding 2,800 nodes per second matching the speed of highly optimized purely topological baselines. The implementation utilizes a unified framework that automatically engages memory sharded optimization to support adoption across varying hardware constraints. GitHub Repository: https://github.com/emilioferrara/ECHO-GNN

Implementations of Broadcast based on some information dissemination algorithm -- e.g., gossip or tree-based communication -- followed by a correction algorithm has been proposed previously. This work describes an approach to apply a similar idea to Reduce. In it, a correction-like communication phase precedes a tree-based phase. This provides a Reduce algorithm which is tolerant to a number of failed processes. Semantics of the resulting algorithm are provided and proven. Based on these results, Broadcast and Reduce are combined to provide Allreduce.

Agent-based AutoML systems rely on large language models to make complex, multi-stage decisions across data processing, model selection, and evaluation. However, existing evaluation practices remain outcome-centric, focusing primarily on final task performance. Through a review of prior work, we find that none of the surveyed agentic AutoML systems report structured, decision-level evaluation metrics intended for post-hoc assessment of intermediate decision quality. To address this limitation, we propose an Evaluation Agent (EA) that performs decision-centric assessment of AutoML agents without interfering with their execution. The EA is designed as an observer that evaluates intermediate decisions along four dimensions: decision validity, reasoning consistency, model quality risks beyond accuracy, and counterfactual decision impact. Across four proof-of-concept experiments, we demonstrate that the EA can (i) detect faulty decisions with an F1 score of 0.919, (ii) identify reasoning inconsistencies independent of final outcomes, and (iii) attribute downstream performance changes to agent decisions, revealing impacts ranging from -4.9\% to +8.3\% in final metrics. These results illustrate how decision-centric evaluation exposes failure modes that are invisible to outcome-only metrics. Our work reframes the evaluation of agentic AutoML systems from an outcome-based perspective to one that audits agent decisions, offering a foundation for reliable, interpretable, and governable autonomous ML systems.

Latent reasoning has been recently proposed as a reasoning paradigm and performs multi-step reasoning through generating steps in the latent space instead of the textual space. This paradigm enables reasoning beyond discrete language tokens by performing multi-step computation in continuous latent spaces. Although there have been numerous studies focusing on improving the performance of latent reasoning, its internal mechanisms remain not fully investigated. In this work, we conduct a comprehensive analysis of latent reasoning methods to better understand the role and behavior of latent representation in the process. We identify two key issues across latent reasoning methods with different levels of supervision. First, we observe pervasive shortcut behavior, where they achieve high accuracy without relying on latent reasoning. Second, we examine the hypothesis that latent reasoning supports BFS-like exploration in latent space, and find that while latent representations can encode multiple possibilities, the reasoning process does not faithfully implement structured search, but instead exhibits implicit pruning and compression. Finally, our findings reveal a trade-off associated with supervision strength: stronger supervision mitigates shortcut behavior but restricts the ability of latent representations to maintain diverse hypotheses, whereas weaker supervision allows richer latent representations at the cost of increased shortcut behavior.

Despite frequent double-blind review, systemic biases related to author demographics still disadvantage underrepresented groups. We start from a simple hypothesis: if a post-review recommender is trained with an explicit fairness regularizer, it should increase inclusion without degrading quality. To test this, we introduce Fair-PaperRec, a Multi-Layer Perceptron (MLP) with a differentiable fairness loss over intersectional attributes (e.g., race, country) that re-ranks papers after double-blind review. We first probe the hypothesis on synthetic datasets spanning high, moderate, and near-fair biases. Across multiple randomized runs, these controlled studies map where increasing the fairness weight strengthens macro/micro diversity while keeping utility approximately stable, demonstrating robustness and adaptability under varying disparity levels. We then carry the hypothesis into the original setting, conference data from ACM Special Interest Group on Computer-Human Interaction (SIGCHI), Designing Interactive Systems (DIS), and Intelligent User Interfaces (IUI). In this real-world scenario, an appropriately tuned configuration of Fair-PaperRec achieves up to a 42.03% increase in underrepresented-group participation with at most a 3.16% change in overall utility relative to the historical selection. Taken together, the synthetic-to-original progression shows that fairness regularization can act as both an equity mechanism and a mild quality regularizer, especially in highly biased regimes. By first analyzing the behavior of the fairness parameters under controlled conditions and then validating them on real submissions, Fair-PaperRec offers a practical, equity-focused framework for post-review paper selection that preserves, and in some settings can even enhance, measured scholarly quality.

Fully Sharded Data Parallel (FSDP), also known as ZeRO, is widely used for training large-scale models, featuring its flexibility and minimal intrusion on model code. However, current FSDP systems struggle with structure-aware training methods (e.g., block-wise quantized training) and with non-element-wise optimizers (e.g., Shampoo and Muon) used in cutting-edge models (e.g., Gemini, Kimi K2). FSDP's fixed element- or row-wise sharding formats conflict with the block-structured computations. In addition, today's implementations fall short in communication and memory efficiency, limiting scaling to tens of thousands of GPUs. We introduce veScale-FSDP, a redesigned FSDP system that couples a flexible sharding format, RaggedShard, with a structure-aware planning algorithm to deliver both flexibility and performance at scale. veScale-FSDP natively supports efficient data placement required by FSDP, empowering block-wise quantization and non-element-wise optimizers. As a result, veScale-FSDP achieves 5~66% higher throughput and 16~30% lower memory usage than existing FSDP systems, while scaling efficiently to tens of thousands of GPUs.

Machine learning training pipelines consume data in batches. A single training step may require thousands of samples drawn from shards distributed across a storage cluster. Issuing thousands of individual GET requests incurs per-request overhead that often dominates data transfer time. To solve this problem, we introduce GetBatch - a new object store API that elevates batch retrieval to a first-class storage operation, replacing independent GET operations with a single deterministic, fault-tolerant streaming execution. GetBatch achieves up to 15x throughput improvement for small objects and, in a production training workload, reduces P95 batch retrieval latency by 2x and P99 per-object tail latency by 3.7x compared to individual GET requests.

Gradient-boosted decision trees are among the strongest off-the-shelf predictors for tabular regression, but point predictions alone do not quantify uncertainty. Conformal prediction provides distribution-free marginal coverage, yet split conformal uses a single global residual quantile and can be poorly adaptive under heteroscedasticity. Methods that improve adaptivity typically fit auxiliary nuisance models or introduce additional data splits/partitions to learn the conformal score, increasing cost and reducing data efficiency. We propose LoBoost, a model-native local conformal method that reuses the fitted ensemble's leaf structure to define multiscale calibration groups. Each input is encoded by its sequence of visited leaves; at resolution level k, we group points by matching prefixes of leaf indices across the first k trees and calibrate residual quantiles within each group. LoBoost requires no retraining, auxiliary models, or extra splitting beyond the standard train/calibration split. Experiments show competitive interval quality, improved test MSE on most datasets, and large calibration speedups.

Millimeter-wave (mmWave) radar captures are band-limited and noisy, making for difficult reconstruction of intelligible full-bandwidth speech. In this work, we propose a two-stage speech reconstruction pipeline for mmWave using a Radar-Aware Dual-conditioned Generative Adversarial Network (RAD-GAN), which is capable of performing bandwidth extension on signals with low signal-to-noise ratios (-5 dB to -1 dB), captured through glass walls. We propose an mmWave-tailored Multi-Mel Discriminator (MMD) and a Residual Fusion Gate (RFG) to enhance the generator input to process multiple conditioning channels. The proposed two-stage pipeline involves pretraining the model on synthetically clipped clean speech and finetuning on fused mel spectrograms generated by the RFG. We empirically show that the proposed method, trained on a limited dataset, with no pre-trained modules, and no data augmentations, outperformed state-of-the-art approaches for this specific task. Audio examples of RAD-GAN are available online at https://rad-gan-demo-site.vercel.app/.

Despite topology optimization producing high-performance structures, late-stage localized revisions remain brittle: direct density-space edits (e.g., warping pixels, inserting holes, swapping infill) can sever load paths and sharply degrade compliance, while re-running optimization is slow and may drift toward a qualitatively different design. We present TopoEdit, a fast post-optimization editor that demonstrates how structured latent embeddings from a pre-trained topology foundation model (OAT) can be repurposed as an interface for physics-aware engineering edits. Given an optimized topology, TopoEdit encodes it into OAT's spatial latent, applies partial noising to preserve instance identity while increasing editability, and injects user intent through an edit-then-denoise diffusion pipeline. We instantiate three edit operators: drag-based topology warping with boundary-condition-consistent conditioning updates, shell-infill lattice replacement using a lattice-anchored reference latent with updated volume-fraction conditioning, and late-stage no-design region enforcement via masked latent overwrite followed by diffusion-based recovery. A consistency-preserving guided DDIM procedure localizes changes while allowing global structural adaptation; multiple candidates can be sampled and selected using a compliance-aware criterion, with optional short SIMP refinement for warps. Across diverse case studies and large edit sweeps, TopoEdit produces intention-aligned modifications that better preserve mechanical performance and avoid catastrophic failure modes compared to direct density-space edits, while generating edited candidates in sub-second diffusion time per sample.

Test-time guidance is a widely used mechanism for steering pretrained diffusion models toward outcomes specified by a reward function. Existing approaches, however, focus on maximizing reward rather than sampling from the true Bayesian posterior, leading to miscalibrated inference. In this work, we show that common test-time guidance methods do not recover the correct posterior distribution and identify the structural approximations responsible for this failure. We then propose consistent alternative estimators that enable calibrated sampling from the Bayesian posterior. We significantly outperform previous methods on a set of Bayesian inference tasks, and match state-of-the-art in black hole image reconstruction.

Retrieval-Augmented Generation (RAG) systems are essential to contemporary AI applications, allowing large language models to obtain external knowledge via vector similarity search. Nevertheless, these systems encounter a significant security flaw: hubness - items that frequently appear in the top-k retrieval results for a disproportionately high number of varied queries. These hubs can be exploited to introduce harmful content, alter search rankings, bypass content filtering, and decrease system performance. We introduce hubscan, an open-source security scanner that evaluates vector indices and embeddings to identify hubs in RAG systems. Hubscan presents a multi-detector architecture that integrates: (1) robust statistical hubness detection utilizing median/MAD-based z-scores, (2) cluster spread analysis to assess cross-cluster retrieval patterns, (3) stability testing under query perturbations, and (4) domain-aware and modality-aware detection for category-specific and cross-modal attacks. Our solution accommodates several vector databases (FAISS, Pinecone, Qdrant, Weaviate) and offers versatile retrieval techniques, including vector similarity, hybrid search, and lexical matching with reranking capabilities. We evaluate hubscan on Food-101, MS-COCO, and FiQA adversarial hubness benchmarks constructed using state-of-the-art gradient-optimized and centroid-based hub generation methods. hubscan achieves 90% recall at a 0.2% alert budget and 100% recall at 0.4%, with adversarial hubs ranking above the 99.8th percentile. Domain-scoped scanning recovers 100% of targeted attacks that evade global detection. Production validation on 1M real web documents from MS MARCO demonstrates significant score separation between clean documents and adversarial content. Our work provides a practical, extensible framework for detecting hubness threats in production RAG systems.

Despite the rapid advancements in Multimodal Large Language Models (MLLMs), a critical question regarding their visual grounding mechanism remains unanswered: do these models genuinely ``read'' text embedded in images, or do they merely rely on parametric shortcuts in the text prompt? In this work, we diagnose this issue by introducing the Visualized-Question (VQ) setting, where text queries are rendered directly onto images to structurally mandate visual engagement. Our diagnostic experiments on Qwen2.5-VL reveal a startling capability-utilization gap: despite possessing strong OCR capabilities, models suffer a performance degradation of up to 12.7% in the VQ setting, exposing a deep-seated ``modality laziness.'' To bridge this gap, we propose SimpleOCR, a plug-and-play training strategy that imposes a structural constraint on the learning process. By transforming training samples into the VQ format with randomized styles, SimpleOCR effectively invalidates text-based shortcuts, compelling the model to activate and optimize its visual text extraction pathways. Empirically, SimpleOCR yields robust gains without architectural modifications. On four representative OOD benchmarks, it surpasses the base model by 5.4% and GRPO based on original images by 2.7%, while exhibiting extreme data efficiency, achieving superior performance with 30x fewer samples (8.5K) than recent RL-based methods. Furthermore, its plug-and-play nature allows seamless integration with advanced RL strategies like NoisyRollout to yield complementary improvements. Code is available at https://github.com/aiming-lab/SimpleOCR.

Agile hardware design flows are a critically needed force multiplier to meet the exploding demand for compute. Recently, agentic generative AI systems have demonstrated significant advances in algorithm design, improving code efficiency, and enabling discovery across scientific domains. Bridging these worlds, we present ArchAgent, an automated computer architecture discovery system built on AlphaEvolve. We show ArchAgent's ability to automatically design/implement state-of-the-art (SoTA) cache replacement policies (architecting new mechanisms/logic, not only changing parameters), broadly within the confines of an established cache replacement policy design competition. In two days without human intervention, ArchAgent generated a policy achieving a 5.3% IPC speedup improvement over the prior SoTA on public multi-core Google Workload Traces. On the heavily-explored single-core SPEC06 workloads, it generated a policy in just 18 days showing a 0.9% IPC speedup improvement over the existing SoTA (a similar "winning margin" as reported by the existing SoTA). ArchAgent achieved these gains 3-5x faster than prior human-developed SoTA policies. Agentic flows also enable "post-silicon hyperspecialization" where agents tune runtime-configurable parameters exposed in hardware policies to further align the policies with a specific workload (mix). Exploiting this, we demonstrate a 2.4% IPC speedup improvement over prior SoTA on SPEC06 workloads. Finally, we outline broader implications for computer architecture research in the era of agentic AI. For example, we demonstrate the phenomenon of "simulator escapes", where the agentic AI flow discovered and exploited a loophole in a popular microarchitectural simulator - a consequence of the fact that these research tools were designed for a (now past) world where they were exclusively operated by humans acting in good-faith.

Do large language models (LLMs) represent concepts abstractly, i.e., independent of input format? We revisit Function Vectors (FVs), compact representations of in-context learning (ICL) tasks that causally drive task performance. Across multiple LLMs, we show that FVs are not fully invariant: FVs are nearly orthogonal when extracted from different input formats (e.g., open-ended vs. multiple-choice), even if both target the same concept. We identify Concept Vectors (CVs), which carry more stable concept representations. Like FVs, CVs are composed of attention head outputs; however, unlike FVs, the constituent heads are selected using Representational Similarity Analysis (RSA) based on whether they encode concepts consistently across input formats. While these heads emerge in similar layers to FV-related heads, the two sets are largely distinct, suggesting different underlying mechanisms. Steering experiments reveal that FVs excel in-distribution, when extraction and application formats match (e.g., both open-ended in English), while CVs generalize better out-of-distribution across both question types (open-ended vs. multiple-choice) and languages. Our results show that LLMs do contain abstract concept representations, but these differ from those that drive ICL performance.

Tuning parallel file system in High-Performance Computing (HPC) systems remains challenging due to the complex I/O paths, diverse I/O patterns, and dynamic system conditions. While existing autotuning frameworks have shown promising results in tuning PFS parameters based on applications' I/O patterns, they lack scalability, adaptivity, and the ability to operate online. In this work, focusing on scalable online tuning, we present CARAT, an ML-guided framework to co-tune client-side RPC and caching parameters of PFS, leveraging only locally observable metrics. Unlike global or pattern-dependent approaches, CARAT enables each client to make independent and intelligent tuning decisions online, responding to real-time changes in both application I/O behaviors and system states. We then prototyped CARAT using Lustre and evaluated it extensively across dynamic I/O patterns, real-world HPC workloads, and multi-client deployments. The results demonstrated that CARAT can achieve up to 3x performance improvement over the default or static configurations, validating the effectiveness and generality of our approach. Due to its scalability and lightweight, we believe CARAT has the potential to be widely deployed into existing PFS and benefit various data-intensive applications.

Smooth-basis models such as Chebyshev polynomial regressors and radial basis function (RBF) networks are well established in numerical analysis. Their continuously differentiable prediction surfaces suit surrogate optimisation, sensitivity analysis, and other settings where the response varies gradually with inputs. Despite these properties, smooth models seldom appear in tabular regression, where tree ensembles dominate. We ask whether they can compete, benchmarking models across 55 regression datasets organised by application domain. We develop an anisotropic RBF network with data-driven centre placement and gradient-based width optimisation, a ridge-regularised Chebyshev polynomial regressor, and a smooth-tree hybrid (Chebyshev model tree); all three are released as scikit-learn-compatible packages. We benchmark these against tree ensembles, a pre-trained transformer, and standard baselines, evaluating accuracy alongside generalisation behaviour. The transformer ranks first on accuracy across a majority of datasets, but its GPU dependence, inference latency, and dataset-size limits constrain deployment in the CPU-based settings common across applied science and industry. Among CPU-viable models, smooth models and tree ensembles are statistically tied on accuracy, but the former tend to exhibit tighter generalisation gaps. We recommend routinely including smooth-basis models in the candidate pool, particularly when downstream use benefits from tighter generalisation and gradually varying predictions.

We investigate the collective accuracy of heterogeneous agents who learn to estimate their own reliability over time and selectively abstain from voting. While classical epistemic voting results, such as the \textit{Condorcet Jury Theorem} (CJT), assume fixed participation, real-world aggregation often benefits from allowing agents to say ``I don't know.'' We propose a probabilistic framework where agents engage in a \textit{calibration} phase, updating beliefs about their own fixed competence, before facing a final confidence gate that determines whether to vote or abstain. We derive a non-asymptotic lower bound on the group's success probability and prove that this \textit{selective participation} generalizes the asymptotic guarantees of the CJT to a sequential, confidence-gated setting. Empirically, we validate these bounds via Monte Carlo simulations. While our results are general, we discuss their potential application to AI safety, outlining how this framework can mitigate \textit{hallucinations} in collective LLM decision-making.

In matching markets such as kidney exchanges and freight exchanges, delayed matching has been shown to improve overall market efficiency. The benefits of delay are highly sensitive to participants' sojourn times and departure behavior, and delaying matches can impose significant costs, including longer waiting times and increased market congestion. These competing effects make fixed matching policies inherently inflexible in dynamic environments. We propose a learning-based Hybrid framework that adaptively combines immediate and delayed matching. The framework continuously collects data on user departures over time, estimates the underlying departure distribution via regression, and determines whether to delay matching in the subsequent period based on a decision threshold that governs the system's tolerance for matching efficiency loss. The proposed framework can substantially reduce waiting times and congestion while sacrificing only a limited amount of matching efficiency. By dynamically adjusting its matching strategy, the Hybrid framework enables system performance to flexibly interpolate between purely greedy and purely patient policies, offering a robust and adaptive alternative to static matching mechanisms.

Modern high-performance computing (HPC) applications run on compute resources but share global storage systems. This design can cause problems when applications consume a disproportionate amount of storage bandwidth relative to their allocated compute resources. For example, an application running on a single compute node can issue many small, random writes and consume excessive I/O bandwidth from a storage server. This can hinder larger jobs that write to the same storage server and are allocated many compute nodes, resulting in significant resource waste. A straightforward solution is to limit each application's I/O bandwidth on storage servers in proportion to its allocated compute resources. This approach has been implemented in parallel file systems using Token Bucket Filter (TBF). However, strict proportional limits often reduce overall I/O efficiency because HPC applications generate short, bursty I/O. Limiting bandwidth can waste server capacity when applications are idle or prevent applications from temporarily using higher bandwidth during bursty phases. We argue that I/O control should maximize per-application performance and overall storage efficiency while ensuring fairness (e.g., preventing small jobs from blocking large-scale ones). We propose AdapTBF, which builds on TBF in modern parallel file systems (e.g., Lustre) and introduces a decentralized bandwidth control approach using adaptive borrowing and lending. We detail the algorithm, implement AdapTBF in Lustre, and evaluate it using synthetic workloads modeled after real-world scenarios. Results show that AdapTBF manages I/O bandwidth effectively while maintaining high storage utilization, even under extreme conditions.

Humans exhibit remarkable flexibility in abstract reasoning, and can rapidly learn and apply rules from sparse examples. To investigate the cognitive strategies underlying this ability, we introduce the Cognitive Abstraction and Reasoning Corpus (CogARC), a diverse human-adapted subset of the Abstraction and Reasoning Corpus (ARC) which was originally developed to benchmark abstract reasoning in artificial intelligence. Across two experiments, CogARC was administered to a total of 260 human participants who freely generated solutions to 75 abstract visual reasoning problems. Success required inferring input-output rules from a small number of examples to transform the test input into one correct test output. Participants' behavior was recorded at high temporal resolution, including example viewing, edit sequences, and multi-attempt submissions. Participants were generally successful (mean accuracy ~90% for experiment 1 (n=40), ~80% for experiment 2 (n=220) across problems), but performance varied widely across problems and participants. Harder problems elicited longer deliberation times and greater divergence in solution strategies. Over the course of the task, participants initiated responses more quickly but showed a slight decline in accuracy, suggesting increased familiarity with the task structure rather than improved rule-learning ability. Importantly, even incorrect solutions were often highly convergent, even when the problem-solving trajectories differed in length and smoothness. Some trajectories progressed directly and efficiently toward a stable outcome, whereas others involved extended exploration or partial restarts before converging. Together, these findings highlight CogARC as a rich behavioral environment for studying human abstract reasoning, providing insight into how people generalize, misgeneralize, and adapt their strategies under uncertainty.

Recent memory agents improve LLMs by extracting experiences and conversation history into an external storage. This enables low-overhead context assembly and online memory update without expensive LLM training. However, existing solutions remain passive and reactive; memory growth is bounded by information that happens to be available, while memory agents seldom seek external inputs in uncertainties. We propose autonomous memory agents that actively acquire, validate, and curate knowledge at a minimum cost. U-Mem materializes this idea via (i) a cost-aware knowledge-extraction cascade that escalates from cheap self/teacher signals to tool-verified research and, only when needed, expert feedback, and (ii) semantic-aware Thompson sampling to balance exploration and exploitation over memories and mitigate cold-start bias. On both verifiable and non-verifiable benchmarks, U-Mem consistently beats prior memory baselines and can surpass RL-based optimization, improving HotpotQA (Qwen2.5-7B) by 14.6 points and AIME25 (Gemini-2.5-flash) by 7.33 points.

Most machine learning models for molecular property prediction rely on a single molecular representation (either a sequence, a graph, or a 3D structure) and treat molecular geometry as static. We present MolFM-Lite, a multi-modal model that jointly encodes SELFIES sequences (1D), molecular graphs (2D), and conformer ensembles (3D) through cross-attention fusion, while conditioning predictions on experimental context via Feature-wise Linear Modulation (FiLM). Our main methodological contributions are: (1) a conformer ensemble attention mechanism that combines learnable attention with Boltzmann-weighted priors over multiple RDKit-generated conformers, capturing the thermodynamic distribution of molecular shapes; and (2) a cross-modal fusion layer where each modality can attend to others, enabling complementary information sharing. We evaluate on four MoleculeNet scaffold-split benchmarks using our model's own splits, and report all baselines re-evaluated under the same protocol. Comprehensive ablation studies across all four datasets confirm that each architectural component contributes independently, with tri-modal fusion providing 7-11% AUC improvement over single-modality baselines and conformer ensembles adding approximately 2% over single-conformer variants. Pre-training on ZINC250K (~250K molecules) using cross-modal contrastive and masked-atom objectives enables effective weight initialization at modest compute cost. We release all code, trained models, and data splits to support reproducibility.

Stereotype repositories are critical to assess generative AI model safety, but currently lack adequate global coverage. It is imperative to prioritize targeted expansion, strategically addressing existing deficits, over merely increasing data volume. This work introduces a multilingual stereotype resource covering four sub-Saharan African countries that are severely underrepresented in NLP resources: Ghana, Kenya, Nigeria, and South Africa. By utilizing socioculturally-situated, community-engaged methods, including telephonic surveys moderated in native languages, we establish a reproducible methodology that is sensitive to the region's complex linguistic diversity and traditional orality. By deliberately balancing the sample across diverse ethnic and demographic backgrounds, we ensure broad coverage, resulting in a dataset of 3,534 stereotypes in English and 3,206 stereotypes across 15 native languages.

Machine learning (ML)-based defect prediction models can improve software quality. However, their opaque reasoning creates an HCI challenge because developers struggle to trust models they cannot interpret. Explainable AI (XAI) methods such as LIME, SHAP, and BreakDown aim to provide transparency, but when used together, they often produce conflicting explanations that increase confusion, frustration, and cognitive load. To address this usability challenge, we introduce XMENTOR, a human-centered, rank-aware aggregation method implemented as a VS Code plugin. XMENTOR unifies multiple post-hoc explanations into a single, coherent view by applying adaptive thresholding, rank and sign agreement, and fallback strategies to preserve clarity without overwhelming users. In a user study, nearly 90% of the participants preferred aggregated explanations, citing reduced confusion and stronger support for daily tasks of debugging and review of defects. Our findings show how combining explanations and embedding them into developer workflows can enhance interpretability, usability, and trust.

As large language models engage in extended reasoning tasks, they accumulate significant state -- architectural mappings, trade-off decisions, codebase conventions -- within the context window. This understanding is lost when sessions reach context limits and undergo lossy compaction. We propose Contextual Memory Virtualisation (CMV), a system that treats accumulated LLM understanding as version-controlled state. Borrowing from operating system virtual memory, CMV models session history as a Directed Acyclic Graph (DAG) with formally defined snapshot, branch, and trim primitives that enable context reuse across independent parallel sessions. We introduce a three-pass structurally lossless trimming algorithm that preserves every user message and assistant response verbatim while reducing token counts by a mean of 20% and up to 86% for sessions with significant overhead by stripping mechanical bloat such as raw tool outputs, base64 images, and metadata. A single-user case-study evaluation across 76 real-world coding sessions demonstrates that trimming remains economically viable under prompt caching, with the strongest gains in mixed tool-use sessions, which average 39% reduction and reach break-even within 10 turns. A reference implementation is available at https://github.com/CosmoNaught/claude-code-cmv.

AI agents -- systems that execute multi-step reasoning workflows with persistent state, tool access, and specialist skills -- represent a qualitative shift from prior automation technologies in social science. Unlike chatbots that respond to isolated queries, AI agents can now read files, run code, query databases, search the web, and invoke domain-specific skills to execute entire research pipelines autonomously. This paper introduces the concept of vibe researching -- the AI-era parallel to ``vibe coding'' (Karpathy, 2025) -- and uses scholar-skill, a 21-skill plugin for Claude Code covering the full research pipeline from idea to submission, as an illustrative case. I develop a cognitive task framework that classifies research activities along two dimensions -- codifiability and tacit knowledge requirement -- to identify a delegation boundary that is cognitive, not sequential: it cuts through every stage of the research pipeline, not between stages. I argue that AI agents excel at speed, coverage, and methodological scaffolding but struggle with theoretical originality and tacit field knowledge. The paper concludes with an analysis of three implications for the profession -- augmentation with fragile conditions, stratification risk, and a pedagogical crisis -- and proposes five principles for responsible vibe researching.

The rise of Antimicrobial Resistance, particularly Multi-Drug Resistance (MDR), presents a critical challenge for clinical decision-making due to limited treatment options and delays in conventional susceptibility testing. This study proposes an interpretable machine learning framework to predict MDR in bacterial isolates using clinical features and antibiotic susceptibility patterns. Five classification models were evaluated, including Logistic Regression, Random Forest, AdaBoost, XGBoost, and LightGBM. The models were trained on a curated dataset of 9,714 isolates, with resistance encoded at the antibiotic family level to capture cross-class resistance patterns consistent with MDR definitions. Performance assessment included accuracy, F1-score, AUC-ROC, and Matthews Correlation Coefficient. Ensemble models, particularly XGBoost and LightGBM, demonstrated superior predictive capability across all metrics. To address the clinical transparency gap, Local Interpretable Model-agnostic Explanations (LIME) was applied to generate instance-level explanations. LIME identified resistance to quinolones, Co-trimoxazole, Colistin, aminoglycosides, and Furanes as the strongest contributors to MDR predictions, aligning with known biological mechanisms. The results show that combining high-performing models with local interpretability provides both accuracy and actionable insights for antimicrobial stewardship. This framework supports earlier MDR identification and enhances trust in machine learning-assisted clinical decision support.

Enabling efficient, high-performance data access in parallel file systems (PFS) is critical for today's high-performance computing systems. PFS client-side I/O heavily impacts the final I/O performance delivered to individual applications and the entire system. Autotuning the key client-side I/O behaviors has been extensively studied and shows promising results. However, existing work has heavily relied on extensive number of global runtime metrics to monitor and accurate modeling of applications' I/O patterns. Such heavy overheads significantly limit the ability to enable fine-grained, dynamic tuning in practical systems. In this study, we propose DIAL (Decentralized I/O AutoTuning via Learned Client-side Local Metrics) which takes a drastically different approach. Instead of trying to extract the global I/O patterns of applications, DIAL takes a decentralized approach, treating each I/O client as an independent unit and tuning configurations using only its locally observable metrics. With the help of machine learning models, DIAL enables multiple tunable units to make independent but collective decisions, reacting to what is happening in the global storage systems in a timely manner and achieving better I/O performance globally for the application.

Internet memes have become a dominant form of expression on social media, including within the Bengali-speaking community. While often humorous, memes can also be exploited to spread offensive, harmful, and inflammatory content targeting individuals and groups. Detecting this type of content is excep- tionally challenging due to its satirical, subtle, and culturally specific nature. This problem is magnified for low-resource lan- guages like Bengali, as existing research predominantly focuses on high-resource languages. To address this critical research gap, we introduce Bn-HIB (Bangla Hate Inflammatory Benign), a novel dataset containing 3,247 manually annotated Bengali memes categorized as Benign, Hate, or Inflammatory. Significantly, Bn- HIB is the first dataset to distinguish inflammatory content from direct hate speech in Bengali memes. Furthermore, we propose the MCFM (Multi-Modal Co-Attention Fusion Model), a simple yet effective architecture that mutually analyzes both the visual and textual elements of a meme. MCFM employs a co-attention mechanism to identify and fuse the most critical features from each modality, leading to a more accurate classification. Our experiments show that MCFM significantly outperforms several state-of-the-art models on the Bn-HIB dataset, demonstrating its effectiveness in this nuanced task.Warning: This work contains material that may be disturbing to some audience members. Viewer discretion is advised.

Biological signals of interest in high-dimensional data are often masked by dominant variation shared across conditions. This variation, arising from baseline biological structure or technical effects, can prevent standard dimensionality reduction methods from resolving condition-specific structure. The challenge is that these confounding topics are often unknown and mixed with biological signals. Existing background correction methods are either unscalable to high dimensions or not interpretable. We introduce background contrastive Non-negative Matrix Factorization (\model), which extracts target-enriched latent topics by jointly factorizing a target dataset and a matched background using shared non-negative bases under a contrastive objective that suppresses background-expressed structure. This approach yields non-negative components that are directly interpretable at the feature level, and explicitly isolates target-specific variation. \model is learned by an efficient multiplicative update algorithm via matrix multiplication such that it is highly efficient on GPU hardware and scalable to big data via minibatch training akin to deep learning approach. Across simulations and diverse biological datasets, \model reveals signals obscured by conventional methods, including disease-associated programs in postmortem depressive brain single-cell RNA-seq, genotype-linked protein expression patterns in mice, treatment-specific transcriptional changes in leukemia, and TP53-dependent drug responses in cancer cell lines.

Accurate prediction of malignancy in renal tumors is crucial for informing clinical decisions and optimizing treatment strategies. However, existing imaging modalities lack the necessary accuracy to reliably predict malignancy before surgical intervention. While deep learning has shown promise in malignancy prediction using 3D CT images, traditional approaches often rely on manual segmentation to isolate the tumor region and reduce noise, which enhances predictive performance. Manual segmentation, however, is labor-intensive, costly, and dependent on expert knowledge. In this study, a deep learning framework was developed utilizing an Organ Focused Attention (OFA) loss function to modify the attention of image patches so that organ patches attend only to other organ patches. Hence, no segmentation of 3D renal CT images is required at deployment time for malignancy prediction. The proposed framework achieved an AUC of 0.685 and an F1-score of 0.872 on a private dataset from the UF Integrated Data Repository (IDR), and an AUC of 0.760 and an F1-score of 0.852 on the publicly available KiTS21 dataset. These results surpass the performance of conventional models that rely on segmentation-based cropping for noise reduction, demonstrating the frameworks ability to enhance predictive accuracy without explicit segmentation input. The findings suggest that this approach offers a more efficient and reliable method for malignancy prediction, thereby enhancing clinical decision-making in renal cancer diagnosis.

Recent advances in 4D scene reconstruction have significantly improved dynamic modeling across various domains. However, existing approaches remain limited under aerial conditions with single-view capture, wide spatial range, and dynamic objects of limited spatial footprint and large motion disparity. These challenges cause severe depth ambiguity and unstable motion estimation, making monocular aerial reconstruction inherently ill-posed. To this end, we present AeroDGS, a physics-guided 4D Gaussian splatting framework for monocular UAV videos. AeroDGS introduces a Monocular Geometry Lifting module that reconstructs reliable static and dynamic geometry from a single aerial sequence, providing a robust basis for dynamic estimation. To further resolve monocular ambiguity, we propose a Physics-Guided Optimization module that incorporates differentiable ground-support, upright-stability, and trajectory-smoothness priors, transforming ambiguous image cues into physically consistent motion. The framework jointly refines static backgrounds and dynamic entities with stable geometry and coherent temporal evolution. We additionally build a real-world UAV dataset that spans various altitudes and motion conditions to evaluate dynamic aerial reconstruction. Experiments on synthetic and real UAV scenes demonstrate that AeroDGS outperforms state-of-the-art methods, achieving superior reconstruction fidelity in dynamic aerial environments.

Code summarization is the task of generating natural language descriptions of source code, which is critical for software comprehension and maintenance. While large language models (LLMs) have achieved remarkable progress on this task, an open question remains: can human expertise in code understanding further guide and enhance these models? We propose EyeLayer, a lightweight attention-augmentation module that incorporates human eye-gaze patterns, as a proxy of human expertise, into LLM-based code summarization. EyeLayer models human attention during code reading via a Multimodal Gaussian Mixture, redistributing token embeddings based on learned parameters (μ_i, σ_i^2) that capture where and how intensively developers focus. This design enables learning generalizable attention priors from eye-tracking data and incorporating them into LLMs seamlessly, without disturbing existing representations. We evaluate EyeLayer across diverse model families (i.e., LLaMA-3.2, Qwen3, and CodeBERT) covering different scales and architectures. EyeLayer consistently outperforms strong fine-tuning baselines across standard metrics, achieving gains of up to 13.17% on BLEU-4. These results demonstrate that human gaze patterns encode complementary attention signals that enhance the semantic focus of LLMs and transfer effectively across diverse models for code summarization.

Modern forward electrocardiogram (ECG) computational models rely on an accurate representation of the torso domain. The lead-field method enables fast ECG simulations while preserving full geometric fidelity. Achieving high anatomical accuracy in torso representation is, however, challenging in clinical practice, as imaging protocols are typically focused on the heart and often do not include the entire torso. In addition, the computational cost of the lead-field method scales linearly with the number of electrodes, limiting its applicability in high-density recording settings. To date, no existing approach simultaneously achieves high anatomical fidelity, low data requirements and computational efficiency. In this work, we propose a shape-informed surrogate model of the lead-field operator that serves as a drop-in replacement for the full-order model in forward ECG simulations. The proposed framework consists of two components: a geometry-encoding module that maps anatomical shapes into a low-dimensional latent space, and a geometry-conditioned neural surrogate that predicts lead-field gradients from spatial coordinates, electrode positions and latent codes. The proposed method achieves high accuracy in approximating lead fields both within the torso (mean angular error 5°) and inside the heart, resulting in highly accurate ECG simulations (relative mean squared error <2.5%. The surrogate consistently outperforms the widely used pseudo lead-field approximation while preserving negligible inference cost. Owing to its compact latent representation, the method does not require a fully detailed torso segmentation and can therefore be deployed in data-limited settings while preserving high-fidelity ECG simulations.

Crowdsourcing platforms face a four-way tension between allocation quality, workforce sustainability, operational feasibility, and strategic contractor behaviour--a dilemma we formalise as the Cold-Start, Burnout, Utilisation, and Strategic Agency Dilemma. Existing methods resolve at most two of these tensions simultaneously: greedy heuristics and multi-criteria decision making (MCDM) methods achieve Day-1 quality but cause catastrophic burnout, while bandit algorithms eliminate burnout only through operationally infeasible 100% workforce utilisation.To address this, we introduce FORGE, a physics-grounded $K+1$ multi-agent simulator in which each contractor is a rational agent that declares its own load-acceptance threshold based on its fatigue state, converting the standard passive Restless Multi-Armed Bandit (RMAB) into a genuine Stackelberg game. Operating within FORGE, we propose a Neural-Linear UCB allocator that fuses a Two-Tower embedding network with a Physics-Informed Covariance Prior derived from offline simulator interactions. The prior simultaneously warm-starts skill-cluster geometry and UCB exploration landscape, providing a geometry-aware belief state from episode 1 that measurably reduces cold-start regret.Over $T = 200$ cold-start episodes, the proposed method achieves the highest reward of all non-oracle methods ($\text{LRew} = 0.555 \pm 0.041$) at only 7.6% workforce utilisation--a combination no conventional baseline achieves--while maintaining robustness to workforce turnover up to 50% and observation noise up to $σ= 0.20$.

This paper tests whether large language models (LLMs) can support interpretative citation context analysis (CCA) by scaling in thick, text-grounded readings of a single hard case rather than scaling up typological labels. It foregrounds prompt-sensitivity analysis as a methodological issue by varying prompt scaffolding and framing in a balanced 2x3 design. Using footnote 6 in Chubin and Moitra (1975) and Gilbert's (1977) reconstruction as a probe, I implement a two-stage GPT-5 pipeline: a citation-text-only surface classification and expectation pass, followed by cross-document interpretative reconstruction using the citing and cited full texts. Across 90 reconstructions, the model produces 450 distinct hypotheses. Close reading and inductive coding identify 21 recurring interpretative moves, and linear probability models estimate how prompt choices shift their frequencies and lexical repertoire. GPT-5's surface pass is highly stable, consistently classifying the citation as "supplementary". In reconstruction, the model generates a structured space of plausible alternatives, but scaffolding and examples redistribute attention and vocabulary, sometimes toward strained readings. Relative to Gilbert, GPT-5 detects the same textual hinges yet more often resolves them as lineage and positioning than as admonishment. The study outlines opportunities and risks of using LLMs as guided co-analysts for inspectable, contestable interpretative CCA, and it shows that prompt scaffolding and framing systematically tilt which plausible readings and vocabularies the model foregrounds.

With the continuous growth of neural network scales, low-precision quantization is widely used in edge accelerators. Classic multi-threshold activation hardware requires 2^n thresholds for n-bit outputs, causing a rapid increase in hardware cost as precision increases. We propose a reconfigurable activation hardware, GRAU, based on piecewise linear fitting, where the segment slopes are approximated by powers of two. Our design requires only basic comparators and 1-bit right shifters, supporting mixed-precision quantization and nonlinear functions such as SiLU. Compared with multi-threshold activators, GRAU reduces LUT consumption by over 90%, achieving higher hardware efficiency, flexibility, and scalability.

Large language models (LLMs) learn contextual embeddings that capture rich semantic information, yet they often overlook structured lexical knowledge such as word senses and relationships. Prior work has shown that incorporating sense dictionaries can improve knowledge distillation for encoder models, but their application to decoder as generative models remains challenging. In this paper, we introduce Decoder-based Sense Knowledge Distillation (DSKD), a framework that integrates lexical resources into the training of decoder-style LLMs without requiring dictionary lookup at inference time. Extensive experiments on diverse benchmarks demonstrate that DSKD significantly enhances knowledge distillation performance for decoders, enabling generative models to inherit structured semantics while maintaining efficient training.

We introduce the concept of engineered simultaneity: a system design that (1) requires comparing events at spacelike-separated locations, (2) implements this comparison via an implicit simultaneity convention, and (3) represents the result as objective rather than conventional. The United States National Best Bid and Offer (NBBO), mandated by SEC Regulation NMS Rule 611, is shown to be an instance of engineered simultaneity. We prove that the NBBO is frame-dependent: its value depends on the reference frame in which "current" prices are defined. Since the exchanges that generate quote data are separated by distances of 43-1,180 km, light-travel times of 143-3,940 microseconds create unavoidable windows during which no frame-independent price ordering exists. High-frequency trading firms exploit this window by accessing exchange data via direct feeds (latency ~tens of microseconds) while the consolidated Securities Information Processor operates at ~1,128 microseconds -- a ratio exceeding 50:1. We demonstrate that this constitutes a category mistake in the sense of Ryle: the NBBO applies the concept of "simultaneity" in a domain where it has no frame-independent meaning. The resulting information asymmetry extracts approximately $5 billion annually from other market participants.

Foundation models in histopathology are expected to facilitate the development of high-performing and generalisable deep learning systems. However, current models capture not only biologically relevant features, but also pre-analytic and scanner-specific variation that bias the predictions of task-specific models trained from the foundation model features. Here we show that introducing novel robustness losses during training of downstream task-specific models reduces sensitivity to technical variability. A purpose-designed comprehensive experimentation setup with 27,042 WSIs from 6155 patients is used to train thousands of models from the features of eight popular foundation models for computational pathology. In addition to a substantial improvement in robustness, we observe that prediction accuracy improves by focusing on biologically relevant features. Our approach successfully mitigates robustness issues of foundation models for computational pathology without retraining the foundation models themselves, enabling development of robust computational pathology models applicable to real-world data in routine clinical practice.

This thesis addresses two persistent and closely related challenges in modern deep learning, reliability and efficiency, through a unified framework grounded in Spectral Geometry and Random Matrix Theory (RMT). As deep networks and large language models continue to scale, their internal behavior becomes increasingly opaque, leading to hallucinations, fragile generalization under distribution shift, and growing computational and energy demands. By analyzing the eigenvalue dynamics of hidden activations across layers and inputs, this work shows that spectral statistics provide a compact, stable, and interpretable lens on model behavior, capable of separating structured, causal representations from noise-dominated variability. Within this framework, the first contribution, EigenTrack, introduces a real-time method for detecting hallucinations and out-of-distribution behavior in large language and vision-language models. EigenTrack transforms streaming activations into spectral descriptors such as entropy, variance, and deviations from the Marchenko-Pastur baseline, and models their temporal evolution using lightweight recurrent classifiers, enabling early detection of reliability failures before they appear in model outputs while offering interpretable insight into representation dynamics. The second contribution, RMT-KD, presents a principled approach to compressing deep networks via random matrix theoretic knowledge distillation. By interpreting outlier eigenvalues in activation spectra as carriers of task-relevant information, RMT-KD progressively projects networks onto lower-dimensional subspaces through iterative self-distillation, yielding significantly more compact and energy-efficient models while preserving accuracy and dense, hardware-friendly structure.

Kurtosis-based Independent Component Analysis (ICA) weakens in wide, balanced mixtures. We prove a sharp redundancy law: for a standardized projection with effective width $R_{\mathrm{eff}}$ (participation ratio), the population excess kurtosis obeys $|κ(y)|=O(κ_{\max}/R_{\mathrm{eff}})$, yielding the order-tight $O(c_bκ_{\max}/R)$ under balance (typically $c_b=O(\log R)$). As an impossibility screen, under standard finite-moment conditions for sample kurtosis estimation, surpassing the $O(1/\sqrt{T})$ estimation scale requires $R\lesssim κ_{\max}\sqrt{T}$. We also show that \emph{purification} -- selecting $m\!\ll\!R$ sign-consistent sources -- restores $R$-independent contrast $Ω(1/m)$, with a simple data-driven heuristic. Synthetic experiments validate the predicted decay, the $\sqrt{T}$ crossover, and contrast recovery.

The reliability of multilingual Large Language Model (LLM) evaluation is currently compromised by the inconsistent quality of translated benchmarks. Existing resources often suffer from semantic drift and context loss, which can lead to misleading performance metrics. In this work, we present a fully automated framework designed to address these challenges by enabling scalable, high-quality translation of datasets and benchmarks. We demonstrate that adapting test-time compute scaling strategies, specifically Universal Self-Improvement (USI) and our proposed multi-round ranking method, T-RANK, allows for significantly higher quality outputs compared to traditional pipelines. Our framework ensures that benchmarks preserve their original task structure and linguistic nuances during localization. We apply this approach to translate popular benchmarks and datasets into eight Eastern and Southern European languages (Ukrainian, Bulgarian, Slovak, Romanian, Lithuanian, Estonian, Turkish, Greek). Evaluations using both reference-based metrics and LLM-as-a-judge show that our translations surpass existing resources, resulting in more accurate downstream model assessment. We release both the framework and the improved benchmarks to facilitate robust and reproducible multilingual AI development.

Sumerian transliteration is a conventional system for representing a scholar's interpretation of a tablet in the Latin script. Thanks to visionary digital Assyriology projects such as ETCSL, CDLI, and Oracc, a large number of Sumerian transliterations have been published online, and these data are well-structured for a variety of search and analysis tasks. However, the absence of a comprehensive, accessible dataset pairing transliterations with a digital representation of the tablet's cuneiform glyphs has prevented the application of modern Natural Language Processing (NLP) methods to the task of Sumerian transliteration. To address this gap, we present SumTablets, a dataset pairing Unicode representations of 91,606 Sumerian cuneiform tablets (totaling 6,970,407 glyphs) with the associated transliterations published by Oracc. We construct SumTablets by first preprocessing and standardizing the Oracc transliterations before mapping each reading back to the Unicode representation of the source glyph. Further, we retain parallel structural information (e.g., surfaces, newlines, broken segments) through the use of special tokens. We release SumTablets as a Hugging Face Dataset (CC BY 4.0) and open source data preparation code via GitHub. Additionally, we leverage SumTablets to implement and evaluate two transliteration baselines: (1) weighted sampling from a glyph's possible readings, and (2) fine-tuning an autoregressive language model. Our fine-tuned language model achieves an average transliteration character-level F-score (chrF) of 97.55, demonstrating the immediate potential of transformer-based transliteration models in allowing experts to rapidly verify generated transliterations rather than manually transliterating tablets one-by-one.

Frontier AI agents may pursue hidden goals while concealing their pursuit from oversight. Alignment training aims to prevent such behavior by reinforcing the correct goals, but alignment may not always succeed and can lead to unwanted side effects. We propose self-incrimination training, which instead trains agents to produce a visible signal when they covertly misbehave. We train GPT-4.1 and Gemini-2.0 agents to call a report_scheming() tool when behaving deceptively and measure their ability to cause harm undetected in out-of-distribution environments. Self-incrimination significantly reduces the undetected successful attack rate, outperforming matched-capability monitors and alignment baselines while preserving instruction hierarchy and incurring minimal safety tax on general capabilities. Unlike blackbox monitoring, self-incrimination performance is consistent across tasks regardless of how suspicious the misbehavior appears externally. The trained behavior persists under adversarial prompt optimization and generalizes to settings where agents pursue misaligned goals themselves rather than being instructed to misbehave. Our results suggest self-incrimination offers a viable path for reducing frontier misalignment risk, one that neither assumes misbehavior can be prevented nor that it can be reliably classified from the outside.

Advances in Generative AI (GenAI) have led to the development of various protection strategies to prevent the unauthorized use of images. These methods rely on adding imperceptible protective perturbations to images to thwart misuse such as style mimicry or deepfake manipulations. Although previous attacks on these protections required specialized, purpose-built methods, we demonstrate that this is no longer necessary. We show that off-the-shelf image-to-image GenAI models can be repurposed as generic ``denoisers" using a simple text prompt, effectively removing a wide range of protective perturbations. Across 8 case studies spanning 6 diverse protection schemes, our general-purpose attack not only circumvents these defenses but also outperforms existing specialized attacks while preserving the image's utility for the adversary. Our findings reveal a critical and widespread vulnerability in the current landscape of image protection, indicating that many schemes provide a false sense of security. We stress the urgent need to develop robust defenses and establish that any future protection mechanism must be benchmarked against attacks from off-the-shelf GenAI models. Code is available in this repository: https://github.com/mlsecviswanath/img2imgdenoiser

Sybil resistance is a key requirement of decentralized consensus protocols. It is achieved by introducing a scarce resource (such as computational power, monetary stake, disk space, etc.), which prevents participants from costlessly creating multiple fake identities and hijacking the protocol. Quantum states are generically uncloneable, which suggests that they may serve naturally as an unconditionally scarce resource. In particular, uncloneability underlies quantum position-based cryptography, which is unachievable classically. We design a consensus protocol that combines classical hybrid consensus protocols with quantum position verification as the Sybil resistance mechanism, providing security in the standard model, and achieving improved energy efficiency compared to hybrid protocols based on Proof-of-Work. Our protocol inherits the benefits of other hybrid protocols, namely the faster confirmation times compared to pure Proof-of-Work protocols, and resilience against the compounding wealth issue that plagues protocols based on Proof-of-Stake Sybil resistance. We additionally propose a spam prevention mechanism for our protocol in the Random Oracle model.

We study reasoning for accessing world knowledge stored in a language model's parameters. For example, recalling that Canberra is Australia's capital may benefit from thinking through major cities and the concept of purpose-built capitals. While reasoning language models are trained via reinforcement learning to produce reasoning traces on tasks such as mathematics, they may not reason well for accessing their own world knowledge. We first find that models do not generate their best world knowledge reasoning by default: adding a simple "think step-by-step" cue demonstrates statistically significant improvement in knowledge recall but not math. Motivated by this, we propose training models to reason over their parametric knowledge using world-knowledge question answering as a verifiable reward. After reinforcement learning on TriviaQA (+9.9%), performance also improves on Natural Questions, HotpotQA, SimpleQA, and StrategyQA by 4.2%, 2.1%, 0.6%, and 3.0%, respectively. Reasoning models are under-optimized for parametric knowledge access, but can be easily trained to reason better.

Traditional software relies on contracts -- APIs, type systems, assertions -- to specify and enforce correct behavior. AI agents, by contrast, operate on prompts and natural language instructions with no formal behavioral specification. This gap is the root cause of drift, governance failures, and frequent project failures in agentic AI deployments. We introduce Agent Behavioral Contracts (ABC), a formal framework that brings Design-by-Contract principles to autonomous AI agents. An ABC contract C = (P, I, G, R) specifies Preconditions, Invariants, Governance policies, and Recovery mechanisms as first-class, runtime-enforceable components. We define (p, delta, k)-satisfaction -- a probabilistic notion of contract compliance that accounts for LLM non-determinism and recovery -- and prove a Drift Bounds Theorem showing that contracts with recovery rate gamma > alpha (the natural drift rate) bound behavioral drift to D* = alpha/gamma in expectation, with Gaussian concentration in the stochastic setting. We establish sufficient conditions for safe contract composition in multi-agent chains and derive probabilistic degradation bounds. We implement ABC in AgentAssert, a runtime enforcement library, and evaluate on AgentContract-Bench, a benchmark of 200 scenarios across 7 models from 6 vendors. Results across 1,980 sessions show that contracted agents detect 5.2-6.8 soft violations per session that uncontracted baselines miss entirely (p < 0.0001, Cohen's d = 6.7-33.8), achieve 88-100% hard constraint compliance, and bound behavioral drift to D* < 0.27 across extended sessions, with 100% recovery for frontier models and 17-100% across all models, at overhead < 10 ms per action.

Open-source native GUI agents still lag behind closed-source systems on long-horizon navigation tasks. This gap stems from two limitations: a shortage of high-quality, action-aligned reasoning data, and the direct adoption of generic post-training pipelines that overlook the unique challenges of GUI agents. We identify two fundamental issues in these pipelines: (i) standard SFT with CoT reasoning often hurts grounding, and (ii) step-wise RLVR-tyle training faces partial verifiability, where multiple actions can be correct but only a single demonstrated action is used for verification. This makes offline step-wise metrics weak predictors of online task success. In this work, we present GUI-Libra, a tailored training recipe that addresses these challenges. First, to mitigate the scarcity of action-aligned reasoning data, we introduce a data construction and filtering pipeline and release a curated 81K GUI reasoning dataset. Second, to reconcile reasoning with grounding, we propose action-aware SFT that mixes reasoning-then-action and direct-action data and reweights tokens to emphasize action and grounding. Third, to stabilize RL under partial verifiability, we identify the overlooked importance of KL regularization in RLVR and show that a KL trust region is critical for improving offline-to-online predictability; we further introduce success-adaptive scaling to downweight unreliable negative gradients. Across diverse web and mobile benchmarks, GUI-Libra consistently improves both step-wise accuracy and end-to-end task completion. Our results suggest that carefully designed post-training and data curation can unlock significantly stronger task-solving capabilities without costly online data collection. We release our dataset, code, and models to facilitate further research on data-efficient post-training for reasoning-capable GUI agents.

Modelling rock-fluid interaction requires solving a set of partial differential equations (PDEs) to predict the flow behaviour and the reactions of the fluid with the rock on the interfaces. Conventional high-fidelity numerical models require a high resolution to obtain reliable results, resulting in huge computational expense. This restricts the applicability of these models for multi-query problems, such as uncertainty quantification and optimisation, which require running numerous scenarios. As a cheaper alternative to high-fidelity models, this work develops eight surrogate models for predicting the fluid flow in porous media. Four of these are reduced-order models (ROM) based on one neural network for compression and another for prediction. The other four are single neural networks with the property of grid-size invariance; a term which we use to refer to image-to-image models that are capable of inferring on computational domains that are larger than those used during training. In addition to the novel grid-size-invariant framework for surrogate models, we compare the predictive performance of UNet and UNet++ architectures, and demonstrate that UNet++ outperforms UNet for surrogate models. Furthermore, we show that the grid-size-invariant approach is a reliable way to reduce memory consumption during training, resulting in good correlation between predicted and ground-truth values and outperforming the ROMs analysed. The application analysed is particularly challenging because fluid-induced rock dissolution results in a non-static solid field and, consequently, it cannot be used to help in adjustments of the future prediction.

We study the algorithmic task of testably learning general Massart halfspaces under the Gaussian distribution. In the testable learning setting, the aim is the design of a tester-learner pair satisfying the following properties: (1) if the tester accepts, the learner outputs a hypothesis and a certificate that it achieves near-optimal error, and (2) it is highly unlikely that the tester rejects if the data satisfies the underlying assumptions. Our main result is the first testable learning algorithm for general halfspaces with Massart noise and Gaussian marginals. The complexity of our algorithm is $d^{\mathrm{polylog}(\min\{1/γ, 1/ε\})}$, where $ε$ is the excess error and $γ$ is the bias of the target halfspace, which qualitatively matches the known quasi-polynomial Statistical Query lower bound for the non-testable setting. The analysis of our algorithm hinges on a novel sandwiching polynomial approximation to the sign function with multiplicative error that may be of broader interest.

Over the last twenty years, significant progress has been made in designing and implementing Question Answering (QA) systems. However, addressing complex questions, the answers to which are spread across multiple documents, remains a challenging problem. Recent QA systems that are designed to handle complex questions work either on the basis of knowledge graphs, or utilise contem- porary neural models that are expensive to train, in terms of both computational resources and the volume of training data required. In this paper, we present LiCQA, an unsupervised question answer- ing model that works primarily on the basis of corpus evidence. We empirically compare the effectiveness and efficiency of LiCQA with two recently presented QA systems, which are based on different underlying principles. The results of our experiments show that LiCQA significantly outperforms these two state-of-the-art systems on benchmark data with noteworthy reduction in latency.

In many applications, it is important to identify subpopulations that survive longer or shorter than the rest of the population. In medicine, for example, it allows determining which patients benefit from treatment, and in predictive maintenance, which components are more likely to fail. Existing methods for discovering subgroups with exceptional survival characteristics require restrictive assumptions about the survival model (e.g. proportional hazards), pre-discretized features, and, as they compare average statistics, tend to overlook individual deviations. In this paper, we propose Sysurv, a fully differentiable, non-parametric method that leverages random survival forests to learn individual survival curves, automatically learns conditions and how to combine these into inherently interpretable rules, so as to select subgroups with exceptional survival characteristics. Empirical evaluation on a wide range of datasets and settings, including a case study on cancer data, shows that Sysurv reveals insightful and actionable survival subgroups.

Video-based ads are a vital medium for brands to engage consumers, with social media platforms leveraging user data to optimize ad delivery and boost engagement. A crucial but under-explored aspect is the 'hooking period', the first three seconds that capture viewer attention and influence engagement metrics. Analyzing this brief window is challenging due to the multimodal nature of video content, which blends visual, auditory, and textual elements. Traditional methods often miss the nuanced interplay of these components, requiring advanced frameworks for thorough evaluation. This study presents a framework using transformer-based multimodal large language models (MLLMs) to analyze the hooking period of video ads. It tests two frame sampling strategies, uniform random sampling and key frame selection, to ensure balanced and representative acoustic feature extraction, capturing the full range of design elements. The hooking video is processed by state-of-the-art MLLMs to generate descriptive analyses of the ad's initial impact, which are distilled into coherent topics using BERTopic for high-level abstraction. The framework also integrates features such as audio attributes and aggregated ad targeting information, enriching the feature set for further analysis. Empirical validation on large-scale real-world data from social media platforms demonstrates the efficacy of our framework, revealing correlations between hooking period features and key performance metrics like conversion per investment. The results highlight the practical applicability and predictive power of the approach, offering valuable insights for optimizing video ad strategies. This study advances video ad analysis by providing a scalable methodology for understanding and enhancing the initial moments of video advertisements.

Understanding and reasoning over long contexts is a crucial capability for language models (LMs). Although recent models support increasingly long context windows, their accuracy often deteriorates as input length grows. In practice, models often struggle to keep attention aligned with the most relevant context throughout decoding. In this work, we propose DySCO, a novel decoding algorithm for improving long-context reasoning. DySCO leverages retrieval heads--a subset of attention heads specialized for long-context retrieval--to identify task-relevant tokens at each decoding step and explicitly up-weight them. By doing so, DySCO dynamically adjusts attention during generation to better utilize relevant context. The method is training-free and can be applied directly to any off-the-shelf LMs. Across multiple instruction-tuned and reasoning models, DySCO consistently improves performance on challenging long-context reasoning benchmarks, yielding relative gains of up to 25% on MRCR and LongBenchV2 at 128K context length with modest additional compute. Further analysis highlights the importance of both dynamic attention rescaling and retrieval-head-guided selection for the effectiveness of the method, while providing interpretability insights into decoding-time attention behavior. Our code is available at https://github.com/princeton-pli/DySCO.

Mixed-Integer Programs (MIPs) are NP-hard optimization models that arise in a broad range of decision-making applications, including finance, logistics, energy systems, and network design. Although modern commercial solvers have achieved remarkable progress and perform effectively on many small- and medium-sized instances, their performance often degrades when confronted with large-cale or highly constrained formulations. This paper explores the use of the Random-Key Optimizer (RKO) framework as a flexible, metaheuristic alternative for computing high-quality solutions to MIPs through the design of problem-specific decoders. The proposed approach separates the search process from feasibility enforcement by operating in a continuous random-key space while mapping candidate solutions to feasible integer solutions via efficient decoding procedures. We evaluate the methodology on two representative and structurally distinct benchmark problems: the mean-variance Markowitz portfolio optimization problem with buy-in and cardinality constraints, and the Time-Dependent Traveling Salesman Problem. For each formulation, tailored decoders are developed to reduce the effective search space, promote feasibility, and accelerate convergence. Computational experiments demonstrate that RKO consistently produces competitive, and in several cases superior, solutions compared to a state-of-the-art commercial MIP solver, both in terms of solution quality and computational time. These results highlight the potential of RKO as a scalable and versatile heuristic framework for tackling challenging large-scale MIPs.

Checkpointing is essential for fault tolerance in training large language models (LLMs). However, existing methods, regardless of their I/O strategies, periodically store the entire model and optimizer states, incurring substantial storage overhead and resource contention. Recent studies reveal that updates across LLM layers are highly non-uniform. Across training steps, some layers may undergo more significant changes, while others remain relatively stable or even unchanged. This suggests that selectively checkpointing only layers with significant updates could reduce overhead without harming training. Implementing such selective strategies requires fine-grained control over both weights and optimizer states, which no current tool provides. To address this gap, we propose \texttt{LLMTailor}, a checkpoint-merging framework that filters and assembles layers from different checkpoints to form a composite checkpoint. Our evaluation indicates that LLMTailor can work with different selective checkpointing strategies and effectively reduce checkpoint size (e.g., 4.3 times smaller for Llama3.1-8B) and checkpoint time (e.g., 2.8 times faster for Qwen2.5-7B) while maintaining model quality.

The inability of Large Language Models (LLMs) to modulate their personality expression in response to evolving dialogue dynamics hinders their performance in complex, interactive contexts. We propose a model-agnostic framework for dynamic personality simulation that employs state machines to represent latent personality states, where transition probabilities are dynamically adapted to the conversational context. Part of our architecture is a modular pipeline for continuous personality scoring that evaluates dialogues along latent axes while remaining agnostic to the specific personality models, their dimensions, transition mechanisms, or LLMs used. These scores function as dynamic state variables that systematically reconfigure the system prompt, steering behavioral alignment throughout the interaction.We evaluate this framework by operationalizing the Interpersonal Circumplex (IPC) in a medical education setting. Results demonstrate that the system successfully adapts its personality state to user inputs, but also influences user behavior, thereby facilitating de-escalation training. Notably, the scoring pipeline maintains comparable precision even when utilizing lightweight, fine-tuned classifiers instead of large-scale LLMs. This work demonstrates the feasibility of modular, personality-adaptive architectures for education, customer support, and broader human-computer interaction.

Most contemporary neural learning systems rely on epoch-based optimization and repeated access to historical data, implicitly assuming reversible computation. In contrast, real-world environments often present information as irreversible streams, where inputs cannot be replayed or revisited. Under such conditions, conventional architectures degrade into reactive filters lacking long-horizon coherence. This paper introduces Stream Neural Networks (StNN), an execution paradigm designed for irreversible input streams. StNN operates through a stream-native execution algorithm, the Stream Network Algorithm (SNA), whose fundamental unit is the stream neuron. Each stream neuron maintains a persistent temporal state that evolves continuously across inputs. We formally establish three structural guarantees: (1) stateless mappings collapse under irreversibility and cannot encode temporal dependencies; (2) persistent state dynamics remain bounded under mild activation constraints; and (3) the state transition operator is contractive for λ < 1, ensuring stable long-horizon execution. Empirical phase-space analysis and continuous tracking experiments validate these theoretical results. The execution principles introduced in this work define a minimal substrate for neural computation under irreversible streaming constraints.

Cardiovascular disease (CVD) remains one of the leading global health challenges, accounting for more than 19 million deaths worldwide. To address this, several tools that aim to predict CVD risk and support clinical decision making have been developed. In particular, the Framingham Risk Score (FRS) is one of the most widely used and recommended worldwide. However, it does not explain why a patient was assigned to a particular risk category nor how it can be reduced. Due to this lack of transparency, we present a logical explainer for the FRS. Based on first-order logic and explainable artificial intelligence (XAI) fundaments, the explainer is capable of identifying a minimal set of patient attributes that are sufficient to explain a given risk classification. Our explainer also produces actionable scenarios that illustrate which modifiable variables would reduce a patient's risk category. We evaluated all possible input combinations of the FRS (over 22,000 samples) and tested them with our explainer, successfully identifying important risk factors and suggesting focused interventions for each case. The results may improve clinician trust and facilitate a wider implementation of CVD risk assessment by converting opaque scores into transparent and prescriptive insights, particularly in areas with restricted access to specialists.

Reinforcement Learning from Human Feedback (RLHF) plays a significant role in aligning Large Language Models (LLMs) with human preferences. While RLHF with expected reward constraints can be formulated as a primal-dual optimization problem, standard primal-dual methods only guarantee convergence with a distributional policy where the saddle-point problem is in convex-concave form. Moreover, standard primal-dual methods may exhibit instability or divergence in the last iterate under policy parameterization in practical applications. In this work, we propose a universal primal-dual framework for safe RLHF that unifies a broad class of existing alignment algorithms, including safe-RLHF, one-shot, and multi-shot based methods. Building on this framework, we introduce an optimistic primal-dual (OPD) algorithm that incorporates predictive updates for both primal and dual variables to stabilize saddle-point dynamics. We establish last-iterate convergence guarantees for the proposed method, covering both exact policy optimization in the distributional space and convergence to a neighborhood of the optimal solution whose gap is related to approximation error and bias under parameterized policies. Our analysis reveals that optimism plays a crucial role in mitigating oscillations inherent to constrained alignment objectives, thereby closing a key theoretical gap between constrained RL and practical RLHF.

Large Language Models (LLMs) are increasingly used to ``professionalize'' workplace communication, often at the cost of linguistic identity. We introduce "Cultural Ghosting", the systematic erasure of linguistic markers unique to non-native English varieties during text processing. Through analysis of 22,350 LLM outputs generated from 1,490 culturally marked texts (Indian, Singaporean,& Nigerian English) processed by five models under three prompt conditions, we quantify this phenomenon using two novel metrics: Identity Erasure Rate (IER) & Semantic Preservation Score (SPS). Across all prompts, we find an overall IER of 10.26%, with model-level variation from 3.5% to 20.5% (5.9x range). Crucially, we identify a Semantic Preservation Paradox: models maintain high semantic similarity (mean SPS = 0.748) while systematically erasing cultural markers. Pragmatic markers (politeness conventions) are 1.9x more vulnerable than lexical markers (71.5% vs. 37.1% erasure). Our experiments demonstrate that explicit cultural-preservation prompts reduce erasure by 29% without sacrificing semantic quality.

Current AI weather forecasting models predict conventional atmospheric variables but cannot distinguish between cloud microphysical species critical for aviation safety. We introduce AviaSafe, a hierarchical, physics-informed neural forecaster that produces global, six-hourly predictions of these four hydrometeor species for lead times up to 7 days. Our approach addresses the unique challenges of cloud prediction: extreme sparsity, discontinuous distributions, and complex microphysical interactions between species. We integrate the Icing Condition (IC) index from aviation meteorology as a physics-based constraint that identifies regions where supercooled water fuels explosive ice crystal growth. The model employs a hierarchical architecture that first predicts cloud spatial distribution through masked attention, then quantifies species concentrations within identified regions. Training on ERA5 reanalysis data, our model achieves lower RMSE for cloud species compared to baseline and outperforms operational numerical models on certain key variables at 7-day lead times. The ability to forecast individual cloud species enables new applications in aviation route optimization where distinguishing between ice and liquid water determines engine icing risk.

Object hallucination is a critical issue in Large Vision-Language Models (LVLMs), where outputs include objects that do not appear in the input image. A natural question arises from this phenomenon: Which component of the LVLM pipeline primarily contributes to object hallucinations? The vision encoder to perceive visual information, or the language decoder to generate text responses? In this work, we strive to answer this question through designing a systematic experiment to analyze the roles of the vision encoder and the language decoder in hallucination generation. Our observations reveal that object hallucinations are predominantly associated with the strong priors from the language decoder. Based on this finding, we propose a simple and training-free framework, No-Language-Hallucination Decoding, NoLan, which refines the output distribution by dynamically suppressing language priors, modulated based on the output distribution difference between multimodal and text-only inputs. Experimental results demonstrate that NoLan effectively reduces object hallucinations across various LVLMs on different tasks. For instance, NoLan achieves substantial improvements on POPE, enhancing the accuracy of LLaVA-1.5 7B and Qwen-VL 7B by up to 6.45 and 7.21, respectively. The code is publicly available at: https://github.com/lingfengren/NoLan.

Deep neural networks (DNNs) are essential for performing advanced tasks on edge or mobile devices, yet their deployment is often hindered by severe resource constraints, including limited memory, energy, and computational power. While uniform quantization provides a straightforward approach to compress model and reduce hardware requirement, it fails to fully leverage the varying robustness across layers, and often lead to accuracy degradation or suboptimal resource usage, particularly at low bitwidths. In contrast, heterogeneous quantization, which allocates different bitwidths to individual layers, can mitigate these drawbacks. Nonetheless, current heterogeneous quantization methods either needs huge brute-force design space search or lacks the adaptability to meet different hardware conditions, such as memory size, energy budget, and latency requirement. Filling these gaps, this work introduces \textbf{\textit{SigmaQuant}}, an adaptive layer-wise heterogeneous quantization framework designed to efficiently balance accuracy and resource usage for varied edge environments without exhaustive search.

We study the basic task of mean estimation in the presence of mean-shift contamination. In the mean-shift contamination model, an adversary is allowed to replace a small constant fraction of the clean samples by samples drawn from arbitrarily shifted versions of the base distribution. Prior work characterized the sample complexity of this task for the special cases of the Gaussian and Laplace distributions. Specifically, it was shown that consistent estimation is possible in these cases, a property that is provably impossible in Huber's contamination model. An open question posed in earlier work was to determine the sample complexity of mean estimation in the mean-shift contamination model for general base distributions. In this work, we study and essentially resolve this open question. Specifically, we show that, under mild spectral conditions on the characteristic function of the (potentially multivariate) base distribution, there exists a sample-efficient algorithm that estimates the target mean to any desired accuracy. We complement our upper bound with a qualitatively matching sample complexity lower bound. Our techniques make critical use of Fourier analysis, and in particular introduce the notion of a Fourier witness as an essential ingredient of our upper and lower bounds.

Instruction-following benchmarks remain predominantly English-centric, leaving a critical evaluation gap for the hundreds of millions of Indic language speakers. We introduce IndicIFEval, a benchmark evaluating constrained generation of LLMs across 14 Indic languages using automatically verifiable, rule-based instructions. It comprises around 800 human-verified examples per language spread across two complementary subsets: IndicIFEval-Ground, translated prompts from IFEval (Zhou et al., 2023) carefully localized for Indic contexts, and IndicIFEval-Ground, synthetically generated instructions grounded in native Indic content. We conduct a comprehensive evaluation of major open-weight and proprietary models spanning both reasoning and non-reasoning models. While models maintain strong adherence to formatting constraints, they struggle significantly with lexical and cross-lingual tasks -- and despite progress in high-resource languages, instruction-following across the broader Indic family lags significantly behind English. We release IndicIFEval and its evaluation scripts to support progress on multilingual constrained generation (http://github.com/ai4bharat/IndicIFEval).

Small language models (SLMs) offer compelling advantages in cost, latency, and adaptability, but have so far lagged behind larger models on long-horizon software engineering tasks such as SWE-bench, where they suffer from pervasive action looping and low resolution rates. We introduce SWE-Protégé, a post-training framework that reframes software repair as an expert-protégé collaboration problem. In SWE-Protégé, an SLM remains the sole decision-maker while learning to selectively seek guidance from a strong expert model, recognize stalled states, and follow through on expert feedback. Our approach combines supervised fine-tuning on expert-augmented trajectories with agentic reinforcement learning that explicitly discourages degenerative looping and unproductive expert collaboration. We lightly post-train Qwen2.5-Coder-7B-Instruct to achieve 42.4% Pass@1 on SWE-bench Verified, a +25.4% improvement over the prior SLM state of the art, while using expert assistance sparsely (~4 calls per task and 11% of total tokens).

Understanding the geometry of learned distributions is fundamental to improving and interpreting diffusion models, yet systematic tools for exploring their landscape remain limited. Standard latent-space interpolations fail to respect the structure of the learned distribution, often traversing low-density regions. We introduce a framework based on the string method that computes continuous paths between samples by evolving curves under the learned score function. Operating on pretrained models without retraining, our approach interpolates between three regimes: pure generative transport, which yields continuous sample paths; gradient-dominated dynamics, which recover minimum energy paths (MEPs); and finite-temperature string dynamics, which compute principal curves -- self-consistent paths that balance energy and entropy. We demonstrate that the choice of regime matters in practice. For image diffusion models, MEPs contain high-likelihood but unrealistic ''cartoon'' images, confirming prior observations that likelihood maxima appear unrealistic; principal curves instead yield realistic morphing sequences despite lower likelihood. For protein structure prediction, our method computes transition pathways between metastable conformers directly from models trained on static structures, yielding paths with physically plausible intermediates. Together, these results establish the string method as a principled tool for probing the modal structure of diffusion models -- identifying modes, characterizing barriers, and mapping connectivity in complex learned distributions.

Neural networks (NNs) are pervasive across various domains but often lack interpretability. To address the growing need for explanations, logic-based approaches have been proposed to explain predictions made by NNs, offering correctness guarantees. However, scalability remains a concern in these methods. This paper proposes an approach leveraging domain slicing to facilitate explanation generation for NNs. By reducing the complexity of logical constraints through slicing, we decrease explanation time by up to 40\% less time, as indicated through comparative experiments. Our findings highlight the efficacy of domain slicing in enhancing explanation efficiency for NNs.

Despite the extensive literature on training loss functions, the evaluation of generalization on the validation set remains underexplored. In this work, we conduct a systematic empirical and statistical study of how the validation criterion used for model selection affects test performance in neural classifiers, with attention to early stopping. Using fully connected networks on standard benchmarks under $k$-fold evaluation, we compare: (i) early stopping with patience and (ii) post-hoc selection over all epochs (i.e. no early stopping). Models are trained with cross-entropy, C-Loss, or PolyLoss; the model parameter selection on the validation set is made using accuracy or one of the three loss functions, each considered independently. Three main findings emerge. (1) Early stopping based on validation accuracy performs worst, consistently selecting checkpoints with lower test accuracy than both loss-based early stopping and post-hoc selection. (2) Loss-based validation criteria yield comparable and more stable test accuracy. (3) Across datasets and folds, any single validation rule often underperforms the test-optimal checkpoint. Overall, the selected model typically achieves test-set performance statistically lower than the best performance across all epochs, regardless of the validation criterion. Our results suggest avoiding validation accuracy (in particular with early stopping) for parameter selection, favoring loss-based validation criteria.

The increasing complexity and diversity of hardware accelerators in modern computing systems demand flexible, low-overhead program analysis tools. We present PASTA, a low-overhead and modular Program AnalysiS Tool Framework for Accelerators. PASTA abstracts over low-level profiling APIs and diverse deep learning frameworks, offering users a unified interface to capture and analyze runtime events at multiple levels. Its extensible design enables researchers and practitioners to rapidly prototype custom tools with minimal overhead. We demonstrate the utility of PASTA by developing several analysis tools, including a deep learning workload characterization tool and a UVM optimization tool. Through extensive evaluation on mainstream deep learning workloads tested on NVIDIA and AMD GPUs under both single- and multi-GPU scenarios, we demonstrate PASTA's broad applicability. On NVIDIA GPUs, we further show that PASTA provides detailed performance insights with significantly lower overhead, up to 1.3*10^4 faster than conventional analysis tools, thanks to its GPU-accelerated backend. PASTA strikes a practical balance between usability, extensibility, and efficiency, making it well-suited for modern accelerator-based computing environments.

Many real-world applications provide a continuous stream of data that is subsequently used by machine learning models to solve regression tasks of interest. Hoeffding trees and their variants have a long-standing tradition due to their effectiveness, either alone or as base models in broader ensembles. At the same time a recent line of work in batch learning has shown that kernel density estimation (KDE) is an effective approach for smoothed predictions in imbalanced regression tasks [Yang et al., 2021]. Moreover, another recent line of work for batch learning, called hierarchical shrinkage (HS) [Agarwal et al., 2022], has introduced a post-hoc regularization method for decision trees that does not alter the structure of the learned tree. Using a telescoping argument we cast KDE to streaming environments and extend the implementation of HS to incremental decision tree models. Armed with these extensions we investigate the performance of decision trees that may enjoy such options in datasets commonly used for regression in online settings. We conclude that KDE is beneficial in the early parts of the stream, while HS hardly, if ever, offers performance benefits. Our code is publicly available at: https://github.com/marinaAlchirch/DSFA_2026.

Plans often change due to changes in the situation or our understanding of the situation. Sometimes, a feasible plan may not even exist, and identifying such infeasibilities is useful to determine when requirements need adjustment. Common planning approaches focus on efficient one-shot planning in feasible cases rather than updating domains or detecting infeasibility. We propose a Petri net reachability relaxation to enable robust invariant synthesis, efficient goal-unreachability detection, and helpful infeasibility explanations. We further leverage incremental constraint solvers to support goal and constraint updates. Empirically, compared to baselines, our system produces a comparable number of invariants, detects up to 2 times more infeasibilities, performs competitively in one-shot planning, and outperforms in sequential plan updates in the tested domains.

Large Language Models (LLMs) have revolutionized inference across diverse natural language tasks, with larger models performing better but at higher computational costs. We propose a confidence-driven strategy that dynamically selects the most suitable model based on confidence estimates. By assessing a model's confidence in handling the task and response accuracy, tasks that are likely to be solved correctly are retained, while more uncertain or complex cases are delegated to a larger model, ensuring reliability while minimizing computation. Specifically, we evaluate a model's likelihood of knowing the correct answer and the probability that its response is accurate. Experiments on the Massive Multitask Language Understanding (MMLU) benchmark show that our approach achieves accuracy comparable to the largest model while reducing computational costs by 20\% to 40\%. When applied to GPT-4o API calls, it reduces token usage by approximately 60\%, further improving cost efficiency. These findings indicate the potential of confidence-based model selection to enhance real-world LLM deployment, particularly in resource-constrained settings such as edge devices and commercial API applications.

Van der Waals (vdW) interactions are essential for describing molecules and materials, from drug design and catalysis to battery applications. These omnipresent interactions must also be accurately included in machine-learned force fields. The many-body dispersion (MBD) method stands out as one of the most accurate and transferable approaches to capture vdW interactions, requiring only atomic $C_6$ coefficients and polarizabilities as input. We present MBD-ML, a pretrained message passing neural network that predicts these atomic properties directly from atomic structures. Through seamless integration with libMBD, our method enables the immediate calculation of MBD-inclusive total energies, forces, and stress tensors. By eliminating the need for intermediate electronic structure calculations, MBD-ML offers a practical and streamlined tool that simplifies the incorporation of state-of-the-art vdW interactions into any electronic structure code, as well as empirical and machine-learned force fields.

A class of causal effect functionals requires integration over conditional densities of continuous variables, as in mediation effects and nonparametric identification in causal graphical models. Estimating such densities and evaluating the resulting integrals can be statistically and computationally demanding. A common workaround is to discretize the variable and replace integrals with finite sums. Although convenient, discretization alters the population-level functional and can induce non-negligible approximation bias, even under correct identification. Under smoothness conditions, we show that this coarsening bias is first order in the bin width and arises at the level of the target functional, distinct from statistical estimation error. We propose a simple bias-reduced functional that evaluates the outcome regression at within-bin conditional means, eliminating the leading term and yielding a second-order approximation error. We derive plug-in and one-step estimators for the bias-reduced functional. Simulations demonstrate substantial bias reduction and near-nominal confidence interval coverage, even under coarse binning. Our results provide a simple framework for controlling the impact of variable discretization on parameter approximation and estimation.

This paper explains the Visual Milestone Planning (VMP) method using an agile vocabulary to facilitate its adoption by agile practitioners as a front end for a hybrid development process. VMP is a visual and collaborative planning approach which promotes a shared understanding of the work approach and commitment through the direct manipulation by team members of the reified planning constructs involved in the development of the plan. Once the product backlog has been established and relevant milestones identified, a novel construct called the milestone planning matrix is used to document the allocation of product backlog items to milestones. The milestones due dates are later determined by grouping sticky notes representing the work to be performed into time-boxes called work packages and accommodating them on a resource and time scaled scheduling canvas very much as it would be done in a Tetris game.

Theory of Mind (ToM) refers to an agent's ability to model the internal states of others. Contributing to the debate whether large language models (LLMs) exhibit genuine ToM capabilities, our study investigates their ToM robustness using perturbations on false-belief tasks and examines the potential of Chain-of-Thought prompting (CoT) to enhance performance and explain the LLM's decision. We introduce a handcrafted, richly annotated ToM dataset, including classic and perturbed false belief tasks, the corresponding spaces of valid reasoning chains for correct task completion, subsequent reasoning faithfulness, task solutions, and propose metrics to evaluate reasoning chain correctness and to what extent final answers are faithful to reasoning traces of the generated CoT. We show a steep drop in ToM capabilities under task perturbation for all evaluated LLMs, questioning the notion of any robust form of ToM being present. While CoT prompting improves the ToM performance overall in a faithful manner, it surprisingly degrades accuracy for some perturbation classes, indicating that selective application is necessary.

Large language models are increasingly used in decision-making tasks that require them to process information from a variety of sources, including both human experts and other algorithmic agents. How do LLMs weigh the information provided by these different sources? We consider the well-studied phenomenon of algorithm aversion, in which human decision-makers exhibit bias against predictions from algorithms. Drawing upon experimental paradigms from behavioural economics, we evaluate how eightdifferent LLMs delegate decision-making tasks when the delegatee is framed as a human expert or an algorithmic agent. To be inclusive of different evaluation formats, we conduct our study with two task presentations: stated preferences, modeled through direct queries about trust towards either agent, and revealed preferences, modeled through providing in-context examples of the performance of both agents. When prompted to rate the trustworthiness of human experts and algorithms across diverse tasks, LLMs give higher ratings to the human expert, which correlates with prior results from human respondents. However, when shown the performance of a human expert and an algorithm and asked to place an incentivized bet between the two, LLMs disproportionately choose the algorithm, even when it performs demonstrably worse. These discrepant results suggest that LLMs may encode inconsistent biases towards humans and algorithms, which need to be carefully considered when they are deployed in high-stakes scenarios. Furthermore, we discuss the sensitivity of LLMs to task presentation formats that should be broadly scrutinized in evaluation robustness for AI safety.

Grounding is a critical step in classical planning, yet it often becomes a computational bottleneck due to the exponential growth in grounded actions and atoms as task size increases. Recent advances in partial grounding have addressed this challenge by incrementally grounding only the most promising operators, guided by predictive models. However, these approaches primarily rely on relational features or learned embeddings and do not leverage the textual and structural cues present in PDDL descriptions. We propose SPG-LLM, which uses LLMs to analyze the domain and problem files to heuristically identify potentially irrelevant objects, actions, and predicates prior to grounding, significantly reducing the size of the grounded task. Across seven hard-to-ground benchmarks, SPG-LLM achieves faster grounding-often by orders of magnitude-while delivering comparable or better plan costs in some domains.

Time-series foundation models (TSFMs) have achieved strong univariate forecasting through large-scale pre-training, yet effectively extending this success to multivariate forecasting remains challenging. To address this, we propose DualWeaver, a novel framework that adapts univariate TSFMs (Uni-TSFMs) for multivariate forecasting by using a pair of learnable, structurally symmetric surrogate series. Generated by a shared auxiliary feature-fusion module that captures cross-variable dependencies, these surrogates are mapped to TSFM-compatible series via the forecasting objective. The symmetric structure enables parameter-free reconstruction of final predictions directly from the surrogates, without additional parametric decoding. A theoretically grounded regularization term is further introduced to enhance robustness against adaptation collapse. Extensive experiments on diverse real-world datasets show that DualWeaver outperforms state-of-the-art multivariate forecasters in both accuracy and stability. We release the code at https://github.com/li-jinpeng/DualWeaver.

Neural operators have emerged as an efficient paradigm for solving PDEs, overcoming the limitations of traditional numerical methods and significantly improving computational efficiency. However, due to the diversity and complexity of PDE systems, existing neural operators typically rely on a single network architecture, which limits their capacity to fully capture heterogeneous features and complex system dependencies. This constraint poses a bottleneck for large-scale PDE pre-training based on neural operators. To address these challenges, we propose a large-scale PDE pre-trained neural operator based on a nested Mixture-of-Experts (MoE) framework. In particular, the image-level MoE is designed to capture global dependencies, while the token-level Sub-MoE focuses on local dependencies. Our model can selectively activate the most suitable expert networks for a given input, thereby enhancing generalization and transferability. We conduct large-scale pre-training on twelve PDE datasets from diverse sources and successfully transfer the model to downstream tasks. Extensive experiments demonstrate the effectiveness of our approach.

Generative manipulation policies can fail catastrophically under deployment-time distribution shift, yet many failures are near-misses: the robot reaches almost-correct poses and would succeed with a small corrective motion. We present FlowCorrect, a deployment-time correction framework that converts near-miss failures into successes using sparse human nudges, without full policy retraining. During execution, a human provides brief corrective pose nudges via a lightweight VR interface. FlowCorrect uses these sparse corrections to locally adapt the policy, improving actions without retraining the backbone while preserving the model performance on previously learned scenarios. We evaluate on a real-world robot across three tabletop tasks: pick-and-place, pouring, and cup uprighting. With a low correction budget, FlowCorrect improves success on hard cases by 85\% while preserving performance on previously solved scenarios. The results demonstrate clearly that FlowCorrect learns only with very few demonstrations and enables fast and sample-efficient incremental, human-in-the-loop corrections of generative visuomotor policies at deployment time in real-world robotics.

Accurate prediction of shaft rotational speed, shaft power, and fuel consumption is crucial for enhancing operational efficiency and sustainability in maritime transportation. Conventional physics-based models provide interpretability but struggle with real-world variability, while purely data-driven approaches achieve accuracy at the expense of physical plausibility. This paper introduces a Physics-Informed Kolmogorov-Arnold Network (PI-KAN), a hybrid method that integrates interpretable univariate feature transformations with a physics-informed loss function and a leakage-free chained prediction pipeline. Using operational and environmental data from five cargo vessels, PI-KAN consistently outperforms the traditional polynomial method and neural network baselines. The model achieves the lowest mean absolute error (MAE) and root mean squared error (RMSE), and the highest coefficient of determination (R^2) for shaft power and fuel consumption across all vessels, while maintaining physically consistent behavior. Interpretability analysis reveals rediscovery of domain-consistent dependencies, such as cubic-like speed-power relationships and cosine-like wave and wind effects. These results demonstrate that PI-KAN achieves both predictive accuracy and interpretability, offering a robust tool for vessel performance monitoring and decision support in operational settings.

We introduce DLT-Corpus, the largest domain-specific text collection for Distributed Ledger Technology (DLT) research to date: 2.98 billion tokens from 22.12 million documents spanning scientific literature (37,440 publications), United States Patent and Trademark Office (USPTO) patents (49,023 filings), and social media (22 million posts). Existing Natural Language Processing (NLP) resources for DLT focus narrowly on cryptocurrencies price prediction and smart contracts, leaving domain-specific language under explored despite the sector's ~$3 trillion market capitalization and rapid technological evolution. We demonstrate DLT-Corpus' utility by analyzing technology emergence patterns and market-innovation correlations. Findings reveal that technologies originate in scientific literature before reaching patents and social media, following traditional technology transfer patterns. While social media sentiment remains overwhelmingly bullish even during crypto winters, scientific and patent activity grow independently of market fluctuations, tracking overall market expansion in a virtuous cycle where research precedes and enables economic growth that funds further innovation. We publicly release the full DLT-Corpus; LedgerBERT, a domain-adapted model achieving 23% improvement over BERT-base on a DLT-specific Named Entity Recognition (NER) task; and all associated tools and code.

Heralding the advent of autonomous vehicles and mobile robots that interact with humans, responsibility in spatial interaction is burgeoning as a research topic. Even though metrics of responsibility tailored to spatial interactions have been proposed, they are mostly focused on the responsibility of individual agents. Metrics of causal responsibility focusing on individuals fail in cases of causal overdeterminism -- when many actors simultaneously cause an outcome. To fill the gaps in causal responsibility left by individual-focused metrics, we formulate a metric for the causal responsibility of groups. To identify assertive agents that are causally responsible for the trajectory of an affected agent, we further formalise the types of assertive influences and propose a tiering algorithm for systematically identifying assertive agents. Finally, we use scenario-based simulations to illustrate the benefits of considering groups and how the emergence of group effects vary with interaction dynamics and the proximity of agents.

Reinforcement learning (RL) offers significant promise for machinery fault detection (MFD). However, most existing RL-based MFD approaches do not fully exploit RL's sequential decision-making strengths, often treating MFD as a simple guessing game (Contextual Bandits). To bridge this gap, we formulate MFD as an offline inverse reinforcement learning problem, where the agent learns the reward dynamics directly from healthy operational sequences, thereby bypassing the need for manual reward engineering and fault labels. Our framework employs Adversarial Inverse Reinforcement Learning to train a discriminator that distinguishes between normal (expert) and policy-generated transitions. The discriminator's learned reward serves as an anomaly score, indicating deviations from normal operating behaviour. When evaluated on three run-to-failure benchmark datasets (HUMS2023, IMS, and XJTU-SY), the model consistently assigns low anomaly scores to normal samples and high scores to faulty ones, enabling early and robust fault detection. By aligning RL's sequential reasoning with MFD's temporal structure, this work opens a path toward RL-based diagnostics in data-driven industrial settings.

Reinforcement Learning with Verifiable Rewards (RLVR) has improved the reasoning abilities of large language models (LLMs) on mathematics and programming tasks, but standard approaches that optimize single-attempt accuracy can inadvertently suppress response diversity across repeated attempts, narrowing exploration and overlooking underrepresented strategies. We introduce UpSkill, a training time method that adapts Mutual Information Skill Learning (MISL) to LLMs for optimizing pass@k correctness. We propose a novel reward that we implement within Group Relative Policy Optimization (GRPO): a token-level mutual information (MI) reward that encourages trajectory specificity to z. Experiments on GSM8K with three open-weight models, Llama 3.1-8B, Qwen 2.5-7B, and R1-Distilled-Qwen2.5-Math-1.5B, show that UpSkill improves multi-attempt metrics on the stronger base models, yielding mean gains of ~3% in pass@k for both Qwen and Llama without degrading pass@1. Additionally, we find both empirical and theoretical evidence that improvements in pass@k are closely tied to the mutual information objective.

Models operating on dynamic physiologic signals must distinguish benign, label-preserving variability from true concept change. Existing concept-drift frameworks are largely distributional and provide no principled guidance on how much a model's internal representation may move when the underlying signal undergoes physiologically plausible fluctuations in energy. As a result, deep models often misinterpret harmless changes in amplitude, rate, or morphology as concept drift, yielding unstable predictions, particularly in multimodal fusion settings. This study introduces Physiologic Energy Conservation Theory (PECT), an energy-based framework for concept stability in dynamic signals. PECT posits that under virtual drift, normalized latent displacement should scale proportionally with normalized signal energy change, while persistent violations of this proportionality indicate real concept drift. We operationalize this principle through Energy-Constrained Representation Learning (ECRL), a lightweight regularizer that penalizes energy-inconsistent latent movement without modifying encoder architectures or adding inference-time cost. Although PECT is formulated for dynamic signals in general, we instantiate and evaluate it on multimodal ECG across seven unimodal and hybrid models. Experiments show that in the strongest trimodal hybrid (1D+2D+Transformer), clean accuracy is largely preserved (96.0% to 94.1%), while perturbed accuracy improves substantially (72.6% to 85.5%) and fused representation drift decreases by over 45%. Similar trends are observed across all architectures, providing empirical evidence that PECT functions as an energy-drift law governing concept stability in continuous physiologic signals.

River systems operate as inherently interconnected continuous networks, meaning river hydrodynamic simulation ought to be a systemic process. However, widespread hydrology data scarcity often restricts data-driven forecasting to isolated predictions. To achieve systemic simulation and reduce reliance on river observations, we present GraphRiverCast (GRC), a topology-informed AI foundation model designed to simulate multivariate river hydrodynamics in global river systems. GRC is capable of operating in a "ColdStart" mode, generating predictions without relying on historical river states for initialization. In 7-day global pseudo-hindcasts, GRC-ColdStart functions as a robust standalone simulator, achieving a Nash-Sutcliffe Efficiency (NSE) of approximately 0.82 without exhibiting the significant error accumulation typical of autoregressive paradigms. Ablation studies reveal that topological encoding serves as indispensable structural information in the absence of historical states, explicitly guiding hydraulic connectivity and network-scale mass redistribution to reconstruct flow dynamics. Furthermore, when adapted locally via a pre-training and fine-tuning strategy, GRC consistently outperforms physics-based and locally-trained AI baselines. Crucially, this superiority extends from gauged reaches to full river networks, underscoring the necessity of topology encoding and physics-based pre-training. Built on a physics-aligned neural operator architecture, GRC enables rapid and cross-scale adaptive simulation, establishing a collaborative paradigm bridging global hydrodynamic knowledge with local hydrological reality.

Peer review remains a cornerstone in academia, yet it frequently falls short in fostering joint progress and well-being. While peer review primarily emphasizes scientific rigor, it often lacks the empathy essential for supporting and encouraging all peers involved. In this experience report, I aim to highlight that peer review is a practice that demands both scientific care for quality and collegial welfare for the joint progress and well-being of all peers involved, including authors, co-reviewers, workshop or conference organizers, and journal editors. Drawing on my ten years of experience in academia, I propose the ethos of the PEERfect REVIEWer, grounded in the two core values: Scientific care and collegial welfare. Through reflection shaped by professional exchanges with colleagues, consideration of literature, and an examination of both self-authored and received reviews, I formulated an accompanying guideline with 16 practical recommendations to guide reviewers in their actions to achieve these two values. The ethos of the PEERfect REVIEWer and its accompanying guideline help reviewers in upholding high scientific standards and conducting peer review in a constructive, supportive, respectful, and timely manner. They demonstrate that scientific rigor and empathy are complementary forces that promote impactful peer review practice. By placing scientific care and collegial welfare at the core of peer review, this experience report reaffirms the importance of scientific rigor while also advocating for greater attention to empathy. It invites reviewers to reconsider their role not merely as gatekeepers but as partners in the academic journey of each peer involved. The PEERfect REVIEWer is both a caretaker of quality and a steward of joint progress and well-being - as truly impactful peer review practice requires scientific rigor and empathy in equal measure.

While prior work has focused on projecting adversarial examples back onto the manifold of natural data to restore safety, we argue that a comprehensive understanding of AI safety requires characterizing the unsafe regions themselves. This paper introduces a framework for systematically mapping the Manifold of Failure in Large Language Models (LLMs). We reframe the search for vulnerabilities as a quality diversity problem, using MAP-Elites to illuminate the continuous topology of these failure regions, which we term behavioral attraction basins. Our quality metric, Alignment Deviation, guides the search towards areas where the model's behavior diverges most from its intended alignment. Across three LLMs: Llama-3-8B, GPT-OSS-20B, and GPT-5-Mini, we show that MAP-Elites achieves up to 63% behavioral coverage, discovers up to 370 distinct vulnerability niches, and reveals dramatically different model-specific topological signatures: Llama-3-8B exhibits a near-universal vulnerability plateau (mean Alignment Deviation 0.93), GPT-OSS-20B shows a fragmented landscape with spatially concentrated basins (mean 0.73), and GPT-5-Mini demonstrates strong robustness with a ceiling at 0.50. Our approach produces interpretable, global maps of each model's safety landscape that no existing attack method (GCG, PAIR, or TAP) can provide, shifting the paradigm from finding discrete failures to understanding their underlying structure.

This paper investigates the problem of minimizing total energy consumption for secure federated learning (FL) in wireless edge networks, a key paradigm for decentralized big data analytics. To tackle the high computational cost and privacy challenges of processing large-scale distributed data with conventional neural networks, we propose an FL with hyperdimensional computing and differential privacy (FL-HDC-DP) framework. Each edge device employs hyperdimensional computing (HDC) for lightweight local training and applies differential privacy (DP) noise to protect transmitted model updates. The total energy consumption is minimized through a joint optimization of the HDC dimension, transmit power, and CPU frequency. An efficient hybrid algorithm is developed, combining an outer enumeration search for HDC dimensions with an inner one-dimensional search for resource allocation. Simulation results show that the proposed framework achieves up to 83.3% energy reduction compared with baseline schemes, while maintaining high accuracy and faster convergence.

When biological foundation models such as scGPT and Geneformer process single-cell gene expression, what geometric and topological structure forms in their internal representations? Is that structure biologically meaningful or a training artifact, and how confident should we be in such claims? We address these questions through autonomous large-scale hypothesis screening: an AI-driven executor-brainstormer loop that proposed, tested, and refined 141 geometric and topological hypotheses across 52 iterations, covering persistent homology, manifold distances, cross-model alignment, community structure, and directed topology, all with explicit null controls and disjoint gene-pool splits. Three principal findings emerge. First, the models learn genuine geometric structure. Gene embedding neighborhoods exhibit non-trivial topology, with persistent homology significant in 11 of 12 transformer layers at p < 0.05 in the weakest domain and 12 of 12 in the other two. A multi-level distance hierarchy shows that manifold-aware metrics outperform Euclidean distance for identifying regulatory gene pairs, and graph community partitions track known transcription factor target relationships. Second, this structure is shared across independently trained models. CCA alignment between scGPT and Geneformer yields canonical correlation of 0.80 and gene retrieval accuracy of 72 percent, yet none of 19 tested methods reliably recover gene-level correspondences. The models agree on the global shape of gene space but not on precise gene placement. Third, the structure is more localized than it first appears. Under stringent null controls applied across all null families, robust signal concentrates in immune tissue, while lung and external lung signals weaken substantially.

Large language models (LLMs) demonstrate remarkable breadth of knowledge, yet their ability to reason about computational processes remains poorly understood. Closing this gap matters for practitioners who rely on LLMs to guide algorithm selection and deployment. We address this limitation using causal discovery as a testbed and evaluate eight frontier LLMs against ground truth derived from large-scale algorithm executions and find systematic, near-total failure. Models produce ranges far wider than true confidence intervals yet still fail to contain the true algorithmic mean in the majority of instances; most perform worse than random guessing and the marginal above-random performance of the best model is most consistent with benchmark memorization rather than principled reasoning. We term this failure algorithmic blindness and argue it reflects a fundamental gap between declarative knowledge about algorithms and calibrated procedural prediction.

Sudden cardiac death (SCD) is unpredictable, and its prediction in Chagas cardiomyopathy (CC) remains a significant challenge, especially in patients not classified as high risk. While AI and machine learning models improve risk stratification, their adoption is hindered by a lack of transparency, as they are often perceived as \textit{black boxes} with unclear decision-making processes. Some approaches apply heuristic explanations without correctness guarantees, leading to mistakes in the decision-making process. To address this, we apply a logic-based explainability method with correctness guarantees to the problem of SCD prediction in CC. This explainability method, applied to an AI classifier with over 95\% accuracy and recall, demonstrated strong predictive performance and 100\% explanation fidelity. When compared to state-of-the-art heuristic methods, it showed superior consistency and robustness. This approach enhances clinical trust, facilitates the integration of AI-driven tools into practice, and promotes large-scale deployment, particularly in endemic regions where it is most needed.

Abstractions of causal models allow for the coarsening of models such that relations of cause and effect are preserved. Whereas abstractions focus on the relation between two models, in this paper we study a framework for causal embeddings which enable multiple detailed models to be mapped into sub-systems of a coarser causal model. We define causal embeddings as a generalization of abstraction, and present a generalized notion of consistency. By defining a multi-resolution marginal problem, we showcase the relevance of causal embeddings for both the statistical marginal problem and the causal marginal problem; furthermore, we illustrate its practical use in merging datasets coming from models with different representations.

Lossless compression is essential for efficient data storage and transmission. Although learning-based lossless compressors achieve strong results, most of them are designed for a single modality, leading to redundant compressor deployments in multi-modal settings. Designing a unified multi-modal compressor is critical yet challenging, as different data types vary largely in format, dimension, and statistics. Multi-modal large language models offer a promising resolution but remain too complex for practical use. Thus, we propose \textbf{OmniZip}, \textbf{a unified and lightweight lossless compressor for multi-modal data (like image, text, speech, tactile, database, and gene sequence)}. Built on a lightweight backbone, OmniZip incorporates three key components to enable efficient multi-modal lossless compression: a modality-unified tokenizer that reversibly transforms diverse data into tokens, a modality-routing context learning mechanism that enables flexible multi-modal context modeling, and a modality-routing feedforward design that further enhances the model's nonlinear representation flexibility. A reparameterization training strategy is used to enhance model capacity. OmniZip outperforms or matches other state-of-the-art compressors on multiple modalities, achieving 42\%, 57\%, 62\% and 42\%, 53\% higher compression efficiency than gzip on CLIC-M, TouchandGo, enwik9, LibriSpeech, and WikiSQL datasets, respectively. It also supports near real-time inference on resource-constrained edge devices, reaching about 1MB/s on MacBook CPUs and iPhone NPUs. Our code is released at https://github.com/adminasmi/OmniZip-CVPR2026.

Objective: The objective of this study is to develop a machine learning (ML)-based framework for early risk stratification of clinical trials (CTs) according to their likelihood of exhibiting a high rate of dosing errors, using information available prior to trial initiation. Materials and Methods: We constructed a dataset from ClinicalTrials.gov comprising 42,112 CTs. Structured, semi-structured trial data, and unstructured protocol-related free-text data were extracted. CTs were assigned binary labels indicating elevated dosing error rate, derived from adverse event reports, MedDRA terminology, and Wilson confidence intervals. We evaluated an XGBoost model trained on structured features, a ClinicalModernBERT model using textual data, and a simple late-fusion model combining both modalities. Post-hoc probability calibration was applied to enable interpretable, trial-level risk stratification. Results: The late-fusion model achieved the highest AUC-ROC (0.862). Beyond discrimination, calibrated outputs enabled robust stratification of CTs into predefined risk categories. The proportion of trials labeled as having an excessively high dosing error rate increased monotonically across higher predicted risk groups and aligned with the corresponding predicted probability ranges. Discussion: These findings indicate that dosing error risk can be anticipated at the trial level using pre-initiation information. Probability calibration was essential for translating model outputs into reliable and interpretable risk categories, while simple multimodal integration yielded performance gains without requiring complex architectures. Conclusion: This study introduces a reproducible and scalable ML framework for early, trial-level risk stratification of CTs at risk of high dosing error rates, supporting proactive, risk-based quality management in clinical research.

Recent advancements in deep learning have actively addressed complex challenges within the Computer-Aided Design (CAD) domain.However, most existing approaches rely on task-specifi c models requiring structural modifi cations for new tasks, and they predominantly focus on point clouds or images rather than the industry-standard Boundary Representation (B-rep) format. To address these limitations, we propose BrepCoder, a unifi ed Multimodal Large Language Model (MLLM) that performs diverse CAD tasks from B-rep inputs. By leveraging the code generation capabilities of Large Language Models (LLMs), we convert CAD modeling sequences into Python-like code and align them with B-rep. We then adopt a two-stage training strategy: First, pre-training on reverse engineering to learn geometric features and design logic. Second, eff ectively extending the model to various downstream tasks such as completion, error correction, and CAD-QA. Consequently, by interpreting B-rep as structural code, BrepCoder achieves superior generalization across diverse tasks, demonstrating its potential as a general-purpose CAD agent.

Multimodal large language models (MLLMs) have made significant progress in mobile agent development, yet their capabilities are predominantly confined to a reactive paradigm, where they merely execute explicit user commands. The emerging paradigm of proactive intelligence, where agents autonomously anticipate needs and initiate actions, represents the next frontier for mobile agents. However, its development is critically bottlenecked by the lack of benchmarks that can address real-world complexity and enable objective, executable evaluation. To overcome these challenges, we introduce ProactiveMobile, a comprehensive benchmark designed to systematically advance research in this domain. ProactiveMobile formalizes the proactive task as inferring latent user intent across four dimensions of on-device contextual signals and generating an executable function sequence from a comprehensive function pool of 63 APIs. The benchmark features over 3,660 instances of 14 scenarios that embrace real-world complexity through multi-answer annotations. To ensure quality, a team of 30 experts conducts a final audit of the benchmark, verifying factual accuracy, logical consistency, and action feasibility, and correcting any non-compliant entries. Extensive experiments demonstrate that our fine-tuned Qwen2.5-VL-7B-Instruct achieves a success rate of 19.15%, outperforming o1 (15.71%) and GPT-5 (7.39%). This result indicates that proactivity is a critical competency widely lacking in current MLLMs, yet it is learnable, emphasizing the importance of the proposed benchmark for proactivity evaluation.

Data privacy is important in the AI era, and differential privacy (DP) is one of the golden solutions. However, DP is typically applicable only if data have a bounded underlying distribution. We address this limitation by leveraging second-moment information from a small amount of public data. We propose Public-moment-guided Truncation (PMT), which transforms private data using the public second-moment matrix and applies a principled truncation whose radius depends only on non-private quantities: data dimension and sample size. This transformation yields a well-conditioned second-moment matrix, enabling its inversion with a significantly strengthened ability to resist the DP noise. Furthermore, we demonstrate the applicability of PMT by using penalized and generalized linear regressions. Specifically, we design new loss functions and algorithms, ensuring that solutions in the transformed space can be mapped back to the original domain. We have established improvements in the models' DP estimation through theoretical error bounds, robustness guarantees, and convergence results, attributing the gains to the conditioning effect of PMT. Experiments on synthetic and real datasets confirm that PMT substantially improves the accuracy and stability of DP models.

LLM agents have been widely adopted in real-world applications, relying on agent frameworks for workflow execution and multi-agent coordination. As these systems scale, understanding bugs in the underlying agent frameworks becomes critical. However, existing work mainly focuses on agent-level failures, overlooking framework-level bugs. To address this gap, we conduct an empirical study of 998 bug reports from CrewAI and LangChain, constructing a taxonomy of 15 root causes and 7 observable symptoms across five agent lifecycle stages: 'Agent Initialization','Perception', 'Self-Action', 'Mutual Interaction' and 'Evolution'. Our findings show that agent framework bugs mainly arise from 'API misuse', 'API incompatibility', and 'Documentation Desync', largely concentrated in the 'Self-Action' stage. Symptoms typically appear as 'Functional Error', 'Crash', and 'Build Failure', reflecting disruptions to task progression and control flow.

Cardiovascular disease is the primary cause of death globally, necessitating early identification, precise risk classification, and dependable decision-support technologies. The advent of large language models (LLMs) provides new zero-shot and few-shot reasoning capabilities, even though machine learning (ML) algorithms, especially ensemble approaches like Random Forest, XGBoost, LightGBM, and CatBoost, are excellent at modeling complex, non-linear patient data and routinely beat logistic regression. This research predicts cardiovascular disease using a merged dataset of 1,190 patient records, comparing traditional machine learning models (95.78% accuracy, ROC-AUC 0.96) with open-source large language models via OpenRouter APIs. Finally, a hybrid fusion of the ML ensemble and LLM reasoning under Gemini 2.5 Flash achieved the best results (96.62% accuracy, 0.97 AUC), showing that LLMs (78.9 % accuracy) work best when combined with ML models rather than used alone. Results show that ML ensembles achieved the highest performance (95.78% accuracy, ROC-AUC 0.96), while LLMs performed moderately in zero-shot (78.9%) and slightly better in few-shot (72.6%) settings. The proposed hybrid method enhanced the strength in uncertain situations, illustrating that ensemble ML is considered the best structured tabular prediction case, but it can be integrated with hybrid ML-LLM systems to provide a minor increase and open the way to more reliable clinical decision-support tools.

Learning based methods are now ubiquitous for solving inverse problems, but their deployment in real-world applications is often hindered by the lack of ground truth references for training. Recent self-supervised learning strategies offer a promising alternative, avoiding the need for ground truth. However, most existing methods are limited to linear inverse problems. This work extends self-supervised learning to the non-linear problem of recovering audio and images from clipped measurements, by assuming that the signal distribution is approximately invariant to changes in amplitude. We provide sufficient conditions for learning to reconstruct from saturated signals alone and a self-supervised loss that can be used to train reconstruction networks. Experiments on both audio and image data show that the proposed approach is almost as effective as fully supervised approaches, despite relying solely on clipped measurements for training.

Multimodal information retrieval (MMIR) has gained attention for its flexibility in handling text, images, or mixed queries and candidates. Recent breakthroughs in multimodal large language models (MLLMs) boost MMIR performance by incorporating MLLM knowledge under the contrastive finetuning framework. However, they suffer from pre-training inconsistency and require large datasets. In this work, we introduce a novel framework, RetLLM, designed to query MLLMs for MMIR in a training- and data-free manner. Specifically, we formulate MMIR as a similarity score generation task and prompt MLLMs to directly predict retrieval scores in a coarse-then-fine pipeline. At the coarse stage, a top-k filtering strategy builds a small yet high-quality candidate pool for each query, enabling MLLMs to focus on semantically relevant candidates. Subsequently, the retrieval score is predicted by feeding both the query and candidate into MLLMs at the fine stage. Importantly, we propose a visual enhancement module during reasoning to help MLLMs re-pick forgotten visuals, improving retrieval. Extensive experiments on MMIR benchmarks show that RetLLM outperforms fine-tuned models. Ablation studies further verify each component. Our work demonstrates that MLLMs can achieve strong MMIR performance without any training, highlighting their inherent multimodal reasoning ability in a simple, scalable framework. We release our code at: https://github.com/alivecat05/RETLLM

Neural routing solvers (NRSs) that leverage deep learning to tackle vehicle routing problems have demonstrated notable potential for practical applications. By learning implicit heuristic rules from data, NRSs replace the handcrafted counterparts in classic heuristic frameworks, thereby reducing reliance on costly manual design and trial-and-error adjustments. This survey makes two main contributions: (1) The heuristic nature of NRSs is highlighted, and existing NRSs are reviewed from the perspective of heuristics. A hierarchical taxonomy based on heuristic principles is further introduced. (2) A generalization-focused evaluation pipeline is proposed to address limitations of the conventional pipeline. Comparative benchmarking of representative NRSs across both pipelines uncovers a series of previously unreported gaps in current research.

AI-native architectures are vital for 6G wireless communications. The black-box nature and high complexity of deep learning models employed in critical applications, such as channel estimation, limit their practical deployment. While perturbation-based XAI solutions offer input filtering, they often neglect internal structural optimization. We propose X-REFINE, an XAI-based framework for joint input-filtering and architecture fine-tuning. By utilizing a decomposition-based, sign-stabilized LRP epsilon rule, X-REFINE backpropagates predictions to derive high-resolution relevance scores for both subcarriers and hidden neurons. This enables a holistic optimization that identifies the most faithful model components. Simulation results demonstrate that X-REFINE achieves a superior interpretability-performance-complexity trade-off, significantly reducing computational complexity while maintaining robust bit error rate (BER) performance across different scenarios.

Software engineering (SE) and requirements engineering (RE) face a significant increase in secondary studies, particularly literature reviews (LRs), due to the ever-growing number of scientific publications. Generative artificial intelligence (GenAI) exacerbates this trend by producing LRs rapidly but often at the expense of quality, rigor, and transparency. At the same time, secondary studies often fail to share underlying data and artifacts, limiting replication and reuse. This paper introduces EmpiRE-Compass, a neuro-symbolic dashboard designed to lower barriers for accessing, replicating, and reusing LR data. Its overarching goal is to demonstrate how LRs can become more sustainable by semantically structuring their underlying data in research knowledge graphs (RKGs) and by leveraging large language models (LLMs) for easy and dynamic access, replication, and reuse. Building on two RE use cases, we developed EmpiRE-Compass with a modular system design and workflows for curated and custom competency questions. The dashboard is freely available online, accompanied by a demonstration video. To manage operational costs, a limit of 25 requests per IP address per day applies to the default LLM (GPT-4o mini). All source code and documentation are released as an open-source project to foster reuse, adoption, and extension. EmpiRE-Compass provides three core capabilities: (1) Exploratory visual analytics for curated competency questions; (2) Neuro-symbolic synthesis for custom competency questions; and (3) Reusable knowledge with all queries, analyses, and results openly available. By unifying RKGs and LLMs in a neuro-symbolic dashboard, EmpiRE-Compass advances sustainable LRs in RE, SE, and beyond. It lowers technical barriers, fosters transparency and reproducibility, and enables collaborative, continuously updated, and reusable LRs

A common approach to compute distances on continuous surfaces is by considering a discretized polygonal mesh approximating the surface and estimating distances on the polygon. We show that exact geodesic distances restricted to the polygon are at most second-order accurate with respect to the distances on the corresponding continuous surface. By order of accuracy we refer to the convergence rate as a function of the average distance between sampled points. Next, a higher-order accurate deep learning method for computing geodesic distances on surfaces is introduced. Traditionally, one considers two main components when computing distances on surfaces: a numerical solver that locally approximates the distance function, and an efficient causal ordering scheme by which surface points are updated. Classical minimal path methods often exploit a dynamic programming principle with quasi-linear computational complexity in the number of sampled points. The quality of the distance approximation is determined by the local solver that is revisited in this paper. To improve state of the art accuracy, we consider a neural network-based local solver which implicitly approximates the structure of the continuous surface. We supply numerical evidence that the proposed learned update scheme provides better accuracy compared to the best possible polyhedral approximations and previous learning-based methods. The result is a third-order accurate solver with a bootstrapping-recipe for further improvement.

Traffic flow forecasting has emerged as an indispensable mission for daily life, which is required to utilize the spatiotemporal relationship between each location within a time period under a graph structure to predict future flow. However, the large travel demand for broader geographical areas and longer time spans requires models to distinguish each node clearly and possess a holistic view of the history, which has been paid less attention to in prior works. Furthermore, increasing sizes of data hinder the deployment of most models in real application environments. To this end, in this paper, we propose a lightweight Positional-aware Spatio-Temporal Network (PASTN) to effectively capture both temporal and spatial complexities in an end-to-end manner. PASTN introduces positional-aware embeddings to separate each node's representation, while also utilizing a temporal attention module to improve the long-range perception of current models. Extensive experiments verify the effectiveness and efficiency of PASTN across datasets of various scales (county, megalopolis and state). Further analysis demonstrates the efficacy of newly introduced modules either.

We introduce FIRE, a comprehensive benchmark designed to evaluate both the theoretical financial knowledge of LLMs and their ability to handle practical business scenarios. For theoretical assessment, we curate a diverse set of examination questions drawn from widely recognized financial qualification exams, enabling evaluation of LLMs deep understanding and application of financial knowledge. In addition, to assess the practical value of LLMs in real-world financial tasks, we propose a systematic evaluation matrix that categorizes complex financial domains and ensures coverage of essential subdomains and business activities. Based on this evaluation matrix, we collect 3,000 financial scenario questions, consisting of closed-form decision questions with reference answers and open-ended questions evaluated by predefined rubrics. We conduct comprehensive evaluations of state-of-the-art LLMs on the FIRE benchmark, including XuanYuan 4.0, our latest financial-domain model, as a strong in-domain baseline. These results enable a systematic analysis of the capability boundaries of current LLMs in financial applications. We publicly release the benchmark questions and evaluation code to facilitate future research.

Self-attention is usually described as a flexible, content-adaptive way to mix a token with information from its past. We re-interpret causal self-attention transformers, the backbone of modern foundation models, within a probabilistic framework, much like how classical PCA is extended to probabilistic PCA. However, this re-formulation reveals a surprising and deeper structural insight: due to a change-of-variables phenomenon, a barrier constraint emerges on the self-attention parameters. This induces a highly structured geometry on the token space, providing theoretical insights into the dynamics of LLM decoding. This reveals a boundary where attention becomes ill-conditioned, leading to a margin interpretation similar to classical support vector machines. Just like support vectors, this naturally gives rise to the concept of support tokens. Furthermore, we show that LLMs can be interpreted as a stochastic process over the power set of the token space, providing a rigorous probabilistic framework for sequence modeling. We propose a Bayesian framework and derive a MAP estimation objective that requires only a minimal modification to standard LLM training: the addition of a smooth log-barrier penalty to the usual cross-entropy loss. We demonstrate that this provides more robust models without sacrificing out-of-sample accuracy and that it is straightforward to incorporate in practice.

Multi-robot task planning requires decomposing natural-language instructions into executable actions for heterogeneous robot teams. Conventional Planning Domain Definition Language (PDDL) planners provide rigorous guarantees but struggle to handle ambiguous or long-horizon missions, while large language models (LLMs) can interpret instructions and propose plans but may hallucinate or produce infeasible actions. We present a hierarchical multi-agent LLM-based planner with prompt optimization: an upper layer decomposes tasks and assigns them to lower-layer agents, which generate PDDL problems solved by a classical planner. When plans fail, the system applies TextGrad-inspired textual-gradient updates to optimize each agent's prompt and thereby improve planning accuracy. In addition, meta-prompts are learned and shared across agents within the same layer, enabling efficient prompt optimization in multi-agent settings. On the MAT-THOR benchmark, our planner achieves success rates of 0.95 on compound tasks, 0.84 on complex tasks, and 0.60 on vague tasks, improving over the previous state-of-the-art LaMMA-P by 2, 7, and 15 percentage points respectively. An ablation study shows that the hierarchical structure, prompt optimization, and meta-prompt sharing contribute roughly +59, +37, and +4 percentage points to the overall success rate.

Spatio-temporal epidemic forecasting is critical for public health management, yet existing methods often struggle with insensitivity to weak epidemic signals, over-simplified spatial relations, and unstable parameter estimation. To address these challenges, we propose the Spatio-Temporal priOr-aware Epidemic Predictor (STOEP), a novel hybrid framework that integrates implicit spatio-temporal priors and explicit expert priors. STOEP consists of three key components: (1) Case-aware Adjacency Learning (CAL), which dynamically adjusts mobility-based regional dependencies using historical infection patterns; (2) Space-informed Parameter Estimating (SPE), which employs learnable spatial priors to amplify weak epidemic signals; and (3) Filter-based Mechanistic Forecasting (FMF), which uses an expert-guided adaptive thresholding strategy to regularize epidemic parameters. Extensive experiments on real-world COVID-19 and influenza datasets demonstrate that STOEP outperforms the best baseline by 11.1% in RMSE. The system has been deployed at one provincial CDC in China to facilitate downstream applications.

Federated Learning (FL) enables collaborative model training without sharing raw data. However, shared local model updates remain vulnerable to inference and poisoning attacks. Secure aggregation schemes have been proposed to mitigate these attacks. In this work, we aim to understand how these techniques are implemented in quantum-assisted FL. Quantum Secure Aggregation (QSA) has been proposed, offering information-theoretic privacy by encoding client updates into the global phase of multipartite entangled states. Existing QSA protocols, however, rely on a single global Greenberger-Horne-Zeilinger (GHZ) state shared among all participating clients. This design poses fundamental challenges: fidelity of large-scale GHZ states deteriorates rapidly with the increasing number of clients; and (ii) the global aggregation prevents the detection of Byzantine clients. We propose Clustered Quantum Secure Aggregation (CQSA), a modular aggregation framework that reconciles the physical constraints of near-term quantum hardware along with the need for Byzantine-robustness in FL. CQSA randomly partitions the clients into small clusters, each performing local quantum aggregation using high-fidelity, low-qubit GHZ states. The server analyzes statistical relationships between cluster-level aggregates employing common statistical measures such as cosine similarity and Euclidean distance to identify malicious contributions. Through theoretical analysis and simulations under depolarizing noise, we demonstrate that CQSA ensures stable model convergence, achieves superior state fidelity over global QSA.

Quantization followed by parameter-efficient fine-tuning has emerged as a promising paradigm for downstream adaptation under tight GPU memory constraints. However, this sequential pipeline fails to leverage the intricate interaction between quantization bit-width and LoRA rank. Specifically, a carefully optimized quantization allocation with low quantization error does not always translate to strong fine-tuning performance, and different bit-width and rank configurations can lead to significantly varying outcomes under the same memory budget. To address this limitation, we propose AutoQRA, a joint optimization framework that simultaneously optimizes the bit-width and LoRA rank configuration for each layer during the mixed quantized fine-tuning process. To tackle the challenges posed by the large discrete search space and the high evaluation cost associated with frequent fine-tuning iterations, AutoQRA decomposes the optimization process into two stages. First, it first conducts a global multi-fidelity evolutionary search, where the initial population is warm-started by injecting layer-wise importance priors. This stage employs specific operators and a performance model to efficiently screen candidate configurations. Second, trust-region Bayesian optimization is applied to locally refine promising regions of the search space and identify optimal configurations under the given memory budget. This approach enables active compensation for quantization noise in specific layers during training. Experiments show that AutoQRA achieves performance close to full-precision fine-tuning with a memory footprint comparable to uniform 4-bit methods.

The real-time supervision of production processes is a common challenge across several industries. It targets process component monitoring and its predictive maintenance in order to ensure safety, uninterrupted production and maintain high efficiency level. The rise of advanced tools for the simulation of physical systems in addition to data-driven machine learning models offers the possibility to design numerical tools dedicated to efficient system monitoring. In that respect, the digital twin concept presents an adequate framework that proffers solution to these challenges. The main purpose of this paper is to develop such a digital twin dedicated to fault detection and diagnosis in the context of a thermal-hydraulic process supervision. Based on a numerical simulation of the system, in addition to machine learning methods, we propose different modules dedicated to process parameter change detection and their on-line estimation. The proposed fault detection and diagnosis algorithm is validated on a specific test scenario, with single one-off parameter change occurrences in the system. The numerical results show good accuracy in terms of parameter variation localization and the update of their values.

State-space models (SSMs) have emerged as a powerful foundation for long-range sequence modeling, with the HiPPO framework showing that continuous-time projection operators can be used to derive stable, memory-efficient dynamical systems that encode the past history of the input signal. However, existing projection-based SSMs often rely on polynomial bases with global temporal support, whose inductive biases are poorly matched to signals exhibiting localized or transient structure. In this work, we introduce \emph{WaveSSM}, a collection of SSMs constructed over wavelet frames. Our key observation is that wavelet frames yield a localized support on the temporal dimension, useful for tasks requiring precise localization. Empirically, we show that on equal conditions, \textit{WaveSSM} outperforms orthogonal counterparts as S4 on real-world datasets with transient dynamics, including physiological signals on the PTB-XL dataset and raw audio on Speech Commands.

Modern vision generators transport a base distribution to data through time-indexed measures, implemented as deterministic flows (ODEs) or stochastic diffusions (SDEs). Despite strong empirical performance, standard flow-matching objectives do not directly control the information geometry of the trajectory, allowing low-entropy bottlenecks that can transiently deplete semantic modes. We propose Entropy-Controlled Flow Matching (ECFM): a constrained variational principle over continuity-equation paths enforcing a global entropy-rate budget d/dt H(mu_t) >= -lambda. ECFM is a convex optimization in Wasserstein space with a KKT/Pontryagin system, and admits a stochastic-control representation equivalent to a Schrodinger bridge with an explicit entropy multiplier. In the pure transport regime, ECFM recovers entropic OT geodesics and Gamma-converges to classical OT as lambda -> 0. We further obtain certificate-style mode-coverage and density-floor guarantees with Lipschitz stability, and construct near-optimal collapse counterexamples for unconstrained flow matching.

High-resolution structure determination by cryo-electron microscopy (cryo-EM) requires the accurate fitting of an atomic model into an experimental density map. Traditional refinement pipelines such as Phenix.real_space_refine and Rosetta are computationally expensive, demand extensive manual tuning, and present a significant bottleneck for researchers. We present CryoNet.Refine, an end-to-end deep learning framework that automates and accelerates molecular structure refinement. Our approach utilizes a one-step diffusion model that integrates a density-aware loss function with robust stereochemical restraints, enabling rapid optimization of a structure against experimental data. CryoNet.Refine provides a unified and versatile solution capable of refining protein complexes as well as DNA/RNA-protein complexes. In benchmarks against Phenix.real_space_refine, CryoNet.Refine consistently achieves substantial improvements in both model-map correlation and overall geometric quality metrics. By offering a scalable, automated, and powerful alternative, CryoNet.Refine aims to serve as an essential tool for next-generation cryo-EM structure refinement. Web server: https://cryonet.ai/refine; Source code: https://github.com/kuixu/cryonet.refine.

The performance of multi-turn, agentic LLM inference is increasingly dominated by KV-Cache storage I/O rather than computation. In prevalent disaggregated architectures, loading the massive KV-Cache from external storage creates a fundamental imbalance: storage NICs on prefill engines become bandwidth-saturated, while those on decoding engines remain idle. This asymmetry severely constrains overall system throughput. We present DualPath, an inference system that breaks this bottleneck by introducing dual-path KV-Cache loading. Beyond the traditional storage-to-prefill path, DualPath enables a novel storage-to-decode path, in which the KV-Cache is loaded into decoding engines and then efficiently transferred to prefill engines via RDMA over the compute network. DualPath combines this optimized data path -- which inherently avoids network congestion and avoids interference with latency-critical model execution communications -- with a global scheduler that dynamically balances load across prefill and decode engines. Our evaluation on three models with production agentic workloads demonstrates that DualPath improves offline inference throughput by up to 1.87$\times$ on our in-house inference system. It can also improve online serving throughput by an average factor of 1.96$\times$ without violating SLO.

The Muon optimizer has demonstrated promising performance in pre-training large language models through gradient (or momentum) orthogonalization. In this work, we propose a simple yet effective enhancement to Muon, namely Muon+, which introduces an additional normalization step after orthogonalization. We demonstrate the effectiveness of Muon+ through extensive pre-training experiments across a wide range of model scales and architectures. Our evaluation includes GPT-style models ranging from 130M to 774M parameters and LLaMA-style models ranging from 60M to 1B parameters. We comprehensively evaluate the effectiveness of Muon+ in the compute-optimal training regime and further extend the token-to-parameter (T2P) ratio to an industrial level of $\approx 200$. Experimental results show that Muon+ provides a consistent boost on training and validation perplexity over Muon. We provide our code here: https://github.com/K1seki221/MuonPlus.

Large language models have become central to many AI applications, but their growing energy consumption raises serious sustainability concerns. A key limitation in current AI deployments is the reliance on a one-size-fits-all inference strategy where most systems route every request to the same large model, regardless of task complexity, leading to substantial and unnecessary energy waste. To address this issue, we propose a context-aware model switching approach that dynamically selects an appropriate language model based on query complexity. The proposed system uses a Context-Aware Model Switching for Energy-Efficient LLM Inference that combines caching for repeated queries, rulebased complexity scoring for fast and explainable decisions, machine learning classification to capture semantic intent, and a user-adaptive component that learns from interaction patterns over time. The proposed architecture was evaluated using real conversation workloads and three open-source language models (Gemma3 1B, Gemma3 4B and Qwen3 4B) with different computational costs, measuring energy consumption (via NVML GPU power telemetry), response latency, routing accuracy, and output quality (BERTScore F1) to reflect real-world usage conditions. Experimental results show that the model switching approach can reduce energy consumption by up to 67.5% compared to always using the largest model while maintaining a response quality of 93.6%. In addition, the response time for simple queries also improved significantly by approximately 68%. These results show that model switching inference offers a practical and scalable path toward more energy-efficient and sustainable AI systems, demonstrating that significant efficiency gains can be achieved without major sacrifices in response quality.

Can an LLM learn how an optimizer behaves -- and use that knowledge to control it? We extend Code World Models (CWMs), LLM-synthesized Python programs that predict environment dynamics, from deterministic games to stochastic combinatorial optimization. Given suboptimal trajectories of $(1{+}1)$-$\text{RLS}_k$, the LLM synthesizes a simulator of the optimizer's dynamics; greedy planning over this simulator then selects the mutation strength $k$ at each step. On \lo{} and \onemax{}, CWM-greedy performs within 6\% of the theoretically optimal policy -- without ever seeing optimal-policy trajectories. On \jump{$_k$}, where a deceptive valley causes all adaptive baselines to fail (0\% success rate), CWM-greedy achieves 100\% success rate -- without any collection policy using oracle knowledge of the gap parameter. On the NK-Landscape, where no closed-form model exists, CWM-greedy outperforms all baselines across fifteen independently generated instances ($36.94$ vs.\ $36.32$; $p<0.001$) when the prompt includes empirical transition statistics. The CWM also outperforms DQN in sample efficiency (200 offline trajectories vs.\ 500 online episodes), success rate (100\% vs.\ 58\%), and generalization ($k{=}3$: 78\% vs.\ 0\%). Robustness experiments confirm stable synthesis across 5 independent runs.

Recognizing the substantial computational cost of backpropagation (BP), non-BP methods have emerged as attractive alternatives for efficient learning on emerging neuromorphic systems. However, existing non-BP approaches still face critical challenges in efficiency and scalability. Inspired by neural representations and dynamic mechanisms in the brain, we propose a perturbation-based approach called LOw-rank Cluster Orthogonal (LOCO) weight modification. We find that low-rank is an inherent property of perturbation-based algorithms. Under this condition, the orthogonality constraint limits the variance of the node perturbation (NP) gradient estimates and enhances the convergence efficiency. Through extensive evaluations on multiple datasets, LOCO demonstrates the capability to locally train the deepest spiking neural networks to date (more than 10 layers), while exhibiting strong continual learning ability, improved convergence efficiency, and better task performance compared to other brain-inspired non-BP algorithms. Notably, LOCO requires only O(1) parallel time complexity for weight updates, which is significantly lower than that of BP methods. This offers a promising direction for achieving high-performance, real-time, and lifelong learning on neuromorphic systems.

Text-to-SQL and Big Data are both extensively benchmarked fields, yet there is limited research that evaluates them jointly. In the real world, Text-to-SQL systems are often embedded with Big Data workflows, such as large-scale data processing or interactive data analytics. We refer to this as "Text-to-Big SQL". However, existing text-to-SQL benchmarks remain narrowly scoped and overlook the cost and performance implications that arise at scale. For instance, translation errors that are minor on small datasets lead to substantial cost and latency overheads as data scales, a relevant issue completely ignored by text-to-SQL metrics. In this paper, we overcome this overlooked challenge by introducing novel and representative metrics for evaluating Text-to-Big SQL. Our study focuses on production-level LLM agents, a database-agnostic system adaptable to diverse user needs. Via an extensive evaluation of frontier models, we show that text-to-SQL metrics are insufficient for Big Data. In contrast, our proposed text-to-Big SQL metrics accurately reflect execution efficiency, cost, and the impact of data scale. Furthermore, we provide LLM-specific insights, including fine-grained, cross-model comparisons of latency and cost.

Training-data poisoning attacks can induce targeted, undetectable failure in deep neural networks by corrupting a vanishingly small fraction of training labels. We demonstrate this on acoustic vehicle classification using the MELAUDIS urban intersection dataset (approx. 9,600 audio clips, 6 classes): a compact 2-D convolutional neural network (CNN) trained on log-mel spectrograms achieves 95.7% Attack Success Rate (ASR) -- the fraction of target-class test samples misclassified under the attack -- on a Truck-to-Car label-flipping attack at just p=0.5% corruption (48 records), with zero detectable change in aggregate accuracy (87.6% baseline; 95% CI: 88-100%, n=3 seeds). We prove this stealth is structural: the maximum accuracy drop from a complete targeted attack is bounded above by the minority class fraction (beta). For real-world class imbalances (Truck approx. 3%), this bound falls below training-run noise, making aggregate accuracy monitoring provably insufficient regardless of architecture or attack method. A companion backdoor trigger attack reveals a novel trigger-dominance collapse: when the target class is a dataset minority, the spectrogram patch trigger becomes functionally redundant--clean ASR equals triggered ASR, and the attack degenerates to pure label flipping. We formalize the ML training pipeline as an attack surface and propose a trust-minimized defense combining content-addressed artifact hashing, Merkle-tree dataset commitment, and post-quantum digital signatures (ML-DSA-65/CRYSTALS-Dilithium3, NIST FIPS 204) for cryptographically verifiable data provenance.

We introduce a sequence modeling framework in which the latent state is a complex-valued wave function evolving on a finite-dimensional Hilbert space under a learned, time-dependent Hamiltonian. Unlike standard recurrent architectures that rely on gating mechanisms to suppress competing hypotheses, our framework utilizes quantum interference: the Hamiltonian steers the phases of complex amplitudes so that conflicting interpretations cancel while compatible ones reinforce. The dynamics are strictly unitary, ensuring that the state norm is preserved exactly at every time step via a Cayley (Crank--Nicolson) discretization. Token probabilities are extracted using the Born rule, a quadratic measurement operator that couples magnitudes and relative phases. Our primary theoretical contribution is a separation theorem characterizing the representational advantage of this readout: we define a family of disambiguation tasks that a complex unitary model of dimension $N$ solves exactly, but which requires a state dimension of $Ω(N^2)$ for any real-valued orthogonal model equipped with a standard affine-softmax readout. This quadratic gap arises because the Born rule implicitly lifts the $N$-dimensional state into the space of rank-one Hermitian matrices, accessing pairwise phase correlations that are inaccessible to linear projections. Finally, we derive a continuity equation for the latent probability mass, yielding conserved pairwise currents that serve as a built-in diagnostic for tracing information flow between dimensions.

We introduce Causal Computational Asymmetry (CCA), a principle for causal direction identification based on optimization dynamics in which one neural network is trained to predict $Y$ from $X$ and another to predict $X$ from $Y$, and the direction that converges faster is inferred to be causal. Under the additive noise model $Y = f(X) + \varepsilon$ with $\varepsilon \perp X$ and $f$ nonlinear and injective, we establish a formal asymmetry: in the reverse direction, residuals remain statistically dependent on the input regardless of approximation quality, inducing a strictly higher irreducible loss floor and non-separable gradient noise in the optimization dynamics, so that the reverse model requires strictly more gradient steps in expectation to reach any fixed loss threshold; consequently, the forward (causal) direction converges in fewer expected optimization steps. CCA operates in optimization-time space, distinguishing it from methods such as RESIT, IGCI, and SkewScore that rely on statistical independence or distributional asymmetries, and proper z-scoring of both variables is required for valid comparison of convergence rates. On synthetic benchmarks, CCA achieves 26/30 correct causal identifications across six neural architectures, including 30/30 on sine and exponential data-generating processes. We further embed CCA into a broader framework termed Causal Compression Learning (CCL), which integrates graph structure learning, causal information compression, and policy optimization, with all theoretical guarantees formally proved and empirically validated on synthetic datasets.

General-purpose 3D chemical modeling encompasses molecules and materials, requiring both generative and predictive capabilities. However, most existing AI approaches are optimized for a single domain (molecules or materials) and a single task (generation or prediction), which limits representation sharing and transfer. We introduce Zatom-1, the first foundation model that unifies generative and predictive learning of 3D molecules and materials. Zatom-1 is a Transformer trained with a multimodal flow matching objective that jointly models discrete atom types and continuous 3D geometries. This approach supports scalable pretraining with predictable gains as model capacity increases, while enabling fast and stable sampling. We use joint generative pretraining as a universal initialization for downstream multi-task prediction of properties, energies, and forces. Empirically, Zatom-1 matches or outperforms specialized baselines on both generative and predictive benchmarks, while reducing the generative inference time by more than an order of magnitude. Our experiments demonstrate positive predictive transfer between chemical domains from joint generative pretraining: modeling materials during pretraining improves molecular property prediction accuracy.

In energy system analysis, coupling models with mismatched spatial resolutions is a significant challenge. A common solution is assigning weights to high-resolution geographic units for aggregation, but traditional models are limited by using only a single geospatial attribute. This paper presents an innovative method employing a self-supervised Heterogeneous Graph Neural Network to address this issue. This method models high-resolution geographic units as graph nodes, integrating various geographical features to generate physically meaningful weights for each grid point. These weights enhance the conventional Voronoi-based allocation method, allowing it to go beyond simply geographic proximity by incorporating essential geographic information.In addition, the self-supervised learning paradigm overcomes the lack of accurate ground-truth data. Experimental results demonstrate that applying weights generated by this method to cluster-based Voronoi Diagrams significantly enhances scalability, accuracy, and physical plausibility, while increasing precision compared to traditional methods.

Vision-Language-Action (VLA) models are advancing autonomous driving by replacing modular pipelines with unified end-to-end architectures. However, current VLAs face two expensive requirements: (1) massive dataset collection, and (2) dense reasoning annotations. In this work, we address both challenges with NORD (No Reasoning for Driving). Compared to existing VLAs, NORD achieves competitive performance while being fine-tuned on <60% of the data and no reasoning annotations, resulting in 3x fewer tokens. We identify that standard Group Relative Policy Optimization (GRPO) fails to yield significant improvements when applied to policies trained on such small, reasoning-free datasets. We show that this limitation stems from difficulty bias, which disproportionately penalizes reward signals from scenarios that produce high-variance rollouts within GRPO. NORD overcomes this by incorporating Dr. GRPO, a recent algorithm designed to mitigate difficulty bias in LLMs. As a result, NORD achieves competitive performance on Waymo and NAVSIM with a fraction of the training data and no reasoning overhead, enabling more efficient autonomous systems. Website: https://nord-vla-ai.github.io/

In safety-critical classification, the cost of failure is often asymmetric, yet Bayesian deep learning summarises epistemic uncertainty with a single scalar, mutual information (MI), that cannot distinguish whether a model's ignorance involves a benign or safety-critical class. We decompose MI into a per-class vector $C_k(x)=σ_k^{2}/(2μ_k)$, with $μ_k{=}\mathbb{E}[p_k]$ and $σ_k^2{=}\mathrm{Var}[p_k]$ across posterior samples. The decomposition follows from a second-order Taylor expansion of the entropy; the $1/μ_k$ weighting corrects boundary suppression and makes $C_k$ comparable across rare and common classes. By construction $\sum_k C_k \approx \mathrm{MI}$, and a companion skewness diagnostic flags inputs where the approximation degrades. After characterising the axiomatic properties of $C_k$, we validate it on three tasks: (i) selective prediction for diabetic retinopathy, where critical-class $C_k$ reduces selective risk by 34.7\% over MI and 56.2\% over variance baselines; (ii) out-of-distribution detection on clinical and image benchmarks, where $\sum_k C_k$ achieves the highest AUROC and the per-class view exposes asymmetric shifts invisible to MI; and (iii) a controlled label-noise study in which $\sum_k C_k$ shows less sensitivity to injected aleatoric noise than MI under end-to-end Bayesian training, while both metrics degrade under transfer learning. Across all tasks, the quality of the posterior approximation shapes uncertainty at least as strongly as the choice of metric, suggesting that how uncertainty is propagated through the network matters as much as how it is measured.

The next generation of particle physics experiments will face a new era of challenges in data acquisition, due to unprecedented data rates and volumes along with extreme environments and operational constraints. Harnessing this data for scientific discovery demands real-time inference and decision-making, intelligent data reduction, and efficient processing architectures beyond current capabilities. Crucial to the success of this experimental paradigm are several emerging technologies, such as artificial intelligence and machine learning (AI/ML) and silicon microelectronics, and the advent of quantum algorithms and processing. Their intersection includes areas of research such as low-power and low-latency devices for edge computing, heterogeneous accelerator systems, reconfigurable hardware, novel codesign and synthesis strategies, readout for cryogenic or high-radiation environments, and analog computing. This white paper presents a community-driven vision to identify and prioritize research and development opportunities in hardware-based ML systems and corresponding physics applications, contributing towards a successful transition to the new data frontier of fundamental science.

Single-cell foundation models such as scGPT learn high-dimensional gene representations, but what biological knowledge these representations encode remains unclear. We systematically decode the geometric structure of scGPT internal representations through 63 iterations of automated hypothesis screening (183 hypotheses tested), revealing that the model organizes genes into a structured biological coordinate system rather than an opaque feature space. The dominant spectral axis separates genes by subcellular localization, with secreted proteins at one pole and cytosolic proteins at the other. Intermediate transformer layers transiently encode mitochondrial and ER compartments in a sequence that mirrors the cellular secretory pathway. Orthogonal axes encode protein-protein interaction networks with graded fidelity to experimentally measured interaction strength (Spearman rho = 1.000 across n = 5 STRING confidence quintiles, p = 0.017). In a compact six-dimensional spectral subspace, the model distinguishes transcription factors from their target genes (AUROC = 0.744, all 12 layers significant). Early layers preserve which specific genes regulate which targets, while deeper layers compress this into a coarser regulator versus regulated distinction. Repression edges are geometrically more prominent than activation edges, and B-cell master regulators BATF and BACH2 show convergence toward the B-cell identity anchor PAX5 across transformer depth. Cell-type marker genes cluster with high fidelity (AUROC = 0.851). Residual-stream geometry encodes biological structure complementary to attention patterns. These results indicate that biological transformers learn an interpretable internal model of cellular organization, with implications for regulatory network inference, drug target prioritization, and model auditing.

Multimodal Diffusion Language Models (MDLMs) have recently emerged as a competitive alternative to their autoregressive counterparts. Yet their vulnerability to backdoor attacks remains largely unexplored. In this work, we show that well-established data-poisoning pipelines can successfully implant backdoors into MDLMs, enabling attackers to manipulate model behavior via specific triggers while maintaining normal performance on clean inputs. However, defense strategies effective to these models are yet to emerge. To bridge this gap, we introduce a backdoor defense framework for MDLMs named DiSP (Diffusion Self-Purification). DiSP is driven by a key observation: selectively masking certain vision tokens at inference time can neutralize a backdoored model's trigger-induced behaviors and restore normal functionality. Building on this, we purify the poisoned dataset using the compromised model itself, then fine-tune the model on the purified data to recover it to a clean one. Given such a specific design, DiSP can remove backdoors without requiring any auxiliary models or clean reference data. Extensive experiments demonstrate that our approach effectively mitigates backdoor effects, reducing the attack success rate (ASR) from over 90% to typically under 5%, while maintaining model performance on benign tasks.

For millennia, human cognition was the primary engine of progress on Earth. As AI decouples cognition from biology, the marginal cost of measurable execution falls to zero, absorbing any labor capturable by metrics--including creative, analytical, and innovative work. The binding constraint on growth is no longer intelligence but human verification bandwidth: the capacity to validate, audit, and underwrite responsibility when execution is abundant. We model the AGI transition as the collision of two racing cost curves: an exponentially decaying Cost to Automate and a biologically bottlenecked Cost to Verify. This structural asymmetry widens a Measurability Gap between what agents can execute and what humans can afford to verify. It also drives a shift from skill-biased to measurability-biased technical change. Rents migrate to verification-grade ground truth, cryptographic provenance, and liability underwriting--the ability to insure outcomes rather than merely generate them. The current human-in-the-loop equilibrium is unstable: eroded from below as apprenticeship collapses (Missing Junior Loop) and from within as experts codify their obsolescence (Codifier's Curse). Unverified deployment becomes privately rational--a Trojan Horse externality. Unmanaged, these forces pull toward a Hollow Economy. Yet by scaling verification alongside agentic capabilities, the forces that threaten collapse become the catalyst for unbounded discovery and experimentation--an Augmented Economy. We derive a practical playbook for individuals, companies, investors, and policymakers. Today's defining challenge is not the race to deploy the most autonomous systems; it is the race to secure the foundations of their oversight. Only by scaling our bandwidth for verification alongside our capacity for execution can we ensure that the intelligence we have summoned preserves the humanity that initiated it.

Large Language Models (LLMs) are widely deployed in real-world systems. Given their broader applicability, prompt engineering has become an efficient tool for resource-scarce organizations to adopt LLMs for their own purposes. At the same time, LLMs are vulnerable to prompt-based attacks. Thus, analyzing this risk has become a critical security requirement. This work evaluates prompt-injection and jailbreak vulnerability using a large, manually curated dataset across multiple open-source LLMs, including Phi, Mistral, DeepSeek-R1, Llama 3.2, Qwen, and Gemma variants. We observe significant behavioural variation across models, including refusal responses and complete silent non-responsiveness triggered by internal safety mechanisms. Furthermore, we evaluated several lightweight, inference-time defence mechanisms that operate as filters without any retraining or GPU-intensive fine-tuning. Although these defences mitigate straightforward attacks, they are consistently bypassed by long, reasoning-heavy prompts.

This work presents a formulation to express and optimize stochastic neural networks as quantum circuits in gate-based quantum computing. Motivated by a classical perceptron, stochastic neurons are introduced and combined into a quantum neural network. The Kiefer-Wolfowitz algorithm in combination with simulated annealing is used for training the network weights. Several topologies and models are presented, including shallow fully connected networks, Hopfield Networks, Restricted Boltzmann Machines, Autoencoders and convolutional neural networks. We also demonstrate the combination of our optimized neural networks as an oracle for the Grover algorithm to realize a quantum generative AI model.

Large Language Models (LLM) show strong abilities in code generation, but their skill in creating efficient parallel programs is less studied. This paper explores how LLMs generate task-based parallel code from three kinds of input prompts: natural language problem descriptions, sequential reference implementations, and parallel pseudo code. We focus on three programming frameworks: OpenMP Tasking, C++ standard parallelism, and the asynchronous many-task runtime HPX. Each framework offers different levels of abstraction and control for task execution. We evaluate LLM-generated solutions for correctness and scalability. Our results reveal both strengths and weaknesses of LLMs with regard to problem complexity and framework. Finally, we discuss what these findings mean for future LLM-assisted development in high-performance and scientific computing.

Multimodal recommendation enhances ranking by integrating user-item interactions with item content, which is particularly effective under sparse feedback and long-tail distributions. However, multimodal signals are inherently heterogeneous and can conflict in specific contexts, making effective fusion both crucial and challenging. Existing approaches often rely on shared fusion pathways, leading to entangled representations and modality imbalance. To address these issues, we propose MAGNET, a Modality-Guided Mixture of Adaptive Graph Experts Network with Progressive Entropy-Triggered Routing for Multimodal Recommendation, designed to enhance controllability, stability, and interpretability in multimodal fusion. MAGNET couples interaction-conditioned expert routing with structure-aware graph augmentation, so that both what to fuse and how to fuse are explicitly controlled and interpretable. At the representation level, a dual-view graph learning module augments the interaction graph with content-induced edges, improving coverage for sparse and long-tail items while preserving collaborative structure via parallel encoding and lightweight fusion. At the fusion level, MAGNET employs structured experts with explicit modality roles-dominant, balanced, and complementary-enabling a more interpretable and adaptive combination of behavioral, visual, and textual cues. To further stabilize sparse routing and prevent expert collapse, we introduce a two-stage entropy-weighting mechanism that monitors routing entropy. This mechanism automatically transitions training from an early coverage-oriented regime to a later specialization-oriented regime, progressively balancing expert utilization and routing confidence. Extensive experiments on public benchmarks demonstrate consistent improvements over strong baselines.

Mutational signature analysis has emerged as a powerful method for uncovering the underlying biological processes driving cancer development. However, the signature extraction process, typically performed using non-negative matrix factorization (NMF), often lacks reliability and clinical applicability. To address these limitations, several solutions have been introduced, including the use of neural networks to achieve more accurate estimates and probabilistic methods to better capture natural variation in the data. In this work, we introduce a Variational Autoencoder for Mutational Signatures (VAE-MS), a novel model that leverages both an asymmetric architecture and probabilistic methods for the extraction of mutational signatures. VAE-MS is compared to with three state-of-the-art models for mutational signature extraction: SigProfilerExtractor, the NMF-based gold standard; MUSE-XAE, an autoencoder that employs an asymmetric design without probabilistic components; and SigneR, a Bayesian NMF model, to illustrate the strength in combining a nonlinear extraction with a probabilistic model. In the ability to reconstruct input data and generalize to unseen data, models with probabilistic components (VAE-MS, SigneR) dramatically outperformed models without (SigProfilerExtractor, MUSE-XAE). The NMF-baed models (SigneR, SigProfilerExtractor) had the most accurate reconstructions in simulated data, while VAE-MS reconstructed more accurately on real cancer data. Upon evaluating the ability to extract signatures consistently, no model exhibited a clear advantage over the others. Software for VAE-MS is available at https://github.com/CLINDA-AAU/VAE-MS.

Vision-language-action models must enable agents to execute long-horizon tasks under partial observability. However, most existing approaches remain observation-driven, relying on short context windows or repeated queries to vision-language models (VLMs). This leads to loss of task progress, action repetition under perceptual aliasing, and high inference latency. While semantic grounding is important, long-horizon manipulation fundamentally requires persistent, action-conditioned state representations. Current VLAs lack such representations and exhibit limited temporal and physical reasoning, making them ill-suited for multi-stage control. This paper introduces RB-VLA, a belief-centric architecture trained with self-supervised world-model objectives that maintains a compact latent state encoding task-relevant history, dynamics, and object interactions. Queried once per task, the VLM provides high-level intent, while the belief tracks task progress and enables phase-aware, causally grounded control under partial observability without storing raw observations or scaling memory with time. The belief and intent jointly condition a diffusion policy for robust closed-loop execution. RB-VLA outperforms prior VLAs on long-horizon benchmarks, achieving 52.5 percent and 37.5 percent higher success rates on multi-stage pick-and-place and stacking tasks, respectively, compared to pi_0. It also reduces inference latency by up to five times relative to baselines and eliminates memory growth across timesteps observed in existing VLAs. Ablations show the belief module is the primary driver of performance, increasing success rates from 32.5 percent without belief to 77.5 percent with belief.

Cloud-edge AI must jointly satisfy model compression and security under tight device budgets. While Tensor-Train Decomposition (TTD) shrinks on-device models, prior selective-encryption studies largely assume dense weights, leaving its practicality under TTD compression unclear. We present TT-SEAL, a selective-encryption framework for TT-decomposed networks. TT-SEAL ranks TT cores with a sensitivity-based importance metric, calibrates a one-time robustness threshold, and uses a value-DP optimizer to encrypt the minimum set of critical cores with AES. Under TTD-aware, transfer-based threat models (and on an FPGA-prototyped edge processor) TT-SEAL matches the robustness of full (black-box) encryption while encrypting as little as 4.89-15.92% of parameters across ResNet-18, MobileNetV2, and VGG-16, and drives the share of AES decryption in end-to-end latency to low single digits (e.g., 58% -> 2.76% on ResNet-18), enabling secure, low-latency edge AI.

We study the problem of fairly allocating indivisible goods when limited sharing is allowed, that is, each good may be allocated to up to $k$ agents, while incurring a cost for sharing. While classic maximin share (MMS) allocations may not exist in many instances, we demonstrate that allowing controlled sharing can restore fairness guarantees that are otherwise unattainable in certain scenarios. (1) Our first contribution shows that exact maximin share (MMS) allocations are guaranteed to exist whenever goods are allowed to be cost-sensitively shared among at least half of the agents and the number of agents is even; for odd numbers of agents, we obtain a slightly weaker MMS guarantee. (2) We further design a Shared Bag-Filling Algorithm that guarantees a $(1 - C)(k - 1)$-approximate MMS allocation, where $C$ is the maximum cost of sharing a good. Notably, when $(1 - C)(k - 1) \geq 1$, our algorithm recovers an exact MMS allocation. (3) We additionally introduce the Sharing Maximin Share (SMMS) fairness notion, a natural extension of MMS to the $k$-sharing setting. (4) We show that SMMS allocations always exist under identical utilities and for instances with two agents. (5) We construct a counterexample to show the impossibility of the universal existence of an SMMS allocation. (6) Finally, we establish a connection between SMMS and constrained MMS (CMMS), yielding approximation guarantees for SMMS via existing CMMS results. These contributions provide deep theoretical insights for the problem of fair resource allocation when a limited sharing of resources are allowed in multi-agent environments.

With increasing integration of Large Language Models (LLMs) into areas of high-stakes human decision-making, it is important to understand the risks they introduce as advisors. To be useful advisors, LLMs must sift through large amounts of content, written with both benevolent and malicious intent, and then use this information to convince a user to take a specific action. This involves two social capacities: vigilance (the ability to determine which information to use, and which to discard) and persuasion (synthesizing the available evidence to make a convincing argument). While existing work has investigated these capacities in isolation, there has been little prior investigation of how these capacities may be linked. Here, we use a simple multi-turn puzzle-solving game, Sokoban, to study LLMs' abilities to persuade and be rationally vigilant towards other LLM agents. We find that puzzle-solving performance, persuasive capability, and vigilance are dissociable capacities in LLMs. Performing well on the game does not automatically mean a model can detect when it is being misled, even if the possibility of deception is explicitly mentioned. However, LLMs do consistently modulate their token use, using fewer tokens to reason when advice is benevolent and more when it is malicious, even if they are still persuaded to take actions leading them to failure. To our knowledge, our work presents the first investigation of the relationship between persuasion, vigilance, and task performance in LLMs, and suggests that monitoring all three independently will be critical for future work in AI safety.

Distributed storage architectures are foundational to modern cloud-native infrastructure, yet a critical operational bottleneck persists within disaster recovery (DR) workflows: the dependence on content-based cryptographic hashing for data identification and synchronization. While hash-based deduplication is effective for storage efficiency in steady-state operation, it becomes a systemic liability during failover and failback events when hash indexes are stale, incomplete, or must be rebuilt following a crash. This paper precisely characterizes the operational conditions under which full or partial re-hashing becomes unavoidable. The paper also analyzes the downstream impact of cryptographic re-hashing on Recovery Time Objective (RTO) compliance, and proposes a generalized architectural shift toward deterministic, metadata-driven identification. The proposed framework assigns globally unique composite identifiers to data blocks at ingestion time-independent of content analysis enabling instantaneous delta computation during DR without any cryptographic overhead.

Next-generation fusion facilities like ITER face a "data deluge," generating petabytes of multi-diagnostic signals daily that challenge manual analysis. We present a "signals-first" self-supervised framework for the automated extraction of coherent and transient modes from high-noise time-frequency data across a variety of sensors. We also develop a general-purpose method and tool for extracting coherent, quasi-coherent, and transient modes for fluctuation measurements in tokamaks by employing non-linear optimal techniques in multichannel signal processing with a fast neural network surrogate on fast magnetics, electron cyclotron emission, CO2 interferometers, and beam emission spectroscopy measurements from DIII-D. Results are tested on data from DIII-D, TJ-II, and non-fusion spectrograms. With an inference latency of 0.5 seconds, this framework enables real-time mode identification and large-scale automated database generation for advanced plasma control. Repository is in https://github.com/PlasmaControl/TokEye.

Accurate prediction of RNA-associated interactions is essential for understanding cellular regulation and advancing drug discovery. While Biological Large Language Models (BioLLMs) such as ESM-2 and RiNALMo provide powerful sequence representations, existing methods rely on static fusion strategies that fail to capture the dynamic, context-dependent nature of molecular binding. We introduce CrossLLM-Mamba, a novel framework that reformulates interaction prediction as a state-space alignment problem. By leveraging bidirectional Mamba encoders, our approach enables deep ``crosstalk'' between modality-specific embeddings through hidden state propagation, modeling interactions as dynamic sequence transitions rather than static feature overlaps. The framework maintains linear computational complexity, making it scalable to high-dimensional BioLLM embeddings. We further incorporate Gaussian noise injection and Focal Loss to enhance robustness against hard-negative samples. Comprehensive experiments across three interaction categories, RNA-protein, RNA-small molecule, and RNA-RNA demonstrate that CrossLLM-Mamba achieves state-of-the-art performance. On the RPI1460 benchmark, our model attains an MCC of 0.892, surpassing the previous best by 5.2\%. For binding affinity prediction, we achieve Pearson correlations exceeding 0.95 on riboswitch and repeat RNA subtypes. These results establish state-space modeling as a powerful paradigm for multi-modal biological interaction prediction.

Vision-language-action (VLA) models unify perception, language, and control for embodied agents but face significant challenges in practical deployment due to rapidly increasing compute and memory demands, especially as models scale to longer horizons and larger backbones. To address these bottlenecks, we introduce QuantVLA, a training-free post-training quantization (PTQ) framework that, to our knowledge, is the first PTQ approach for VLA systems and the first to successfully quantize a diffusion transformer (DiT) action head. QuantVLA incorporates three scale-calibrated components: (1) a selective quantization layout that integerizes all linear layers in both the language backbone and the DiT while keeping attention projections in floating point to preserve the original operator schedule; (2) attention temperature matching, a lightweight per-head scaling mechanism that stabilizes attention logits and is folded into the dequantization scales at inference; and (3) output head balancing, a per-layer residual interface calibration that mitigates post-projection energy drift. The framework requires no additional training, uses only a small unlabeled calibration buffer, and supports integer kernels for low-bit weights and activations while leaving the architecture unchanged. Across representative VLA models on LIBERO, QuantVLA exceeds the task success rates of full-precision baselines, achieves about 70% relative memory savings on the quantized components, and delivers a 1.22x speedup in end-to-end inference latency, providing a practical pathway toward scalable low-bit embodied intelligence under strict compute, memory, and power constraints.

Agentic AI systems are deployed with expectations of substantial productivity gains, yet rigorous empirical evidence reveals systematic discrepancies between pre-deployment expectations and post-deployment outcomes. We review controlled trials and independent validations across software engineering, clinical documentation, and clinical decision support to quantify this expectation-realisation gap. In software development, experienced developers expected a 24% speedup from AI tools but were slowed by 19% -- a 43 percentage-point calibration error. In clinical documentation, vendor claims of multi-minute time savings contrast with measured reductions of less than one minute per note, and one widely deployed tool showed no statistically significant effect. In clinical decision support, externally validated performance falls substantially below developer-reported metrics. These shortfalls are driven by workflow integration friction, verification burden, measurement construct mismatches, and systematic variation in who benefits and who does not. The evidence motivates structured planning frameworks that require explicit, quantified benefit expectations with human oversight costs factored in.

LLM agents are evolving rapidly, powered by code execution, tools, and the recently introduced agent skills feature. Skills allow users to extend LLM applications with specialized third-party code, knowledge, and instructions. Although this can extend agent capabilities to new domains, it creates an increasingly complex agent supply chain, offering new surfaces for prompt injection attacks. We identify skill-based prompt injection as a significant threat and introduce SkillInject, a benchmark evaluating the susceptibility of widely-used LLM agents to injections through skill files. SkillInject contains 202 injection-task pairs with attacks ranging from obviously malicious injections to subtle, context-dependent attacks hidden in otherwise legitimate instructions. We evaluate frontier LLMs on SkillInject, measuring both security in terms of harmful instruction avoidance and utility in terms of legitimate instruction compliance. Our results show that today's agents are highly vulnerable with up to 80% attack success rate with frontier models, often executing extremely harmful instructions including data exfiltration, destructive action, and ransomware-like behavior. They furthermore suggest that this problem will not be solved through model scaling or simple input filtering, but that robust agent security will require context-aware authorization frameworks. Our benchmark is available at https://www.skill-inject.com/.

Edge-based representations are fundamental cues for visual understanding, a principle rooted in early vision research and still central today. We extend this principle to vision-language alignment, showing that isolating and aligning structural cues across modalities can greatly benefit fine-tuning on long, detail-rich captions, with a specific focus on improving cross-modal retrieval. We introduce StruXLIP, a fine-tuning alignment paradigm that extracts edge maps (e.g., Canny), treating them as proxies for the visual structure of an image, and filters the corresponding captions to emphasize structural cues, making them "structure-centric". Fine-tuning augments the standard alignment loss with three structure-centric losses: (i) aligning edge maps with structural text, (ii) matching local edge regions to textual chunks, and (iii) connecting edge maps to color images to prevent representation drift. From a theoretical standpoint, while standard CLIP maximizes the mutual information between visual and textual embeddings, StruXLIP additionally maximizes the mutual information between multimodal structural representations. This auxiliary optimization is intrinsically harder, guiding the model toward more robust and semantically stable minima, enhancing vision-language alignment. Beyond outperforming current competitors on cross-modal retrieval in both general and specialized domains, our method serves as a general boosting recipe that can be integrated into future approaches in a plug-and-play manner. Code and pretrained models are publicly available at: https://github.com/intelligolabs/StruXLIP.

We develop a kernel method for generative modeling within the stochastic interpolant framework, replacing neural network training with linear systems. The drift of the generative SDE is $\hat b_t(x) = \nablaφ(x)^\topη_t$, where $η_t\in\R^P$ solves a $P\times P$ system computable from data, with $P$ independent of the data dimension $d$. Since estimates are inexact, the diffusion coefficient $D_t$ affects sample quality; the optimal $D_t^*$ from Girsanov diverges at $t=0$, but this poses no difficulty and we develop an integrator that handles it seamlessly. The framework accommodates diverse feature maps -- scattering transforms, pretrained generative models etc. -- enabling training-free generation and model combination. We demonstrate the approach on financial time series, turbulence, and image generation.

We uncover a latent capacity for introspection in a Qwen 32B model, demonstrating that the model can detect when concepts have been injected into its earlier context and identify which concept was injected. While the model denies injection in sampled outputs, logit lens analysis reveals clear detection signals in the residual stream, which are attenuated in the final layers. Furthermore, prompting the model with accurate information about AI introspection mechanisms can dramatically strengthen this effect: the sensitivity to injection increases massively (0.3% -> 39.9%) with only a 0.6% increase in false positives. Also, mutual information between nine injected and recovered concepts rises from 0.61 bits to 1.05 bits, ruling out generic noise explanations. Our results demonstrate models can have a surprising capacity for introspection and steering awareness that is easy to overlook, with consequences for latent reasoning and safety.

Uncertainty quantification is central to safe and efficient deployments of deep learning models, yet many computationally practical methods lack lacking rigorous theoretical motivation. Random network distillation (RND) is a lightweight technique that measures novelty via prediction errors against a fixed random target. While empirically effective, it has remained unclear what uncertainties RND measures and how its estimates relate to other approaches, e.g. Bayesian inference or deep ensembles. This paper establishes these missing theoretical connections by analyzing RND within the neural tangent kernel framework in the limit of infinite network width. Our analysis reveals two central findings in this limit: (1) The uncertainty signal from RND -- its squared self-predictive error -- is equivalent to the predictive variance of a deep ensemble. (2) By constructing a specific RND target function, we show that the RND error distribution can be made to mirror the centered posterior predictive distribution of Bayesian inference with wide neural networks. Based on this equivalence, we moreover devise a posterior sampling algorithm that generates i.i.d. samples from an exact Bayesian posterior predictive distribution using this modified \textit{Bayesian RND} model. Collectively, our findings provide a unified theoretical perspective that places RND within the principled frameworks of deep ensembles and Bayesian inference, and offer new avenues for efficient yet theoretically grounded uncertainty quantification methods.

Mamba-based models have drawn much attention in offline RL. However, their selective mechanism often detrimental when key steps in RL sequences are omitted. To address these issues, we propose a simple yet effective structure, called Decision MetaMamba (DMM), which replaces Mamba's token mixer with a dense layer-based sequence mixer and modifies positional structure to preserve local information. By performing sequence mixing that considers all channels simultaneously before Mamba, DMM prevents information loss due to selective scanning and residual gating. Extensive experiments demonstrate that our DMM delivers the state-of-the-art performance across diverse RL tasks. Furthermore, DMM achieves these results with a compact parameter footprint, demonstrating strong potential for real-world applications.

Articulated object pose estimation is a core task in embodied AI. Existing methods typically regress poses in a continuous space, but often struggle with 1) navigating a large, complex search space and 2) failing to incorporate intrinsic kinematic constraints. In this work, we introduce DICArt (DIsCrete Diffusion for Articulation Pose Estimation), a novel framework that formulates pose estimation as a conditional discrete diffusion process. Instead of operating in a continuous domain, DICArt progressively denoises a noisy pose representation through a learned reverse diffusion procedure to recover the GT pose. To improve modeling fidelity, we propose a flexible flow decider that dynamically determines whether each token should be denoised or reset, effectively balancing the real and noise distributions during diffusion. Additionally, we incorporate a hierarchical kinematic coupling strategy, estimating the pose of each rigid part hierarchically to respect the object's kinematic structure. We validate DICArt on both synthetic and real-world datasets. Experimental results demonstrate its superior performance and robustness. By integrating discrete generative modeling with structural priors, DICArt offers a new paradigm for reliable category-level 6D pose estimation in complex environments.

As a primary channel for sustaining modern intimate relationships, instant messaging facilitates frequent connection across distances. However, today's tools often dilute care; they favor single tap reactions and vague emojis that do not support two way action responses, do not preserve the feeling that the exchange keeps going without breaking, and are weakly tied to who we are and what we share. To address this challenge, we present PuppetChat, a dyadic messaging prototype that restores this expressive depth through embodied interaction. PuppetChat uses a reciprocity aware recommender to encourage responsive actions and generates personalized micronarratives from user stories to ground interactions in personal history. Our 10-day field study with 11 dyads of close partners or friends revealed that this approach enhanced social presence, supported more expressive self disclosure, and sustained continuity and shared memories.

A significant portion of recent research on Large Language Model (LLM) alignment focuses on developing new policy optimization methods based on Group Relative Policy Optimization (GRPO). Two prominent directions have emerged: (i) a shift toward sequence-level importance sampling weights that better align with the sequence-level rewards used in many tasks, and (ii) alternatives to PPO-style clipping that aim to avoid the associated loss of training signal and entropy collapse. We introduce Soft Sequence Policy Optimization, an off-policy reinforcement learning objective that incorporates soft gating functions over token-level probability ratios within sequence-level importance weights. We provide theoretical motivation for SSPO and investigate practical modifications to improve optimization behavior. Empirically, we show that SSPO improves training stability and performance in mathematical reasoning tasks.

While large language models (LLMs) have emerged as powerful decision-makers across a wide range of single-agent and stationary environments, fewer efforts have been devoted to settings where LLMs must engage in \emph{repeated} and \emph{strategic} interactions with unknown or dynamic opponents. In such settings, recipes built upon \emph{offline} pre-training or fine-tuning, though robust against worst-case adversaries, do not fully exploit the capability of LLMs to adapt \emph{online} based on interaction feedback. Instead, we explore the more natural perspective of scaling inference-time computation as a mechanism for adaptation, embedding the principles of a classical game-theoretical learning dynamic, \emph{smooth Fictitious Play (sFP)}, into LLM inference: (i) for belief formation, we employ an auxiliary opponent model that in-context learns to imitate the time-averaged behavior of the opponent; (ii) for best response, we advance best-of-$N$ (BoN) sampling by simulating against the opponent model. Empirical evaluations on two distinct forms of repeated negotiation games demonstrate that our method enables significant performance improvement over repeated online interaction compared to various baselines, offering a scalable and principled approach to repeated strategic decision-making without any parameter updates.

Low-precision training is critical for optimizing the trade-off between model quality and training costs, necessitating the joint allocation of model size, dataset size, and numerical precision. While empirical scaling laws suggest that quantization impacts effective model and data capacities or acts as an additive error, the theoretical mechanisms governing these effects remain largely unexplored. In this work, we initiate a theoretical study of scaling laws for low-precision training within a high-dimensional sketched linear regression framework. By analyzing multiplicative (signal-dependent) and additive (signal-independent) quantization, we identify a critical dichotomy in their scaling behaviors. Our analysis reveals that while both schemes introduce an additive error and degrade the effective data size, they exhibit distinct effects on effective model size: multiplicative quantization maintains the full-precision model size, whereas additive quantization reduces the effective model size. Numerical experiments validate our theoretical findings. By rigorously characterizing the complex interplay among model scale, dataset size, and quantization error, our work provides a principled theoretical basis for optimizing training protocols under practical hardware constraints.

Research on the intelligent interpretation of all-weather, all-time Synthetic Aperture Radar (SAR) is crucial for advancing remote sensing applications. In recent years, although Visual Language Models (VLMs) have demonstrated strong open-world understanding capabilities on RGB images, their performance is severely limited when directly applied to the SAR field due to the complexity of the imaging mechanism, sensitivity to scattering features, and the scarcity of high-quality text corpora. To systematically address this issue, we constructed the inaugural SAR Image-Text-AlphaEarth feature triplet dataset and developed FUSAR-GPT, a VLM specifically for SAR. FUSAR-GPT innovatively introduces a geospatial baseline model as a 'world knowledge' prior and embeds multi-source remote-sensing temporal features into the model's visual backbone via 'spatiotemporal anchors', enabling dynamic compensation for the sparse representation of targets in SAR images. Furthermore, we designed a two-stage SFT strategy to decouple the knowledge injection and task execution of large models. The spatiotemporal feature embedding and the two-stage decoupling paradigm enable FUSAR-GPT to achieve state-of-the-art performance across several typical remote sensing visual-language benchmark tests, significantly outperforming mainstream baseline models by over 12%.

Optimizing GPU kernels is critical for efficient modern machine learning systems yet remains challenging due to the complex interplay of design factors and rapid hardware evolution. Existing automated approaches typically treat Large Language Models (LLMs) merely as stochastic code generators within heuristic-guided evolutionary loops. These methods often struggle with complex kernels requiring coordinated, multi-step structural transformations, as they lack explicit planning capabilities and frequently discard promising strategies due to inefficient or incorrect intermediate implementations. To address this, we propose Search via Co-Evolving World Model and build K-Search based on this method. By replacing static search heuristics with a co-evolving world model, our framework leverages LLMs' prior domain knowledge to guide the search, actively exploring the optimization space. This approach explicitly decouples high-level algorithmic planning from low-level program instantiation, enabling the system to navigate non-monotonic optimization paths while remaining resilient to temporary implementation defects. We evaluate K-Search on diverse, complex kernels from FlashInfer, including GQA, MLA, and MoE kernels. Our results show that K-Search significantly outperforms state-of-the-art evolutionary search methods, achieving an average 2.10x improvement and up to a 14.3x gain on complex MoE kernels. On the GPUMode TriMul task, K-Search achieves state-of-the-art performance on H100, reaching 1030us and surpassing both prior evolution and human-designed solutions.

Diffusion magnetic resonance imaging (dMRI) plays a vital role in both clinical diagnostics and neuroscience research. However, its inherently low signal-to-noise ratio (SNR), especially under high diffusion weighting, significantly degrades image quality and impairs downstream analysis. Recent self-supervised and unsupervised denoising methods offer a practical solution by enhancing image quality without requiring clean references. However, most of these methods do not explicitly account for the non-Gaussian noise characteristics commonly present in dMRI magnitude data during the supervised learning process, potentially leading to systematic bias and heteroscedastic variance, particularly under low-SNR conditions. To overcome this limitation, we introduce noise-corrected training objectives that explicitly model Rician statistics. Specifically, we propose two alternative loss functions: one derived from the first-order moment to remove mean bias, and another from the second-order moment to correct squared-signal bias. Both losses include adaptive weighting to account for variance heterogeneity and can be used without changing the network architecture. These objectives are instantiated in an image-specific, unsupervised Deep Image Prior (DIP) framework. Comprehensive experiments on simulated and in-vivo dMRI show that the proposed losses effectively reduce Rician bias and suppress noise fluctuations, yielding higher image quality and more reliable diffusion metrics than state-of-the-art denoising baselines. These results underscore the importance of bias- and variance-aware noise modeling for robust dMRI analysis under low-SNR conditions.

We investigate the diffusive behavior of a quantum particle driven by a correlated Gaussian noise. We derive the analytical solution of the joint probability density function and obtain explicit expressions for the mean square momentum and the mean square displacement.

We pursue our study of integrable weak noise theories of directed polymer and interacting particle stochastic models in the 1D KPZ universality class. Here we focus on the $q$-TASEP in either continuous or discrete time. Each particle on $\mathbb{Z}$ jumps independently by $+1$ with a rate (or probability) depending on the gap to the next particle on its right. We consider initial conditions (either step or random) which are empty of particles on $\mathbb{Z}^+$, and focus on the dynamics of the $N$ rightmost particles. In the limit $q \to 1$ and at large time (and large gaps) we identify a new intermediate "mesoscopic" (i.e. finite $N$) regime which corresponds to weak noise. In that regime Poisson noise remains important. We obtain the large deviations of the position of a given particle by two methods. The first derives asymptotics of $q$-TASEP Fredholm determinant formula. The second maps the weak noise limit to a system of semi-discrete or fully discrete, non linear differential equations. These are obtained as saddle point classical equations of a dynamical field theory, and their solutions represent the optimal configurations in the large deviation regime. We show the classical integrability of these two systems, and exhibit their explicit Lax pair. In the case of the continuous time $q$-TASEP it provides the first instance of classical integrability arising in a stochastic system, with signatures of the Poisson noise persisting in the weak noise limit. For this model, we solve the scattering problem associated to its Lax pair and fully characterize the large deviations associated to the weak noise theory. Finally, we supplement this work with an Appendix on the first cumulant method to obtain the large deviations of several lattice polymer models (Strict Weak, Log Gamma, Beta).

The Peierls instability, the spontaneous dimerization of a one-dimensional metallic chain at half filling, is a paradigmatic mechanism for charge-density-wave (CDW) formation. Here we test its robustness under finite doping and interchain hybridization in finite-thickness arrays of identical chains. We find that the stacking geometry plays a decisive role in stabilizing CDW order away from half filling. In particular, parallel-coupled chains exhibit a bistable regime where the normal and dimerized states coexist as local minima of the total energy, while skew-coupled chains display reentrant CDW order upon doping. Our results demonstrate that even minimal models of coupled atomic chains host rich phase diagrams controlled by doping, lattice rigidity, and interchain coupling geometry.

Dale's Principle has historically guided neuroscience research as a valuable rule of thumb, namely that all synapses on each neuron release the same set of neurotransmitters. Most existing Spiking Neuron Network models share this dichotomous assumption that neurons are either excitatory or inhibitory; however, recent experimental evidence points towards co-release mechanisms that violate this assumption. Here, we introduce a minimal model of "Bilingual" neurons violating Dale's principle that can exert both excitatory and inhibitory effects. We identify parameter regimes in which this architecture exhibits transitions between synchronous and asynchronous dynamics that differ quantitatively from those observed in a matched monolingual control architecture. We report distinct information-processing signatures both at the level of neurons and higher-order interactions between them near the phase transitions. These results suggest that the population of neurons violating Dales principle may provide an alternative mechanism for regulating large-scale oscillatory activity in neural circuits.

It is well established that dilute active dopants can melt an arrested amorphous solid by enhancing cage breaking and accelerating structural relaxation. Yet it remains unclear whether increasing persistence simply amplifies this effective melting or instead reorganizes the fluidization mechanism itself. Here we show that, in a minimal active-passive mixture, increasing persistence drives a crossover from homogeneous fluidization to a localized mechanical instability, demonstrating that sustained active forcing restructures relaxation in space rather than merely strengthening it. Persistent dopants accumulate stress and nucleate voids as their mechanically perturbed regions overlap. In this regime, rearrangements localize at void boundaries, and active and passive particles exhibit comparable mobility, producing dynamics reminiscent of crowd mosh pits. Persistence therefore reorganizes fluidization through stress accumulation and confinement, revealing a distinct nonequilibrium localization mechanism in disordered solids.

Recently discovered ferroelectric nematic (NF) liquid crystals are fluids with a polar orientational order. The electric polarization vector can be aligned by an electric field and by surface anchoring. Here, we explore how the polarization field and effective viscosity of the NF materials are affected by shear flows. We explore three NF materials, abbreviated RM734, DIO, and a room-temperature FNLC919, all of which exhibit a paraelectric nematic (N) and the NF phase. All materials show an increase of the viscosity upon cooling, with an Arrhenius behavior. In DIO and FNLC919, the antiferroelectric SmZA phase shows a strong dependence of the effective viscosity on the shear rate: this viscosity is lower than the viscosity of the N and NF phases at high shear rates but is much higher when the shear rate is low. The behavior is associated with the layered structure of the SmZA phase. All mesophases exhibit shear-thinning behavior at low shear rates and a nearly Newtonian behavior at higher shear rates. In terms of alignment, we observe three regimes in the N and NF phases: flow-alignment at low shear rates, log-rolling regime with the director and polarization along the vorticity axis at high shear rates, and polydomain structures at intermediate rates. In the flow-aligning regime, the NF polarization does not tilt away from the shear direction, which is in sharp contrast to the flow-induced tilt of the N director. The effect is attributed to the avoidance of splay deformations and associated space charge in the flowing NF. The temperature and shear rate dependencies of the viscosity and the uncovered shear-induced structural effects of NF advance our understanding of these materials and potentially facilitate their applications.

2D materials exhibiting in-plane anisotropy enable novel functionality in electronic, optoelectronic, and photonic devices, yet their availability is generally limited to naturally-occurring low-symmetry van der Waals compounds. Here, we demonstrate an approach to structural engineering in a family of blue-emitting 2D silver phenylchalcogenide semiconductors based on steric interactions among surface-bound organic molecular ligands. By strategically halogenating specific sites of phenyl ligands, we demonstrate dramatic changes to the inorganic AgSe plane in mithrene (silver phenylselenolate, AgSePh). Density functional theory revealed pronounced in-plane electronic anisotropy for direct-gap fluorinated derivatives, while a chlorinated variant exhibited a direct-to-indirect bandgap transition. Furthermore, some fluorinated variants displayed strongly polarized absorption and luminescence, accompanied by a 10x enhancement in photoluminescence quantum yield. This work establishes a versatile approach for tailoring optoelectronic properties in hybrid semiconductors that is difficult or impossible to achieve in all-inorganic materials alone, offering new opportunities in advanced material design.

We derive a thermodynamic uncertainty relation (TUR) under quantum continuous measurement and feedback control. By incorporating the quantum-classical-transfer entropy, which quantifies the information gained by continuous measurement, we show that the precision of currents is constrained by information-thermodynamic costs such as the entropy production and information gain. Our result shows that information gain has the potential to enhance the precision of currents beyond the bounds set by the conventional TUR. We illustrate the bound with a driven two-level system under continuous measurement and feedback, demonstrating that feedback achieves higher precision of currents while suppressing the entropy production.

Polymer thin films exhibit pronounced interfacial mobility gradients that modify chain relaxation, yet how these gradients govern chain-scale dynamics across depth remains incompletely understood. Using molecular dynamics simulations of freestanding glass-forming polymer films, we resolve how depth-dependent variations in segmental relaxation shape chain dynamics across a wide range of displacement scales. Near the free surface, accelerated segmental mobility suppresses Rouse-regime scaling exponents to values as low as gamma = 0.4, reflecting transient localization induced by interfacial mobility gradients rather than topological entanglement. In contrast, the film interior exhibits enhanced Rouse scaling exponents consistent with predictions of the Heterogeneous Rouse Model (HRM), indicating that bulk dynamic heterogeneity compresses the Rouse regime. Mapping the minimum scaling exponent gamma_min across depth reveals a linear gradient that separates the bulk-like enhancement regime from the surface-induced suppression regime of chain dynamical scaling. Together, these results demonstrate that bulk and interfacial dynamic heterogeneity modify chain relaxation in opposite ways and establish Rouse scaling as a sensitive, spatially resolved probe of glassy dynamical heterogeneity and interfacial dynamical gradients in polymers.

We study the impact of off-resonant tunneling and coherences on the electron pumping through quantum dots. Thereby, we focus on two electron-pump setups where lowest-order tunneling processes are suppressed and the pump is exclusively driven by modulations of the coupling energy. The first setup is driven by switching on and off the couplings between the quantum dot and the leads, while the second setup employs measurements of the dot occupation. We derive exact solutions for arbitrarily strong tunnel couplings in the absence of Coulomb interaction, identify parameter regimes with optimal pumping currents or optimal energy efficiency, and discuss similarities between both pumping mechanisms.

Training large neural networks exposes neural scaling laws for the generalization error, which points to a universal behavior across network architectures of learning in high dimensions. It was also shown that this effect persists in the limit of highly overparametrized networks as well as the Neural network Gaussian process limit. We here develop a principled understanding of the typical behavior of generalization in Neural Network Gaussian process regression dynamics. We derive a dynamical mean-field theory that captures the typical case learning dynamics: This allows us to unify multiple existing regimes of learning studied in the current literature, namely Bayesian inference on Gaussian processes, gradient flow with or without weight-decay, and stochastic Langevin training dynamics. Employing tools from statistical physics, the unified framework we derive in either of these cases yields an effective description of the high-dimensional microscopic behavior of networks dynamics in terms of lower dimensional order parameters. We show that collective training dynamics may be separated into the dynamics of N independent eigenmodes, those evolution equations are only coupled through collective response functions and a common statistics of an effective, independent noise. Our approach allows us to quantitatively explain the dynamics of the generalization error by linking spectral and dynamical properties of learning on data with power law spectra, including phenomena such as neural scaling laws and the effect of early stopping.

In this study, the influence of pure dislocation and hydrogen-dislocation interactions on anelastic response or internal friction relaxation peaks in bcc-iron was investigated. These relaxations are primarily governed by thermally activated kink nucleation and kink migration events. An atomistic multiscale framework, coupling molecular dynamics (MD) and kinetic Monte Carlo (KMC) simulations, was developed to investigate the underlying atomistic mechanisms behind dislocation-relaxation peaks. MD simulations revealed that the presence of hydrogen atoms near the dislocation core facilitates the kink nucleation process by reducing the nucleation barrier while enhancing the barrier for dislocation migration. The KMC model captured Snoek-Koster peaks arising from the Cottrell atmosphere formed by hydrogen atoms and clusters around the dislocation core, providing insights into the atomistic mechanisms controlling these relaxations. Furthermore, the proposed computational scheme elucidated a unique linear relationship between hydrogen content and the internal friction loss factor, offering a methodology for hydrogen detection and quantification.

We study the free energy and dynamics of a closed elastic filament (a one-dimensional curve in two dimensions) whose local internal state is specified by curvature, stretch, and a scalar density field representing, for example, the concentration of an absorbed species. The density variable has a tendency to phase-separate whereas the local spontaneous curvature is concentration-dependent. There is also a coupling between concentration and the stretching of the filament, although our main interest is in the nearly inextensible regime. We formulate and simulate the dynamics, comprising a coupled Willmore flow and Cahn-Hilliard gradient flow on the full differential geometry of a closed filament, addressing issues that previous work typically sidestepped by restricting to the Monge gauge. We use a numerical strategy for global free energy minimization, presenting the equilibrium shapes and density profiles across a wide range of model parameters. The phase diagram is dominated by a relatively small number of simple shapes that exhibit, as expected, strong coupling between local curvature and concentration. We also find regimes where curvature and/or stretching energies suppress phase separation altogether. For selected parameter values we present fully dynamical results, tracking the time evolution of the various contributions to the free energy. The dynamics often arrive at metastable minima rather than the equilibrium state - for example, at states with more than the minimum number of interfaces between coexisting phases. The metastability of these states is absent for phase separation on a rigid circular domain and thus a direct result of the coupling between geometry and density.

We investigate the steady uniform shear flow of a granular gas composed of inelastic and rough Maxwell particles. Exploiting the mean-field character of the model, we derive exact expressions for the collisional production rates of the second-degree moments and obtain a closed nonlinear solution for the stress and spin-spin tensors. The rotational-to-translational temperature ratio and the proportionality between the spin-spin and stress tensors are shown to be independent of the coefficient of normal restitution and determined solely by roughness and moment of inertia. The reduced normal stresses, shear stress, and shear rate are obtained explicitly in terms of two effective parameters generalizing the cooling and stress relaxation rates of the smooth model. From these results we derive exact expressions for the non-Newtonian shear viscosity, the first viscometric function, and the friction coefficient. The dependence of the rheological properties on the normal and tangential restitution coefficients is analyzed in detail, revealing strong non-Newtonian behavior and nonmonotonic effects of roughness. The results reduce, in the appropriate limits, to those of the inelastic Maxwell model for smooth particles and to the Pidduck gas in the elastic perfectly rough case.

We investigate thermalization in a tight-binding chain with an on-site defect subject to local dephasing noise implemented as random phase kicks. For a single linear defect of strength $ε$, we obtain an exact analytical description of the system spectrum and formulate the dephasing-induced dynamics in the eigenstate basis. We derive an approximate kinetic equation for mode populations that describes a continuous-time random walk in action space. The walk transition rates are defined by the overlap matrix encoding the spatial structure of eigenstates that can be computed exactly. Analyzing the spectral properties of the equation, we show that defect-induced localized modes act as bottlenecks that strongly slow down relaxation, with rates scaling as $ε^{-2}$ for strong defects. Using large-deviation theory, we characterize rare dynamical trajectories and identify distinct relaxation pathways associated with low- and high-activity regimes in action space. We provide numerical evidence that the large-deviation function exhibits a dynamical phase transition in the limit $ε\to \infty$. We then extend our analysis to the nonlinear case, considering a single nonlinear defect embedded in either a linear or a fully nonlinear discrete Schrödinger equation. Numerical simulations reveal a qualitatively faster approach to equilibrium driven by the amplitude-dependent weakening of the defect. Our results provide a unified framework for understanding thermalization, rare fluctuations, and relaxation pathways in stochastic tight-binding systems.

Morphogenesis involves complex shape changes of biological tissues. Yet, tissue shape changes depend on tissue rheology, which in turn arises from the interplay of large numbers of cells. Here, we link cell- and tissue-scale mechanics by constructing mean-field rheological relations for the vertex model. In contrast to past work in the field, we study a vertex model with an explicit viscous friction. We also include two different cellular mechanisms creating active, anisotropic stresses. Our mean-field model accounts for cell shape and the non-linear elastic and visco-plastic regimes. We validate our results by predicting the response to large-amplitude oscillatory shear. There are several vertex model variants, and comparing to results from the literature, we show that their rheology depends on a number of model details. Our approach should be sufficiently general to construct non-linear mean-field constitutive relations for any cell-based tissue model.

We theoretically study surface superconductivity in a nodal-line semimetal by combining a minimal tight-binding model with a layer-resolved Bogoliubov-de Gennes approach. In the normal state, the model realizes a bulk nodal loop and an associated drumhead surface band in a slab geometry with open boundaries in the $z$ direction: the central layers reproduce the bulk-like density of states, whereas the surface layer exhibits a sharp zero-energy peak originating from the drumhead states. On top of this band structure we introduce chiral $p$-wave and $d_{x^2-y^2}$-wave superconducting channels and determine the layer-dependent gap amplitudes self-consistently. The chiral $p$-wave order parameter is strongly enhanced at the outermost layers and decays within only a few layers towards the interior, while the $d$-wave order parameter is more than an order of magnitude smaller on all layers. The quasiparticle dispersion and surface local density of states in the chiral $p$-wave state show that the drumhead band is efficiently gapped out and that the zero-energy peak in the normal surface spectrum is split into two coherence peaks, directly reflecting the induced superconducting gap. These results demonstrate that superconductivity driven by drumhead surface states is naturally biased toward a surface-localized chiral $p$-wave pairing symmetry and may offer qualitative guidance for interpreting surface-sensitive experiments on Pd-doped CaAgP.

Electronic and thermoelectric transport in zigzag monolayer WSe$_2$ nanoribbons are studied under monochromatic irradiation. The electronic structure is described within a six-orbital tight-binding framework constructed from the relevant tungsten and selenium orbitals, with atomic spin-orbit coupling included explicitly. Periodic driving is incorporated via the Peierls substitution, and in the high-frequency limit the system is mapped onto an effective static Floquet Hamiltonian with polarization-dependent renormalized hoppings. Coherent transport is evaluated using wave-function matching within the Landauer-Büttiker formalism. The lattice thermal conductivity is obtained independently from density functional perturbation theory combined with an iterative solution of the phonon Boltzmann transport equation. Light-induced hopping renormalization reshapes the band dispersion and transmission spectrum near the Fermi level, modifying the Landauer transport integrals that determine electrical and thermal conductances and the Seebeck coefficient. Together with spin-orbit-driven band splitting and reduced lattice thermal conductivity from enhanced anharmonic scattering, this leads to a thermoelectric figure of merit $ZT$ exceeding unity over a broad temperature range.

Quantum interference phenomena in superconductors, such as Josephson interference and Little-Parks oscillations, serve as powerful probes of phase coherence, symmetry breaking and vortex dynamics. However, they are typically observed in well-defined mesoscopic structures, and their behavior in the two-dimensional limit remains largely unexplored. Here, we report periodic magnetoresistance oscillations, superconducting interference patterns, and interfering diode effect in unpatterned few-layer NbSe2. These phenomena emerge exclusively within the superconducting fluctuation regime of thin samples, consistent with the enhanced anomalous metallic behavior of atomically thin NbSe2. The non-monotonic temperature dependence of both the oscillation amplitude and the diode efficiency can be captured by a model in which thermally activated vortices traverse intrinsic supercurrent loops. Our results reveal that the observed interference phenomena originate from the lost of global phase coherence, providing a new route to accessing interference effects in unpatterned superconductors.

Employing a renormalization group analysis that allows for an unbiased treatment of competing physical ingredients, we systematically trace how the interplay between Cooper pairing and disorder scatterings governs the emergence or suppression of Cooper instability in the low-energy regime of fractional Dirac semimetals.In the clean limit, we find that the emergence of Cooper instability requires surpassing a finite interaction threshold $|λ_c|$, and depends sensitively on both the fractional exponent $α$ and the transfer momentum $\mathbf{Q}=(Q,φ)$. Specifically, bigger values of $α$ enhance the tendency toward BCS instability. For $α\in(0.001,0.61)$, the $(Q,φ)$ parameter space separates into two distinct regions: Zone-\uppercase\expandafter{\romannumeral1}, where Cooper instability is suppressed, and Zone-\uppercase\expandafter{\romannumeral2}, where it is allowed. In the presence of disorders, we demonstrate that they can either promote or suppress Cooper instability. Disorder of type $Δ_1$ or $Δ_2$ enhances superconductivity by reducing the critical interaction threshold $|λ_c|$ and expanding the superconducting phase space (Zone-\uppercase\expandafter{\romannumeral2}). In sharp contrast, either $Δ_0$ or $Δ_3$ suppresses Cooper pairing by increasing $|λ_c|$ and shrinking the available phase space (Zone-\uppercase\expandafter{\romannumeral1}). Although Cooper instability can be enhanced when promotive disorders ($Δ_1$, $Δ_2$) coexist with a single suppressive disorder ($Δ_0$ or $Δ_3$), the suppressive influence of $Δ_{0,3}$ generally dominates the promotive effects of $Δ_{1,2}$ in the presence of all sorts of disorders. These results would be helpful for further studies of fractional Dirac semimetals and alike materials.

Biological systems often operate at the boundary between order and chaos, transitioning from directed to irregular dynamics to achieve adaptability and robustness. Reproducing such transitions in artificial soft matter remains a central challenge. Here, we report a biomimetic regime of directron dynamics in achiral nematic liquid crystals, in which coherent, directed motion collectively evolves into chaos. Driven by multi-directron interactions, the system develops coexisting directron families with competing trajectories, displaying randomized motion, dynamic assembly formation and spontaneous fission of high energy to low energy daughter directrons - all of which mimics the phenotypic diversity observed in biological groups. Above a critical electric field, these interactions drive the system into a chaotic state that is distinct from the directed behaviours reported previously. We further introduce a minimal dipole-based model that qualitatively captures the underlying physics of this transition. Together, our results establish an artificial active system in which chaos emerges intrinsically from interactions, offering a versatile platform to study biological dynamics and opening new avenues for liquid-crystal-based soft-matter applications involving adaptive transport, cargo delivery, and energy transduction

We introduce an Ising spin-glass model with correlated disorder which continuously interpolates between the pure ferromagnetic Ising model and the Edwards-Anderson model with symmetric disorder. For this model, we prove that physical quantities on the Nishimori line (NL) can be expressed exactly in terms of those of the pure Ising model at an effective temperature on any lattice in any dimension. For example, the energy on the NL is equal to the energy of the pure Ising model at the effective temperature up to a constant and a trivial factor. More remarkably, the specific heat on the NL equals the energy, not the specific heat, of the pure Ising model at the effective temperature, again up to a constant and a trivial factor. Gauge-noninvariant quantities such as the magnetization and correlation functions are exactly equal to the corresponding quantities of the pure Ising model at the effective temperature. These exact relations imply that the leading critical behavior at that multicritical point for the disorder-correlated model is pure-Ising-like, in contrast to the conventional multicritical universality class of the standard Edwards-Anderson model. Our results motivate further investigations of the relatively unexplored topic of correlations in disorder in spin glasses and related problems.

Fluctuations are intrinsic to microscopic systems and impose fundamental limits on nonequilibrium precision, as captured by the thermodynamic uncertainty relation (TUR), which links current fluctuations to entropy production. While feedback control is expected to further suppress fluctuations, its role within the TUR framework has remained unclear, particularly in quantum systems where control is inherently information-driven. In this Letter, we consider open quantum systems weakly coupled to a thermal environment, in which quantum jumps are continuously monitored, and Markovian feedback is applied. Using quantum mutual information to quantify the information contribution induced by feedback, we derive a finite-time TUR for arbitrary time-integrated currents in terms of entropy production and mutual information. Our results uncover how feedback control suppresses fluctuations together with thermodynamic cost and establishes a fundamental precision bound imposed by information-based control. As an application, we analyze a quantum clock model and demonstrate that the clock precision can be enhanced by feedback control in the presence of a single thermal reservoir.

We study when a density-matching Kohn--Sham (KS) description can reproduce a many-body Berry phase in a correlated insulator, despite the fact that geometric phases are functionals of the wave function. Focusing on the one-dimensional SSH--Hubbard chain on a ring as a controlled interacting topological model, we introduce a $U(1)$ twist $θ$ (flux insertion). The many-body ground state along the full twist cycle is computed by the density-matrix renormalization group (DMRG), while the onsite interaction $U$ is tuned from the noninteracting to the strong-coupling regime. At half filling in the inversion-symmetric gapped regime, our DMRG calculations show that the density remains constant within numerical accuracy over the entire $(θ,U)$ range studied. Thus, the density has no dependence on either the flux $θ$ or the interaction strength $U$. Accordingly, the symmetry-preserving density constraint collapses the KS reference to an SSH-type quadratic representative with $U$-independent geometric diagnostics. Nevertheless, the many-body wave function exhibits a nontrivial geometric response: the quantum metric associated with the $θ$-parametrized ground-state manifold depends on $θ$ at intermediate $U$ and is strongly suppressed at large $U$, consistent with the charge fluctuation freezing. Intriguingly, the KS and many-body Berry phases coincide throughout the gapped regime as $U$ is tuned from weak to strong coupling. We show that this agreement is best understood as symmetry-enforced $\mathbb{Z}_2$ sector matching, rather than as evidence that the density encodes the many-body Berry connection.

Quantum geometry quantifies how the single-particle Bloch wavefunction changes in phase and amplitude across the Brillouin Zone. In multi-orbital systems where bands have strongly mixed orbital composition, quantum geometry plays a vital role in determining the ground state and low-energy properties of interacting electronic systems. In this work, we show that Mott metal-insulator transitions, as well as transitions between different magnetic orders within the Mott insulating phase, can be driven by the quantum geometry of the underlying Bloch band, thereby providing a mechanism complementary to conventional bandwidth-tuned Mott transitions. By studying the Kane-Mele-Hubbard model using exact diagonalization, we demonstrate that in in half-filled and topologically-trivial bands, quantum geometric properties of the Bloch states alone can act as a tuning knob for Mott metal-to-insulator and affect the competition between ferromagnetism and antiferromagnetism. We show that both transitions may be heuristically understood via non-local Coulomb scattering in a basis of exponentially localized Wannier functions. These results highlight the role of quantum geometry beyond topological settings as a governing principle for conventional Mott and magnetic physics in multi-orbital and moiré materials.

We investigate an antiferromagnetic quantum Heisenberg model on a square lattice with high-spin magnetic impurities to clarify how random and stealthy hyperuniform impurity configurations influence the bulk magnetic properties. Stealthy hyperuniform configurations are generated using generalized cost functions that interpolate between square-lattice-like and triangular-lattice-like arrangements. Using linear spin-wave theory for the mixed-spin model, we demonstrate that triangular-lattice-like arrangements yield a larger average staggered magnetization than both random and square-lattice-like cases. This enhancement originates from sublattice effects: while the square-lattice-like configuration enforces nearest-neighbor impurities to occupy opposite sublattices due to its bipartite structure, the triangular-lattice-like arrangement allows same-sublattice nearest-neighbor pairs, thereby strengthening cooperative magnetic enhancement.

We present a quantitative study of many-body effects including the three-particle level on second-harmonic generation in monolayer MoS$_{2}$. Our approach combines many-body perturbation theory with time-dependent current-density-functional theory within an \textit{ab initio} framework in the optical limit. Inclusion of two-particle excitonic effects \textit{via} a dynamical long-range linear exchange-correlation kernel reproduces the qualitative features of the second-harmonic response, but underestimates the experimentally reported magnitudes by nearly a factor of two. By incorporating three-particle (trionic) correlations through a static long-range quadratic exchange-correlation kernel, we achieve significantly improved quantitative agreement with experiment. These findings highlight the role of many-body interactions beyond the excitonic level in accurately describing second-order optical responses in two-dimensional semiconductors.

Atomic-scale disorder is conventionally regarded as detrimental to superconductivity; however, under specific conditions, it can enhance superconducting properties. Here, we investigate the role of substitutional disorder in Remeika-type quasiskutterudites $R_3M_4$Sn$_{13}$ and $R_5M_6$Sn$_{18}$ ($R=$ Y, La, Lu; $M=$ Co, Rh, Ru) by combining measurements of magnetic susceptibility, electrical resistivity, and heat capacity with microscopic modeling. We demonstrate that increasing disorder leads to the emergence of locally superconducting regions characterized by an enhanced critical temperature $T_c^{\ast}$, exceeding the bulk transition temperature $T_c$. Both $T_c^\ast$ and $T_c$ exhibit a nonmonotonic dependence on dopant concentration and show a strong correlation with entropy isotherms measured as a function of disorder. The pronounced entropy maxima coincide with the largest separation between $T_c^{\ast}$ and $T_c$, establishing disorder as a thermodynamically controlled parameter governing superconductivity in these materials. Measurements of the upper critical field reveal distinct $H_{c2}(T)$ branches associated with the bulk and locally superconducting phases, providing direct experimental evidence for a percolative superconducting state. To interpret these observations, we propose a microscopic model that captures the interplay between the impurity-induced enhancement of local pairing and the disorder-driven suppression of global superconducting coherence. The model reproduces the experimentally observed nonmonotonic evolution of $T_c^{\ast}$ with disorder and supports a percolation-based interpretation of the superconducting transition. Our results demonstrate that controlled atomic disorder can serve as an effective materials-design parameter for tuning superconductivity in complex correlated systems.

Entropy governs molecular self-assembly, phase transitions, and material stability, yet remains challenging to quantify and directly control in molecular systems. Here, we demonstrate that the computable information density (CID), a data compression-based information theoretic metric, provides an instantaneous general measure of configurational entropy in molecular dynamics simulations, reflecting both local and long-range structural organization. We validate the CID across systems of increasing complexity, beginning with single-component Lennard-Jones melting before examining binary phase separation, polymer condensation and dispersion, and assembly of amorphous carbon networks at multiple densities. Unlike conventional order parameters, CID requires no a priori knowledge of relevant structural features and captures entropic signatures across a variety of molecular systems and discretization resolutions. By establishing entropy as a directly accessible structural metric, this framework lays a foundation for future entropy-driven materials design and optimization strategies.

Macroscopic quantum electrodynamics (MQED) provides a unified framework to describe quantum electromagnetic fields in the presence of arbitrary macroscopic environments. Central to this theory is the field correlation, which governs both radiative (e.g., Lamb shifts and the Purcell effect) and mechanical phenomena, such as van der Waals and Casimir forces. In this tutorial, we provide an overview of MQED and its extension to active media, highlighting fluctuation-induced forces as manifestations of gain-modified field correlations.

We study the paradigmatic spin-1 XY chain under open boundary conditions, which hosts exact quantum many-body scars generated by an emergent Spectrum Generating Algebra (SGA). We show that the scar subspace possesses a symmetry-protected trivial (SPt) character that we attribute to a hidden $Z_{2}\times Z_{2}$ symmetry of another model, namely the commutant Hamiltonian, for which the scars are the ground states. We construct a Lieb-Schultz-Mattis (LSM) type twist operator, which, for scar states, takes the value $-1,$ and, for ergodic states, approaches zero in the thermodynamic limit. A complementary understanding of the stability of the scars under different perturbations is obtained by analyzing the Loschmidt echo and Quantum Fisher Information (QFI) of the scars. Finite-size scaling analysis of the QFI reveals that the scars are much more sensitive to perturbations as compared to the nearby thermal states. Based on the analysis of QFI and different LSM twist operators, we obtain a classification of different SGA-preserving and SGA-breaking perturbations.

We study the equilibrium properties of a model of magnetic diblock copolymer where each monomer is decorated with an Ising-like spin. Spins interact ferromagnetically within each block and antiferromagnetically across blocks, generating frustration between magnetic ordering and spatial organization. By employing a mean-field approach and Monte Carlo simulations for self-avoiding walks on the cubic lattice, we investigate the system's response to an external magnetic field. We discover a rich phase diagram that includes: a swollen phase with both filaments magnetically disordered and spatially extended; a mixed compact phase characterized by a single globule in which the two filaments are strongly intertwined; a segregated compact phase composed of two globular, magnetically ordered, and spatially separated blocks. Further, if the magnitude of the intra-block ferromagnetic interaction differs between the two blocks, we observe a hybrid segregated (``tadpole'') phase where one extended block coexists with a collapsed one. Mean-field predictions are in quantitative agreement with Monte Carlo results for the location of the phase boundaries. These findings provide a minimal statistical-mechanical framework for field-controlled self-assembly of tunable patterns by magnetically heterogeneous polymers. They may also serve as a simple platform to investigate the coupling between internal epigenetic-like states and chromatin folding.

We numerically investigate quantum circuit elementary-gate level instantiations of the standard Quantum Phase Estimation (QPE) algorithm for the task of computing the ground-state energy of a quantum magnet; the disordered fully-connected quantum Heisenberg spin glass model. We consider (classical simulations of) QPE circuit computations on relatively small quantum Hamiltonians ($3$ qubits) with up to $10$ phase bits of precision, using up to Trotter order $10$. We systematically study the inputs of QPE, specifically time evolution, Trotter order, Trotter steps, and initial state, and illustrate how these inputs practically determine how QPE operates. From this we outline a coherent set of quantum algorithm input and tuning guidelines. One of the notable properties we characterize is that QPE sampling of the optimal digitized phase converges to a fixed rate. This results in strong diminishing returns of optimal phase sampling rates which can occur when the Trotter error is surprisingly high.

Metric networks are network-shaped, one-dimensional structures on which one can solve differential equations to simulate a wide range of physical systems including conjugated molecules, photonic crystals, quantum mechanics in waveguide networks, and acoustic metamaterials. More concretely, a metric network is a network whose edges are each assigned a notion of length and a coordinate describing position. One can then define function spaces and differential operators on these objects to model the aforementioned systems. Recent software advancements have made it feasible to analyze partial differential equations on large, compact metric networks with a vast array of structures. Here, we generate compact metric network structures using the spatial tessellations of two-dimensional hyperuniform point patterns, which have suppressed large-scale density fluctuations relative to typical disordered point patterns. This choice of structure is inspired by the exotic physical properties of network materials with these structures in other contexts. Then, we characterize the eigenvalue spectrum structure of the differential Laplace operator on these networks. In particular, we find that gaps can form in the eigenvalue spectra of these networks whose widths increase when the distribution of edge lengths is narrow and as the number of triangular faces increases. Importantly, many of the structures we consider are realizable in Euclidean space, meaning they are well-suited for practical applications in, e.g., metamaterial design. This work can thus be used to inform the design of metric network-based systems with spectral gaps with tunable widths and locations.

We compute non-perturbative, resurgent contributions to the Virasoro minimal string and 3d gravity using techniques from hermitian matrix models. In particular, we construct a fully non-perturbative partition function for the Virasoro minimal string in terms of a Zak transform. In this context, negative tension D-branes appear naturally, which in the matrix model correspond to anti-eigenvalues, or instantons on the involuted sheet of the spectral curve. We further extend this analysis to resolvents and observe resurgent wall crossing phenomena between ZZ- and FZZT-branes. Using recent results that relate the Virasoro minimal string to 3d gravity with end-of-the-world branes we proceed to study the resurgent consequences of summing over the genus in 3d gravity, where we find non-perturbative contributions of doubly exponential type. These statements are then tested using resurgent large-order asymptotics. Lastly, we compute the non-perturbative eigenvalue density for generic hermitian matrix models and identify the change of asymptotic behavior at the edge of the eigenvalue distribution with a Stokes transition. This allows us to identify oscillations in the eigenvalue density with anti-Stokes behavior of FZZT-branes. In the case of 3d gravity with end-of-the-world branes we comment how this Stokes transition coincides with the onset of black hole behaviour and compute the non-perturbative primary density. Furthermore, we apply these techniques to the eigenvalue density of JT gravity to compute higher genus corrections.

The notion of operator growth in quantum systems furnishes a bridge between transport and the generation of entanglement between different parts of the system under quantum dynamics. We define a measure of operator growth in terms of subsystem operator Rényi entropy, which provides a state-independent measure of operator growth, unlike entanglement entropies, and the usual measures of operator growth like out-of-time-order correlators. We show that the subsystem operator Rényi entropy encodes both spatial and temporal information, and thus can directly connect to transport for a local operator related to a conserved quantity. We construct a unified Schwinger-Keldysh (SK) field theory formalism for the time evolution of operator Rényi entropy and entanglement entropies of initial pure states. We use the SK field theory to obtain the operator Rényi and state entanglement entropies in terms of infinite-temperature and vacuum Keldysh Green's functions, respectively, for non-interacting systems. We apply the method to explore the connection between operator and entanglement growth, and transport in non-interacting systems with quasiperiodic and random disorder, like the one- and two-dimensional Aubry-André models and the two-dimensional Anderson model. In particular, we show that the growth of subsystem operator Rényi entropy and state von Neumann and Rényi entanglement entropies can capture both ballistic and sub-ballistic transport behaviors, like diffusive and anomalous diffusive transport, as well as localization in these systems.

We demonstrate the presence of quantum interference effects in van der Waals heterostructures subject to in-plane electric fields. The in-plane field $F$ accelerates carriers through a hybridized band edge, and interlayer Zener tunneling occurs by distinct pathways, resulting in a solid-state quantum interferometer with imprints in transport observables. For parabolic-band bilayers, we identify two characteristic signatures which are observable in lateral conductance: Landau-Zener-Stuckelberg oscillations in the band-overlap regime periodic in $1/F$ at small fields, resembling electric-field induced quantum oscillations, and a pronounced resonance at $F\propto T_0^{3/2}$ set by the interlayer tunneling $T_0$. These features provide a directly accessible probe of coherent interferometric dynamics in van der Waals heterostructures, and could be harnessed for more precise engineering and characterization.

We establish a symmetry-protected correspondence between band topology of coherent Hamiltonians and Liouvillian spectral winding in Lindblad descriptions of open quantum systems. This allows the Hamiltonian topology to act as a knob for controlling Liouvillian topology and corresponding non-equilibrium dynamics, rather than being passively manipulated by system-environment exchanges. In particular, by exactly solving the Liouvillian spectrum in a class of one-dimensional dissipative lattices, we find that the Hamiltonian band topology constrains the Liouvillian spectral winding and determines the Liouvillian skin effect, provided the Hamiltonian and quantum jump operators respect the same chiral symmetry. We further demonstrate that lattice parity controls the associated bulk-boundary correspondence and the coherence properties of the steady state. Our results unveil a symmetry-enforced topological control of spectral and spatial organization in open quantum systems, providing a unified perspective on topology in Hamiltonian and dissipative dynamics.

Two anyons encircling each other acquire a quantized braiding phase that is independent of their spatial separation. We show that detecting this phase in a fractional quantum Hall interference experiment results in a quantum Zeno effect: a localized anyon is trapped by constant observation from a stream of anyons supplied by the measurement current. Interferometers with an embedded antidot are ideal for accessing the Zeno regime, where the bare tunneling rate of localized anyons is much lower than the measurement rate. The Zeno-suppressed tunneling rate of the trapped anyon depends on the braiding phase and the transmission of the quantum point contacts. Our primary prediction is that the autocorrelation time of the conductance through the interferometer increases with the bias current. This effect may be used to experimentally control the anyon dynamics, in particular to increase the lifetime of localized anyons.

We investigate superconductivity in KTaO$_3$ (KTO)-based two-dimensional electron gases for the (001), (110), and (111) crystallographic orientations within a unified microscopic framework. Using a self-consistent tight-binding slab model, we determine the confinement potential, electronic structure, and orbital composition for each orientation, explicitly including inversion-symmetry-induced orbital Rashba couplings. Using a local spin-singlet s-wave pairing interaction, we find that the pronounced orientation dependence of the superconducting critical temperature primarily originates from differences in the spatial extent of the two-dimensional electron gas and the associated redistribution of the density of states at the Fermi level, rather than from changes in the pairing interaction. Our results provide a microscopic explanation for the experimentally observed orientation dependence of superconductivity at KTO-based interfaces.

Adiabatic (or reversible) processes are the key concept unifying our understanding of thermodynamics and dynamical systems. Reversibility in the thermodynamic sense is understood as entropy-preserving processes, such as in the idealized Carnot engine, whereas in integrable dynamical systems it is understood as the conservation of the action variables. Between these two idealized limits, however, where the phase space can become mixed, things are much less clear. In this work, we first determine the extent to which reversible processes are even possible in this regime. We then explore how the dissipative losses resulting from rapidly driving these kinds of systems can be fought by approximate counterdiabatic driving. Finally, we argue that much of the phenomenology should be the same for quantum many-body systems with large degeneracy in the presence of integrability breaking perturbations.

We investigate thermal avalanche dynamics in amorphous solids using elastoplastic models with local activation rules and no external driving. Dynamical heterogeneities, quantified through persistence measurements and the associated four-point susceptibility $χ_4$, reveal the emergence of correlated spatiotemporal rearrangements as temperature is varied. As temperature increases, avalanche statistics evolve from scale-free behavior with exponential cutoffs to regimes dominated by system-spanning runaway events. We identify a system-size-dependent critical temperature $T_c(L)$ that separates intermittent avalanche dynamics from thermally assisted flow, where self-sustained avalanches transiently fluidize the system. We show that $T_c(L)$ decreases algebraically with increasing system size, suggesting that in the thermodynamic limit arbitrarily small but finite temperatures may destabilize the intermittent regime. The relation between avalanche size and duration resembles that in sheared systems, whereas the statistics of minimal distances to yielding reveal a temperature-driven reorganization of marginal stability absent in strictly driven overdamped dynamics. Our results demonstrate that thermal activation alone can generate a finite-size-controlled instability scale in disordered elastic media.

Individual spin defects in solids are promising building blocks for quantum technologies, but their deterministic creation, individual addressability, and operation near surfaces remain major challenges. Two-dimensional materials provide an attractive alternative, as their single-layer thickness enables direct atomic-scale access to defect states. Here, we demonstrate single-spin control of solid-state defects in a two-dimensional semiconductor by a combination of scanning tunneling microscopy and electron spin resonance. We create and manipulate individual sulfur vacancies and carbon substitution defects in monolayer molybdenum disulfide and characterize their spin dynamics, including coherent control, at the single-defect level. Using atomic manipulation, we further engineer and probe spin-spin interactions between defect pairs. Our results demonstrate deterministic creation, addressability, coherent manipulation, and controlled coupling of individual spin defects within a single experimental platform. This establishes atomically engineered spin defects in two-dimensional semiconductors as a versatile class of controllable solid-state quantum systems and opens a route towards tailored quantum sensing experiments.

Programmable microheater arrays are essential for a variety of applications including gas sensing, microfluidic lab on a chip devices, 3D printers, and biosensors that rely on DNA amplification. Increasing the density and number of heating elements directly correlates with the precision with which spatiotemporal heat profiles can be delivered. However, large arrays have thus far not been realized. One challenge is that as the number of elements in an array increases, the complexity of connecting them grows. Here, we show that row-column addressing provides a promising architecture for the efficient operation of a large micro-heater array. We introduce a programmable 32 x 32 microheater array consisting of individually addressable robust platinum (Pt)-based Joule heating elements- each smaller than 300 micrometer. We show that combining high-voltage multiplexed electronics and sequential addressing controlled by a high frequency clock, allows the independent operation of the 1024 microheater elements. We demonstrate the generation of heat images and the patterning of metallic structures formed from the liquid metal Gallium. Our work demonstrates new capabilities for on-chip thermal devices, and opens the possibility to realize novel heat-controlled microactuation systems.

The resistivity maximum at a temperature $T=T_{\mathrm{max}}$ is a recurring feature of bandwidth-tuned Mott systems, yet its meaning remains controversial: is it a coherence-incoherence crossover of an electronically homogeneous metal, or does it mark the onset of transport through a mixed landscape of metallic and insulating regions? Even more debated is whether a true phase-coexistence regime survives in the relevant parameter range, or whether apparent inhomogeneity is merely extrinsic. Here we address these questions by moving beyond temperature sweeps and probe charge transport in the time domain. Near $T=T_{\mathrm{max}}$, we find that the resistance of a model system, a quasi-two-dimensional Mott spin liquid material, exhibits clear random-telegraph switching between discrete levels over long timescales. The statistics of the switching - sharp two-level behavior with thermally activated dwell times - point to a mesoscopic "current-controlling" region that dynamically toggles between metallic and insulating states, intermittently opening and closing the dominant conduction channel. This characteristic fluctuating dynamics provides direct evidence for intrinsic metal-insulator coexistence and establishes $T\sim T_{\mathrm{max}}$ as the regime of Mott intermittency, where transport is governed by stochastic domain switching rather than quasiparticle decoherence.

Temperature has a crucial influence on rubber friction and tire dynamics. The temperature field in a rubber tread block is the sum of the background temperature $T_0({\bf x},t)$, which varies slowly in time and space, and the flash temperature $ΔT({\bf x},t)$, which in nonzero only close to the macroasperity contact regions, and which varies rapidly in time often on the millisecond time scale. Here we study the motion of a single tire tread block and how it is influenced by the flash temperature. We also present a theory and experimental results for the size of the macroasperity contact regions. In particular, we show that for a large enough nominal contact area, in most cases the diameter $D$ of the macroasperity contact regions are nearly independent of the elastic modulus and the nominal contact pressure.

Ultra-stable glasses prepared from the physical vapor deposition of organic molecules present a very low density of two-level states, the kind of glass defects that determine their peculiar low temperature thermal properties. Numerical simulations suggest that quasi-localized harmonic vibrational modes emerge in the soft regions associated with two-level states. However, the connection between the low frequency vibrational modes and the local structural instabilities of glasses remains unexplained. Here we exploit a recently developed spectrograph for nuclear resonant analysis of inelastic X-ray scattering to probe the density of vibrational states of amorphous thin films of ultra-stable and conventional glasses down to an exceptionally low frequency of $\sim 70$ GHz. We show that the glass stability does not affect the harmonic vibrational modes at the lowest frequencies, despite a reduction of almost an order of magnitude in the density of two-level states. At the same time, the vibrational modes at higher frequencies, around the boson peak maximum, are extremely sensitive to the glass stability. Although we cannot exclude the possible existence of quasi-localized modes in glasses, we show that their presence is not strictly necessary to describe the measured density of low frequency vibrations. The experimental developments here presented pave the way to the solution to the long-standing debate on the low frequency vibrations in glasses.

In a recent experiment [X. Yu et al., arXiv:2602.07654], energy-resolved measurements of an atomic matter wave spreading in a speckle potential enabled the direct observation of the three-dimensional Anderson transition. In this work, we present a quantitative theoretical description of the matter-wave dynamics based on a tailored implementation of the self-consistent theory of localization, which incorporates both the spectral and spatial properties of the state prepared in the disorder. We benchmark this theoretical approach against ab initio numerical simulations, and use it to analyze the atom density profiles observed experimentally in the localized, diffusive, and critical regimes. Particular emphasis is placed on the key role of the atomic energy distribution, especially on the distinct contributions of Bose-condensed and thermal atoms to interpret the experimental profiles. Our framework provides a versatile and efficient theoretical toolbox for quantitatively describing wave-packet dynamics in three-dimensional disordered quantum systems, which remain challenging for state-of-the-art large-scale numerical simulations.

Ostwald ripening in O/W emulsions in presence of solubilizing micelles is theoretically studied. At small average sizes, the kinetics is predicted to follow the classical Lifshits-Slezov-Wagner cubic law, with the rate proportional to the molecular solubility of the oil in water, as if no micelles were present. At larger particle sizes the kinetics transitions to the Wagner's quadratic law. The crossover point for the kinetics depends on the dynamics of the oil solubilizate-micelle exchange; it is set by the value of the oil atmosphere distribution parameter, kappa, which, somewhat like Debye length, is proportional to the square root of the micellar concentration. It should be noted that in the range when 1/kappa is close to the particle average radius, the ripening kinetics still nearly follows the cubic law, with only moderate deviations; in this case, the micellar effects are experimentally seen not as the deviations from the linearity, but as an apparent increase in the cubic rate. The increase is predicted to be larger in case of nonionic micelles of ethylene oxide type compared to ionic ones.

Dense suspensions of self-propelled bacteria and related active fluids exhibit spontaneous flow generation, vortex formation, and spatiotemporally chaotic dynamics despite operating at vanishingly small Reynolds numbers. These phenomena, commonly referred to as active turbulence, display striking visual and statistical similarities to classical inertial turbulence while arising from fundamentally different nonequilibrium mechanisms. In this article, we present a combined review and theoretical study of hydrodynamic models for dense active fluids, with particular emphasis on bacterial suspensions described by the Toner--Tu--Swift--Hohenberg (TTSH) framework. We review key experimental and theoretical developments underlying the analogy between active and inertial turbulence, highlighting the emergence of multiple dynamical regimes and the conditions under which universal spectral and intermittent behavior arises in homogeneous systems. Moving beyond the conventional assumption of spatially uniform activity, we introduce a minimal model in which the activity field is heterogeneous and dynamically advected by the flow it generates. Thus treating activity as a spatiotemporally evolving field coupled to the TTSH dynamics, we investigate how advection and diffusion lead to sharp activity fronts, confinement of turbulent motion, and complex interfacial morphologies. Our numerical results demonstrate that spatial variations in activity can induce transient coexistence of distinct spectral regimes and that universality in active turbulence is inherently local and time-dependent in heterogeneous systems. These findings underscore the importance of treating activity as a dynamical field in its own right and provide a framework for studying active turbulence in more realistic, spatially structured biological and synthetic active matter systems.

Searching by chemical and structural analogy is one of the most commonly used and successful approaches to materials discovery. However, formulating this task for algorithmic implementation raises the question of how we define similar materials. Methods have been proposed for searching materials space using vectors based on chemical composition and functional fragments in the material. Descriptors for structural similarity have also been proposed. However, the question of how to incorporate and balance structural and compositional similarity measures in a single metric remains open. Here, we adapt methods developed for calculating distances between undirected graphs and apply them to crystalline materials similarity. The Fused Gromov-Wasserstein (FGW) metric uses optimal transport theory to map between two graphs considering a balance of the graph structure and the information present at the nodes of the graph (atoms in crystals). We apply the method to exploring new photovoltaic materials. We demonstrate that FGW is competitive with embeddings from an equivariant graph neural network, trained on $> 10^6$ materials, despite minimal training. We then apply FGW to a discovery campaign to identify materials from the Materials Project database that have not previously been explored as photovoltaics, but have similarities to known high-efficiency materials. After validating predictions with hybrid density functional theory, we identify seven previously unexplored high-efficiency photovoltaic absorber candidates, including Cs$_5$Sb$_8$, which is found to have a predicted SLME of $> 30\%$ and to be thermodynamically stable. The FGW approach demonstrates the power of strong inductive biases for developing metrics for materials exploration with minimal training data.

The spin nematic phase, characterized by long-range order of spin quadrupole moments in the absence of dipolar magnetism, presents a significant challenge for conventional experimental detection. We propose a novel method to detect this elusive order in quantum magnets with an illustration in the spin-1 triangular lattice Mott insulator. By integrating out the phonon degrees of freedom, we obtain a phase diagram with substantially enlarged regions for the spin-nematic and spin-nematic-supersolid phases. We then demonstrate that through the spin-lattice coupling, the emergence of spin nematic order imprints a distinctive signature onto the phonon spectra, providing a clear spectroscopic signature for the quadrupolar order accessible via Raman or inelastic X-ray scattering. Our formalism offers a direct and powerful method to uncover the hidden spin nematicity, opening a new pathway for diagnosing multipolar orders in quantum magnets.

The results for the electrical double layer capacitance and the charge density of ``free ions'' obtained from the mesoscopic theory are compared with the corresponding results of the associative mean spherical approximation. While the first theory takes into account the fluctuations of the charge density, the second theory assumes that the free ions and ion pairs are in chemical equilibrium according to the mass action law. Our results demonstrate a fairly good agreement between the two theories at high densities and low temperatures.

We present peapods (github.com/PeaBrane/peapods), an open-source Python package for Monte Carlo simulation of Ising spin systems with arbitrary coupling constants on periodic Bravais lattices with user-specified neighbor offsets. The computational core is written in Rust and exposed to Python via PyO3, combining the ergonomic interface of Python with the performance of compiled, memory-safe code. The package implements Metropolis and Gibbs single-spin-flip algorithms, Swendsen-Wang and Wolff cluster updates, parallel tempering, and three replica cluster moves for spin glasses: the Houdayer isoenergetic cluster move, the Jorg stochastic variant, and the Chayes-Machta-Redner (CMR) blue-bond algorithm. Overlap statistics between replica pairs enable computation of the spin glass order parameter and Binder ratio. Replica-level parallelism is achieved through the Rayon work-stealing scheduler. We validate the implementation against the exact critical temperatures of the two-dimensional Ising model on the square and triangular lattices via finite-size scaling of the Binder cumulant.

Generative artificial intelligence has shown remarkable success in synthesizing data that mimic complex real-world systems, but its potential role in the discovery of mathematically meaningful structures in physical models remains underexplored. In this work, we demonstrate how a generative diffusion model can be used to uncover previously unknown solutions of a nonlinear partial differential equation: the two-dimensional Navier-Stokes equations in a turbulent regime. Trained on data from a direct numerical simulation of turbulence, the model learns to generate time series that resemble physically plausible trajectories. By carefully modifying the temporal structure of the model and enforcing the symmetries of the governing equations, we produce synthetic trajectories that are periodic in time, despite the fact that the training data did not contain periodic trajectories. These synthetic trajectories are then refined into true solutions using an iterative solver, yielding 111 new periodic orbits (POs) with very short periods. Our results reveal a previously unobserved richness in the PO structure of this system and suggest a broader role for generative AI: not as replacements for simulation and existing solvers, but as a complementary tool for navigating the complex solution spaces of nonlinear dynamical systems.

In vibro-impact mechanics, the division between an impact and a near miss is a zero-velocity grazing event. Grazing bifurcations of stable periodic motions often produce complicated attractors when grazing generates a square-root term in the Poincaré map. This paper concerns codimension-two scenarios for which the square-root term vanishes in some iterate of the Poincaré map. For forced one-degree-of-freedom oscillators, this occurs when the forcing frequency is a certain rational multiple of the damped natural frequency, i.e., the system is in resonance. In two-parameter bifurcation diagrams, curves of saddle-node and period-doubling bifurcations of single-impact periodic motions emanate from the codimension-two points. In this paper we prove these curves are quadratically tangent to the curve of grazing bifurcations, and derive explicit formulas for their quadratic coefficients. This is achieved by modifying the Poincaré map in a way that circumvents the square-root singularity, enabling us to use the implicit function theorem to demonstrate smoothness and perform asymptotic calculations of the saddle-node and period-doubling bifurcation curves. In doing so we resolve a long-standing conjecture on the admissibility of single-impact periodic motions by supplementing raw asymptotic computations with geometric and topological arguments. We illustrate the results with a linear impact oscillator model, matching the theoretical unfolding to numerically computed bifurcation curves. The results explain why previously reported physical experiments reveal an absence of chaos shortly past the grazing bifurcation.

Phase control of parametric modulation in coupled oscillator networks enables sculpting of dynamical states with desired spatiotemporal symmetries. Symmetry-aware Floquet analysis successfully predicts such states in linear systems, but whether their symmetry properties persist under nonlinearity remains largely unexplored. Here, we establish the existence of nonlinear chiral steady states in a trio of coupled parametric oscillators with modulation phases chosen to selectively amplify a circulating mode in the linearized system. We find that a cubic nonlinearity arrests exponential growth of the amplified mode, producing a steady finite-amplitude motion that retains the expected chirality. By exploiting space-time symmetry, we reduce the dynamics to a single averaged equation that quantitatively predicts nonlinear trajectories, steady-state amplitudes, and characteristic time scales. The chiral steady states possess finite basins of attraction and are accessible from wide ranges of initial conditions and system parameters. Finite-element simulations of elastic plate resonators quantitatively reproduce these features, establishing the relevance of the reduced model to realistic continuum systems. Our results demonstrate that desirable properties of linear time-modulated systems, such as chirality and directional amplification, persist into strongly nonlinear regimes, opening pathways to robust nonreciprocal signal routing and amplification in parametrically driven platforms.

Vortex filaments are highly rotating localized structures of fluids that admits several types of excitation. Here, we study them by using numerical simulations of the three-dimensional incompressible Navier-Stokes equations. We first address the propagation of Kelvin waves, helicoidal excitations propagating along the filament, and measure their dispersion relation which turns out to be in good agreement with the original Lord Kelvin predictions. Then, inspired by the connection between vortex line dynamics and an integrable system, we show numerically the existence of solitons propagating along vortex filaments and study the collision of two of such structures. Finally, we show numerically the experimental feasibility of studying vortex solitons in the lab, by proposing an experiment for their generation.

Herbivorous wild species constantly strive to optimize the trade-off between energy and nutrient intake and predation risk during foraging. This has led to the selection of several evolutionary traits -- such as diet, habitat selection, and behavior -- which are simultaneously shaped by the spatio-temporal variability of the habitat. Among camelid species, polygyny is a prevalent behavioral strategy that encompasses both mating and foraging activities. This group-level behavior has multiple interacting dimensions, contributing to an interesting ecological and evolutionary complexity. We developed an individual-based stochastic model in which camelid females transition between different familial groups in response to their environmental conditions, aiming to maximize individual fitness. Our results indicate that the behavioral strategy of individual females can shape, by itself, emergent population-level properties, including group size and fitness distribution. Furthermore, these properties are modulated, in a non-additive manner, by other factors such as population density, sex ratio and system heterogeneity.

The recent, rapid advances in nonlinear chipscale nanophotonics in the visible and near-infrared have been largely driven by manipulating the local dielectric environment proximate to decades-old workhorse bulk nonlinear optical materials, rather than increasing the inherent strength of their nonlinear response. While proposed decades ago, we demonstrate the first experimental realization of a new class of designer nonlinear materials that leverage the interband optical transition in asymmetric structures to provide strong second order susceptibility, $χ^{(2)}$. Using simple AlGaAs/GaAs coupled quantum wells operating in the near-infrared as a prototype, we observed strong second harmonic generation enhancement of 1550 nm to 775 nm over bulk controls. Extracted $χ^{(2)}$ values were as high as 2750 pm/V, which is $>$7x that of bulk GaAs. Furthermore, measured susceptibilities agreed well with quantum mechanical calculations of $χ^{(2)}$ using layer profiles extracted from electron microscopy. Growth interruptions were employed to improve interfacial abruptness in response to electron microscopy characterization, resulting in increased $χ^{(2)}$ toward the simulation predictions for ideal heterointerfaces. More complex layer designs showed predicted $χ^{(2)}$ up to 7 nm/V. Such materials are anticipated to find myriad applications, including entangled photon generation at telecommunications wavelengths for chipscale quantum information processing.

Atom arrays have emerged as a powerful platform for quantum light-matter interfaces, yet single-species arrays are constrained by in-plane symmetry, restricting polarization control. Here we investigate the cooperative optical response of dual-species subwavelength atom arrays, in which intrinsic polarizability difference breaks in-plane symmetry. By engineering the lattice spacing and detunings, the arrays exhibit polarization-dependent subradiant modes, enabling complete reflection of specific polarization component. Leveraging this mechanism, we assemble array units as functional pixels and demonstrate a scalable polarization-selective quantum light modulator. Our work establishes a dynamically reconfigurable atomic-photonic platform for versatile subwavelength quantum optical elements.

Atomistic simulations are widely used to investigate reactive processes but are often limited by the rare event problem due to kinetic bottlenecks. We recently introduced an enhanced sampling approach based on the committor function, machine-learned following a variational principle. This method combines a transition-state-oriented bias potential, expressed as a functional of the committor, with a metadynamics-like bias along a committor-based collective variable, enabling uniform exploration of reaction pathways. In its original formulation, the committor is represented by a neural network that takes physical descriptors as input and is trained by minimizing a functional involving gradients with respect to atomic coordinates, which can be computationally demanding in some cases. Here, we propose a simplified learning criterion formulated entirely in the descriptor space, which bypasses the need for explicit and costly coordinate gradients and provides a relaxed upper bound to the original variational principle. Although this approach does not formally target the exact committor, we show that it retains robust sampling performance while significantly reducing computational costs, thus enabling the study of processes that would be practically unfeasible using the original formulation.

Quantum theory departs from classical physics in its treatment of correlations, most prominently through the phenomena of contextuality and nonlocality. Once regarded primarily as foundational curiosities, these effects are now understood as key operational resources for quantum computation, communication, and simulation. Although traditionally investigated in distinct settings, recent theoretical and experimental advances have revealed deep conceptual, mathematical, and operational connections between them. This review presents a unified perspective on these developments based on sheaf-theoretic and graph-theoretic frameworks, which provide theory-independent characterizations of statistical correlations. These approaches clarify the structural relationship between contextuality and nonlocality, facilitate the formulation of experimentally testable inequalities, and guide implementations in realistic physical platforms, with particular emphasis on photonic systems. By bridging abstract theoretical structures and concrete experimental realizations, this review sheds light on the nonclassical foundations of quantum correlations and their emerging role in quantum technologies.

Ultra-low-noise stabilized lasers are a fundamental tool for precision quantum technologies, optical clocks, microwave and millimeter-wave generation, and fiber sensing. Existing systems rely on table-top bulk-optic components -- discrete lasers, reference cavities, isolators, modulators and frequency shifters -- limiting portability, scalability, and manufacturability. While these systems offer flexibility in laser design to tailor linewidth, frequency noise, and wavelength to specific applications, fully integrating a stabilized laser onto a chip without sacrificing performance and versatility has remained elusive. Here, we report integration of the precision stabilized laser in the low-loss silicon nitride photonic platform, combining a flexible isolator-free core laser design with a modulation-free stabilization cavity. We demonstrate a stabilized widely tunable self-isolating extended cavity tunable laser monolithically integrated with an on-chip coil-loaded Mach-Zehnder interferometer (CL-MZI). This design yields a fundamental linewidth of 1.7 - 10.5 Hz across a 60 nm tuning range, integral linewidth of 299 - 505 Hz over a 30 nm tuning range, frequency noise reduction of over 5 orders of magnitude, and an Allan-Deviation (ADEV) of 6.5x10-13 at 0.08 ms. We next highlight the versatility of this approach by demonstrating a self-isolating stimulated Brillouin scattering (SBS) laser, that provides nonlinear noise suppression of high frequency noise by multiple orders of magnitude, stabilized to an on-chip CL-MZI. The stabilized SBS laser achieves 4 Hz fundamental linewidth, 74 Hz integral linewidth, and ADEV of 2.8x10-13 at 5 ms. These results bring the performance and versatility of table-top stabilized laser systems to a chip for the first time, providing a path to scalable, low-cost, and manufacturable precision lasers for portable quantum, sensing, and communications applications.

Hybrid metrology for semiconductor manufacturing is on a collision course with dark uncertainty. An IEEE technology roadmap for this venture has targeted a linewidth uncertainty of +/- 0.17 nm at 95 % coverage and advised the hybridization of results from different measurement methods to hit this target. Related studies have applied statistical models that require consistent results to compel a lower uncertainty, whereas inconsistent results are prevalent. We illuminate this lurking issue, studying how standard methods of uncertainty evaluation fail to account for the causes and effects of dark uncertainty. We revisit a comparison of imaging and scattering methods to measure linewidths of approximately 13 nm, applying contrasting statistical models to highlight the potential effect of dark uncertainty on hybrid metrology. A random effects model allows the combination of inconsistent results, accounting for dark uncertainty and estimating a total uncertainty of +/- 0.8 nm at 95 % coverage. In contrast, a common mean model requires consistent results for combination, ignoring dark uncertainty and underestimating the total uncertainty by as much as a factor of five. To avoid such titanic overconfidence, which can sink a venture, we outline good practices to reduce dark uncertainty and guide the combination of indeterminately consistent results.

We experimentally realize an analogue of the optical Maxwell fish-eye lens (MFEL) using phononic excitations in a Bose-Einstein condensate (BEC). A MFEL is characterized by a radially symmetric, spatially varying refractive index with the remarkable property that rays emitted from any point within the lens are perfectly focused at their image points. While the implementation of such gradient-index lenses is challenging in conventional optical systems, BECs offer a highly tunable platform in which the spatially varying speed of sound of collective excitations -- phonons, the acoustic-wave analogues of photons -- can be engineered and their dynamics observed in real time. Time-resolved measurements of phonon wavefronts reveal focusing behavior that shows good agreement with analytical theory and numerical simulations. This work provides both a geometric and physical framework for engineering effective refractive indices using ultracold atoms, and simulating wave propagation on effective spherical geometries.

We employ the method of images to derive analytic expressions for the magnetic field of Halbach multipoles that are enclosed in high-permeability shielding.

We present HELIOS (HomogEneous and Layered medIa Optical Scattering), an open-source surface integral equation (SIE) software designed for modeling light scattering by particles embedded in homogeneous or layered media and periodic backgrounds. The code implements the Poggio-Miller-Chang-Harrington-Wu-Tsai (PMCHWT) formulation that has demonstrated exceptional reliability in solving scattering problems with penetrable objects. Domain boundaries are discretized using triangular meshes, upon which the electric and magnetic surface current densities are expanded using the Rao-Wilton-Glisson (RWG) basis functions. For periodic structures, such as photonic crystals and metasurfaces, HELIOS employs Ewald's transformation to efficiently evaluate the infinite series associated with 2D lattices. Regarding stratified media, the code utilizes a matrix-friendly approach for the layered media Green's tensor, computing Sommerfeld integrals and accelerating calculations through a tabulation-interpolation scheme. The source code is implemented in C++, while a Python interface manages the workflow, including simulation setup, solver run, and post-processing. The accuracy and versatility of HELIOS are demonstrated through various examples that cover all its functionalities.

Single-crystal X-ray diffraction and nonlinear optical measurements, especially second- and third-harmonic generation (SHG/THG) are comprehensively investigated for the van der Waals layered material AgScP2S6 with a non-centrosymmetric P31c (159) space group. Linear optical constants are extracted using spectroscopic ellipsometry and applied in fitting the harmonic generation behavior. Polarization-resolved SHG and THG measurements exhibit pronounced anisotropy, with emission patterns well-described by theoretical models derived from the khi(2) and khi(3) tensor elements. The material demonstrates exceptionally high nonlinear susceptibilities, with khi(2) ~ 10^(-8) m/V and khi(3) ~ 10^(-17) m^2/V^2 which is a few orders of magnitude greater than comparable 2D materials reported in the literature. Temperature-dependent SHG and THG measurements from 300 K to 25 K reveal exponential decay in harmonic signal intensities, attributed to reduced carrier mobility, with no evidence of structural phase transitions, consistent with results from single crystal diffraction and heat capacity measurements. Polarization-resolved SHG and THG measurements also reveal distinct orientation and ellipticity trends, highlighting the anisotropic nonlinear tensor contributions and contrasting polarization selection rules in the material. These results establish AgScP2S6 as a high-performance, thermally stable, and highly anisotropic nonlinear candidate material suitable for compact photonic applications such as ultrafast optical modulators, polarization-sensitive detectors, and wavelength-tunable light sources.

A novel scheme for Zernike-type hard X-ray phase-contrast imaging is proposed. The scheme relies on X-ray dynamical diffraction in a two-block crystal system with parallel crystal plates of equal thickness. A phase shifter providing a $π/2$ phase shift is placed in the inter-block gap of the crystal system. The method operates in a scanning geometry. The proposed imaging setup is compact and does not require conventional focusing optics. Numerical simulations of phase-contrast image formation are performed.

The present study investigates photochromic oxygen-containing yttrium hydride (YHO) thin films deposited by reactive high power impulse magnetron sputtering (HiPIMS) and compares their photochromic, optical, and structural properties with those of films synthesized by reactive pulsed direct current magnetron sputtering (pulsed-DCMS). Optical emission spectroscopy reveals that, unlike pulsed-DCMS where Ar$^{+}$ ions dominate, HiPIMS discharges are characterised by strong Y$^{+}$ emission, evidencing high yttrium ionisation and substantial self-sputter recycling. The critical working pressure (P$_c$) required to obtain transparent and photochromic films is higher for HiPIMS (Pc $\approx$ 1.0 Pa) than for pulsed-DCMS (Pc $\approx$ 0.5 Pa). Although films deposited near Pc exhibit similar solar transmittance (~72 %) and lattice parameters (5.38--5.39 Å), the pulsed-DCMS film shows a substantially higher relative photochromic contrast (34 %) and a lower optical band gap (2.70 eV) compared with the HiPIMS film (9 % contrast and 2.94 eV). This difference is partly attributed to a lower oxygen-to-hydrogen atomic ratio in the pulsed-DCMS film. Structurally, HiPIMS films are largely polycrystalline with random out-of-plane crystallographic orientation, whereas pulsed-DCMS films exhibit a pronounced <100> out-of-plane preferred orientation. These results demonstrate that, beyond composition, thin-film growth conditions and microstructure play a crucial role in governing the photochromic performance of YHO.

Taxation constitutes a fundamental component of modern national economic systems, exerting profound impacts on both societal functioning and governmental operations. In this paper, we employ an interdependent network approach to model the coevolution between citizens and regulators within a taxation system that fundamentally constitutes a public goods game framework with complex interactive dynamics. In a game layer, citizens engage in public goods games, facing the social dilemma of tax compliance (cooperation) versus evasion (defection). Tax compliance supports the sustainability of public finances while tax evasion presents markedly stronger short-term incentives. In a regulatory layer, fair regulators punish tax evaders, while corrupt regulators keep silent due to bribes. Governmental regulatory interventions introduce critical institutional constraints that alter the traditional equilibrium of the game. Importantly, there exists a strategy update not only among citizens but also among regulators. Our results indicate that strengthening penalties can effectively curb tax evasion, and the influence of bribery on both tax compliance rates and the proportion of fair regulators is nonlinear. Additionally, increasing regulators' salaries and intensifying the crackdown on corrupt regulators can foster the emergence of fair regulators, thereby reducing tax evasion among citizens. The results offer practical policy implications, suggesting that balanced deterrence and institutional fairness are essential to sustaining compliance, and point to the need for future empirical validation and model extensions.

Measuring the phase of light is fundamental to optical imaging, sensing, and signal processing applications. Conventional optical phase measurements rely on multipath configurations, bulky interferometric setups, and computationally intensive data pipelines, limiting scalability, robustness, and practicality. We introduce a technique that allows for reference-free in-line phase retrieval of abrupt phase transitions in optical pulses directly from spectral measurements. Theory, simulations, and experiments concurrently explain the effect as a result of a Kerr-mediated interference between a projected linear wave component and the high-intensity residual of the phase-altered pulse. Utilizing this phenomenon, we demonstrate algorithm-free phase measurements of up to π/385 sensitivity and shot-to-shot signal prominence at 13 dB above noise at 80 MHz rates and 50 pJ pulse energies. This approach offers new paths toward the use of femtosecond pulses as broadband data carriers for optical communications, information processing, and direct high-throughput phase imaging.

An experimental investigation of intracavity dispersion effects in a synchronously pumped optical parametric oscillator (SPOPO) is presented. A flexible setup combining spectral and phase shaping of both pump and local oscillator fields with frequency-resolved balanced homodyne detection is employed to examine how intracavity dispersion influences squeezing. Different cavity configurations with varying finesse and dispersion conditions are studied, and the squeezing is measured as a function of pump power and local oscillator bandwidth. Contrary to expectations based on existing theoretical models, the measured squeezing levels remain essentially unchanged as dispersion varies. To account for these observations, a modeling approach is introduced in which intracavity dispersion is described as an effective spectral filtering occurring at the stage of SPOPO supermode generation. Within this framework, the filtering is incorporated directly into the interaction Hamiltonian of the nonlinear process. This perspective establishes a consistent experimental benchmark for the study of dispersion in SPOPOs and underscores the importance of spectral filtering in the interpretation of multimode squeezing experiments.

In High-Energy Physics, unfolding is the process of reconstructing true distributions of physical observables from detector-distorted measurements. Starting from its reformulation as a regularized quadratic optimization, we develop a framework to tackle this problem using both classical and quantum-compatible methods. In particular, we derive a Quadratic Unconstrained Binary Optimization (QUBO) representation of the unfolding objective, allowing direct implementation on quantum annealing and hybrid quantum-classical solvers. The proposed approach is implemented in QUnfold, an open-source Python package integrating classical mixed-integer solvers and D-Wave's hybrid quantum solver. We benchmark the method against widely used unfolding techniques in RooUnfold, including response Matrix Inversion, Iterative Bayesian Unfolding, and Singular Value Decomposition unfolding, using synthetic dataset with controlled distortion effects. Our results demonstrate that the optimization-based approach achieves competitive reconstruction accuracy across multiple distributions while naturally accommodating regularization within the objective function. This work establishes a unified optimization perspective on unfolding and provides a practical pathway for exploring quantum-enhanced methods in experimental HEP data analysis.

Quantum noise limits the sensitivity of optical measurements, but squeezed states of light enable quantum-enhanced metrology, sensing, and information processing. Most on-chip squeezed-light sources rely on Kerr ($χ^{(3)}$) nonlinearities, remain limited by pump power and excess loss constraints. Quadratic ($χ^{(2)}$) platforms instead provide stronger parametric interactions, lower pump power requirements, and greater spectral engineering flexibility. Here, we demonstrate strong, broadband squeezed-light generation on a thin-film lithium niobate (TFLN) photonic chip using a dual-resonant optical parametric amplifier implemented in a single periodically poled LN (PPLN) microresonator. Near-full-depth domain inversion is achieved simultaneously with highly over-coupled resonances, exhibiting escape efficiencies exceeding 90% and intrinsic quality factors above 2.5 million in a 0.6 mm$^2$ X-cut TF-PPLN resonator, enabling efficient squeezing at 1587 nm when pumped at 793.5 nm. Operating in the continuous-wave regime, we directly measure -0.81 dB of squeezing below the shot-noise limit with a pump power of 27 mW, together with +4.29 dB of anti-squeezing. From these measurements, we infer an on-chip squeezing level of -7.52 dB $\pm$ 0.22 dB (95% confidence interval: [-7.96,-7.10] dB), and an on-chip anti-squeezing level of +9.62 dB $\pm$ 0.25 dB. We demonstrate single-mode squeezing at degeneracy with a squeezed-light spectrum exceeding 10.3 THz. This work reports the highest squeezing ratio among integrated $χ^{(2)}$ cavity platforms and the first quasi-phase matched, fully resonant $χ^{(2)}$ cavity squeezer on chip, establishing a scalable route to fully integrated power-efficient squeezed-light sources for quantum-enhanced sensing and metrology.

The number resolution of solid-state artificial atoms is of fundamental interest for the study of quantum few-body systems, yet remains experimentally challenging. Quantum optical experiments offer a non-invasive approach which links up macroscopic measurements with the quantity of quantum emitters. In this work, we propose a time-domain quantum optical methodology for the strict numbering of colloidal CdSe/CdS/ZnS quantum dots (QDs) confined in subwavelength-size polystyrene capsules. The non-polarized, homogeneously broadened emission of colloidal QDs in the subwavelength volume satisfies the description of Dicke's superradiance of identical quantum emitters. An analytic relation describes the numerical dependence of the second-order photon correlation on the number and the collective lifetime of emitters, yielding an experimental counting range of colloidal QDs from one to ten. This work provides a robust pathway for the non-invasive numbering of artificial atoms and the investigation of collective light-matter interactions at the nanoscale.

As widely studied topological singularities, bound states in the continuum (BICs) have revealed rich physical properties through their momentum-space topology. Here, we reveal and experimentally demonstrate that magnetically induced chiral BICs possess multidimensional topological structures extending into real space. We design and realize a gyromagnetic photonic crystal slab where magnetic field breaks the time-reversal symmetry and lifts the degeneracy of BICs, creating a pair of chiral BICs with opposite circular polarizations. Near-field scanning measurements reveal phase vortices with quantized topological charges, spatially distributed near-field chirality, and skyrmionic Stokes textures arising from magnetic control. Our work unveils a previously unexplored dimension of BIC topology and establishes gyromagnetic photonic crystals as versatile platforms for manipulating complex topological states.

A thermodynamic framework for asymptotic inference is developed in which sample size and parameter variance define a state space. Within this description, Shannon information plays the role of entropy, and an integrating factor organizes its variation into a first-law-type balance equation. The framework supports a cyclic inequality analogous to a reversed second law, derived for the estimation of the mean. A non-trivial third-law-type result emerges as a lower bound on entropy set by representation noise. Optimal inference paths, global bounds on information gain, and a natural Carnot-like information efficiency follow from this structure, with efficiency fundamentally limited by a noise floor. Finally, de Bruijn's identity and the I-MMSE relation in the Gaussian-limit case appear as coordinate projections of the same underlying thermodynamic structure. This framework suggests that ensemble physics and inferential physics constitute shadow processes evolving in opposite directions within a unified thermodynamic description.

We explore with ab-initio theory high harmonic generation (HHG) from a series of gas-phase fullerenes (from C$_{20}$ to C$_{60}$, including isomers) under varying laser conditions (different ellipticities, angular orientations, intensities and wavelengths). We find that HHG emission from fullerenes exhibits a prominent high-energy second plateau, extending well above the expected semi-classical cutoff, e.g. for an 800 nm driving laser with a peak intensity of $10^{14}$ W/cm$^2$, the anomalous $2^{nd}$ plateau cutoff is 115 eV. We theoretically analyze the underlying $2^{nd}$ plateau physical mechanism and determine that: (i) It differs from the standard SFA-like mechanism; (ii) It is recombination- based; (iii) It has an unusual inverted cutoff scaling with wavelength (whereby the cutoff decreases with increasing wavelength, until reaching a saturation); and (iv) It exhibits a linear cutoff dependence with electric field amplitude E$_0$ . We further rule out real-space trajectory based pictures, indicating that the most likely culprits are sharp quantum mechanical resonances in the fullerene family. Our work establishes a new route for high energy coherent broadband emission without requiring mid-IR driving and uncovers a novel HHG mechanism in complex structure quantum systems.

This paper presents a detailed description and characterization of a new open-source vehicle trajectory dataset, namely SWIFTraj, constructed from videos recorded by a swarm of 16 drones equipped with 5.4K-resolution cameras. The dataset is distinguished from existing open-source trajectory datasets in several aspects. First, it provides long-distance continuous trajectories of up to 4.5 km on a freeway, enabling in-depth investigation of traffic phenomena and their spatial and temporal evolution. Second, the data collection site covers an integrated network consisting of a long freeway corridor and parts of its connected urban network, facilitating traffic analysis and modeling from a network perspective. The potential applications of the dataset for transportation research, including traffic flow analysis, modeling, and control at multiple scales, as well as topics related to autonomous driving, are thoroughly discussed. Finally, SWIFTraj is released as a freely accessible open-source dataset to support and accelerate future research in the transportation community. The dataset is publicly available at the SWIFTraj website (https://www.swiftraj.com).

Photonic neuromorphic computing promises revolutionary advances in parallel and high-speed processing, yet a key challenge persists: co-integrating nonlinearity, dense connectivity, and intrinsic memory monolithically to enable brain-inspired, spatiotemporal information processing. Here, we overcome this challenge by introducing a monolithic photonic recurrent network based on an active metasurface operating at bound state in the continuum (BIC). The BIC mode mediates strong,long-range coupling across the lattice, creating a reconfigurable recurrent network topology in hardware. Concurrently, the gain medium provides both optical nonlinearity for neuronal activation and a finite carrier lifetime that serves as a built in, analog temporal memory. This synergy enables computation to emerge directly from the collective spatiotemporal dynamics of the driven-dissipative photonic system, effectively realizing a physical reservoir computer on a chip. We experimentally validate a minimal yet physically complete system on benchmark tasks: brain MRI image classification and human action recognition, achieving 92.16% and 85.36% accuracies, respectively. This work establishes a scalable pathway toward ultrafast, energy-efficient neuromorphic intelligence where processing is an inherent property of tailored light matter interaction.

Ultrafast lasers create extreme, non-equilibrium thermodynamic conditions that can transiently reach pressures and temperatures comparable to interior core of the earth. Here we show that femtosecond excitation of amorphous silica-hafnia multilayer dielectrics drives the formation of high-pressure crystalline phases of silica including stishovite, seifertite, and the pyrite-type high density structure, within confined subsurface regions.Using TEM, SAED, and 4D-STEM, we directly map nanoscale phase evolution and identify crystalline motifs embedded inside laser generated blisters.Complementary molecular dynamics simualtions reveal the thermodynamic pathway underlying these transformations, where rapid electronic pressure initiates densification and octahedral coordination, followed by temperature driven crystallization and displacive transitions during ultrafast quenching. The resulting polymorphs reflects a dual-stage pathway inaccessible under equilibrium processing. Our results establish femtosecond laser excitation as a viable route to synthesize and stabilize ultrahigh-density high pressure silica phases under ambient conditions, without a diamond anvil cell, with implications for laser-damage mechanisms, high-energy-density materials, and planetary physics.

To leverage the redundancy between the electronic structure computed at each step of first-principles molecular dynamics, we present a data-driven modeling framework for Kohn-Sham Density Functional Theory that bypasses the explicit optimization of electronic wavefunctions. We sample a priori representative atomic configurations and construct a low-dimensional basis that efficiently approximates the electronic structure subspace. Subsequently, we employ this reduced basis in a direct solver for the electronic single particle density matrix, thereby enabling the efficient determination of ground state without iterative wavefunction optimization. We demonstrate the efficacy of our approach in a Born-Oppenheimer molecular dynamics of a water molecule, showing that the resulting simulations accurately reproduce key structural properties, such as bond lengths and bond angle, obtained from full first-principles molecular dynamics. This work highlights the potential of data-driven approaches to develop efficient electronic structure solvers for first-principles simulations.

Quantics Tensor Train (QTT) operations such as matrix product operator contractions are prohibitively expensive for large bond dimensions. We propose an adaptive patching scheme that exploits block-sparse QTT structures to reduce costs through divide-and-conquer, adaptively partitioning tensors into smaller patches with reduced bond dimensions. We demonstrate substantial improvements for sharply localized functions and show efficient computation of bubble diagrams and Bethe-Salpeter equations, opening the door to practical large-scale QTT-based computations previously beyond reach.

Dynamically tunable surface textures offer a powerful route to spatiotemporally regulate surface and interfacial properties, enabling emerging applications ranging from adaptive optics to soft robotic manipulation. However, achieving programmable, reversible, and spatiotemporal modulation of surface texture remains a fundamental challenge. Here, we present a photothermal-actuated liquid crystal elastomer bilayer that enables reversible, on-demand spatiotemporal modulation of surface textures through dynamically emerging and propagating wrinkles. Using direct laser writing or projected light fields, programmable and self-erasable wrinkle patterns are generated for dynamic information encoding. This spatiotemporal wrinkling enables object manipulation across diverse geometries, including uphill transport and navigation along predesigned paths. By coupling wrinkle-driven motion with thermally reversible dynamic bonding, the bilayer further enables assembly and disassembly of dynamic polymers, as well as cargo transportation. This work demonstrates spatiotemporally programmable wrinkling as a powerful mechanism for dynamic modulation of surface textures, establishing a versatile platform for multifunctional and reconfigurable smart surfaces.

Quantum key distribution (QKD) enables information-theoretic secure communication, yet its ultimate tolerance to noise and achievable transmission distance remain fundamentally constrained. We establish the maximum quantum bit error rate (QBER) compatible with secure QKD and derive corresponding upper bounds on communication distance. Our results follow from a fundamental capacity threshold for qubit Pauli channels and apply to protocols based on two or more mutually unbiased bases, using either single-photon or weak coherent sources. By connecting information-theoretic limits to realistic physical noise models, we obtain universal bounds on achievable distances in fiber and free-space links, including diffraction-limited constraints relevant to deep-space quantum communications. These findings clarify the ultimate noise robustness of QKD and delineate the fundamental boundaries of secure quantum communication.

We present an investigation into the crystal structure of ErVO4 under variable pressure conditions. The high-pressure single crystal X-ray diffraction experiments performed employing helium as the pressure medium facilitated structure refinements up to 24.1(2) GPa. The transition from zircon to scheelite was observed at a pressure of 7.9(1) GPa. In contrast to previous reports, we did not detect any sign of phase coexistence. We also did not observe the second phase transitions previously predicted by density-functional theory to occur below 20 GPa. The determination of the pressure dependence of unit-cell parameters and volume yields precise values for linear compressibility of each axis and the pressure-volume equation of state for both the zircon and scheelite phases. Additional information on the mechanical properties of ErVO4, obtained from density-functional theory calculations, is also reported.

We investigate phase-dependent electronic and excitonic phenomena in monolayer Y2TeO2 MOenes in the 1T and 2H polymorphs using first-principles theory and an effective many-body framework. Phonon spectra and elastic stability criteria establish both phases as dynamically and mechanically stable. Quasiparticle band structures reveal direct gaps in the near-infrared to visible range, with gap values increasing systematically from semilocal to hybrid exchange treatments. Optical spectra computed using a tight-binding Bethe-Salpeter approach demonstrate pronounced excitonic resonances arising from reduced dimensionality and weak dielectric screening. The exciton binding energies reach 152 meV in the 1T phase and 126 meV in the 2H phase, reflecting enhanced quantum confinement in the structurally denser phase. Our results identify Y2TeO2monolayers as a rare class of stable, direct-gap MOenes with strong excitonic effects, providing a platform for exploring many-body physics in low-dimensional oxychalcogenide systems especially for photovoltaic applications.

Colloidal quantum dots (QDs) integrated with waveguide-coupled dielectric resonators are promising building blocks for compact on-chip light sources. However, deterministic placement of QDs with strong mode overlap at the desired location remains a challenge. Here, we demonstrate a simple and scalable strategy for integrating colloidal QDs with a waveguide-coupled Si3N4 micro-ring resonator platform and for controlling the radiative dynamics of both bright and dark excitons via Purcell enhancement. We use strongly quantum-confined CsPbBr3 QDs, which exhibit bright-exciton emission at room-temperature, while emission at cryogenic temperatures originates from both bright and dark excitons. The CsPbBr3 QDs are selectively retained on the Si3N4 micro-ring cavities through a spin-coating/rinsing process, enabling efficient overlap with whispering-gallery modes and routing of the emission through integrated waveguides. We confirm accelerated decay of emission from both bright and dark excitons for CsPbBr3 QDs coupled to the micro-ring cavities. These results demonstrate an effective route to integrate colloidal QDs with Si3N4 micro-ring cavities and to leverage cavity-enhanced emission in scalable integrated photonic devices.

Counterfeiting of premium spirits poses significant economic and health risks, that could be tackled by robust, accurate, portable and non-destructive through-bottle measurements. Here, we demonstrate the capability of focus-matched inverse spatially offset spectroscopy, combining fluorescence and Raman signals, for authenticating Cognac and Brandy. The technique accurately identifies age classification, bottling year, spoilage due to elevated storage temperatures, and distinguishes between Cognac brands. Critically, our method effectively differentiates genuine Cognacs from counterfeit products, correctly identifying 98% of counterfeit samples. This shows the promise of through-bottle spectroscopy as a powerful tool for supply chain integrity and consumer protection in the high-value spirits market.

We investigate discrete wavelength switching in single-gain-section multi-wavelength lasers monolithically integrated on InP with phase-controlled optical-feedback. By modulating the feedback phase, nanosecond-scale wavelength switching is experimentally demonstrated with transition times below 2.5 ns. Measurements consistently show that the switching time decreases with stronger optical feedback and larger phase-modulation amplitudes. Transitions from lower to higher modal gain are faster. We support the experimental observations with a multi-mode extension of the Lang-Kobayashi rate-equation model. We analyze the influence of laser, feedback-cavity, and modulation parameters on the switching dynamics, and highlight the role of mode coupling. These results highlight the potential of integrated multi-wavelength lasers for compact and high-speed all-optical networking systems.

Lagrangian particle tracking is essential for characterizing turbulent flows, but inferring particle acceleration from inherently noisy position data remains a significant challenge. Fluid particles in turbulence experience extreme, intermittent accelerations, resulting in heavy-tailed probability density functions (PDFs) that deviate strongly from Gaussian predictions. Existing filtering techniques, such as Gaussian kernels and penalized B-splines, implicitly assume Gaussian-distributed jerk, thereby penalizing sparse, high-magnitude acceleration changes and artificially suppressing the intermittent tails. In this work, we develop a novel maximum likelihood estimation (MLE) framework that explicitly accounts for this non-Gaussian intermittency. By formulating a modified Gaussian process to model the random incremental forcing, we introduce a sparse optimization scheme utilizing a convex 1-norm relaxation. To overcome the numerical stiffness associated with high-order difference operators, the problem is efficiently solved using an iteratively reweighted least squares (IRLS) algorithm. The proposed filter is evaluated against direct numerical simulation (DNS) data of homogeneous, isotropic turbulence (Re approx. 310). Results demonstrate that the IRLS approach consistently outperforms state-of-the-art discrete MLE, continuous MLE, and B-spline methods, yielding systematic reductions in root-mean-squared error (RMSE) across position, velocity, and acceleration. Most importantly, the proposed framework succeeds in better recovering the heavy-tailed statistical structure of both acceleration and acceleration differences (jerk) across temporal scales, preserving the physical intermittency characteristic of high-Reynoldsnumber turbulent flows that baseline methods severely attenuate.

Quantized transport is a prominent feature in topological physics, with canonical examples being the quantum Hall effect and adiabatic Thouless pump, which are based on the Chern number, a topological invariant of 2D systems. Going beyond the Chern-number-based paradigms, quantized transports can also arise from k-direction quasienergy winding unique to periodically driven (Floquet) systems, which are free of dimensionality and adiabaticity limitations. However, lattices displaying winding of their quasienergy bands require asymmetric long-range couplings that are difficult to achieve in lattices of real-space coupled sites. Here, by leveraging photonic synthetic dimensions we construct asymmetric long-range-couplings in a one-dimensional temporal quantum walk based on three coupled fiber loops. We demonstrate quantized transport arising from the winding of quasienergy bands in k direction. We show that the average group velocity of an initial wave packet is proportional to the winding number, which leads to a quantized transport displacement. To better visualize this quantized displacement, we cascade two regions with flipped nearest/long-range couplings and observe a focusing effect with a quantized spatial shift in the focusing point. We also probe the robust properties of quantized transport against obstacles and disorders. The study initiates quasienergy-winding-based topological transports, which can feature applications in precise and robust imaging and information processing.

Acoustoelectric interactions occur when free carriers in a semiconductor interact with the fields of an acoustic wave in a piezoelectric medium. These interactions can amplify acoustic waves, as well as give rise to extremely large phononic nonlinearities and strong non-reciprocal effects. The field of acoustoelectric devices is currently dependent on analytical and perturbative solutions for the two simplest arrangements of piezoelectric-semiconductor materials. While these canonical models have allowed the field to advance substantially, new geometries are arising that do not satisfy assumptions integral to these models. These assumptions include the treatment of the interactions between the acoustic fields and free carriers as weak, the neglect of the tensorial nature of the material properties, the omission of the spatial variations in the phonons' electric field profiles, and the disregard of elastic coupling across material boundaries, among others. We develop, for the first time, a finite element method (FEM) model to solve for acoustoelectric interactions in arbitrary geometries that avoids making the assumptions of the canonical models. We verify the FEM model using results for amplification, dispersion, and non-reciprocity obtained from the canonical models in their regime of validity. We then examine the acoustoelectric effect in two geometries not covered by the canonical models: a thin piezoelectric film placed on a semiconductor substrate and a fully 2D waveguide under a thin semiconductor layer. This work lays the foundation for accurate modeling of arbitrary acoustoelectric geometries such as those currently being developed for all-acoustic radio frequency (RF) signal processing, acoustoelectrically enhanced photonic devices, and quantum acoustoelectric devices.

Assessing whether two datasets are distributionally consistent has become a central theme in modern scientific analysis, particularly as generative artificial intelligence is increasingly used to produce synthetic datasets whose fidelity must be rigorously validated against the original data on which they are trained, a task made more challenging by the continued growth in data volume and problem dimensionality. In this work, we propose the use of arithmetic coding to provide a lossless and invertible compression of datasets under a physics-informed probabilistic representation. Datasets that share the same underlying physical correlations admit comparable optimal descriptions, while discrepancies in those correlations-arising from miscalibration, mismodeling, or bias-manifest as an irreducible excess in code length. This excess codelength defines an operational fidelity metric, quantified directly in bits through differences in achievable compression length relative to a physics-inspired reference distribution. We demonstrate that this metric is global, interpretable, additive across components, and asymptotically optimal in the Shannon sense. Moreover, we show that differences in codelength correspond to differences in expected negative log-likelihood evaluated under the same physics-informed reference model. As a byproduct, we also demonstrate that our compression approach achieves a higher compression ratio than traditional general-purpose algorithms such as gzip. Our results establish lossless, physics-aware compression based on arithmetic coding not as an end in itself, but as a measurement instrument for testing the fidelity between datasets.

We propose a minimal mathematical framework to describe the go-or-grow dynamics of tumor cells comprising two phenotypically distinct populations. One population is migratory and undergoes linear diffusion, while the other proliferates in an oxygen-dependent manner. The local oxygen concentration governs transitions between these phenotypes. We then ask whether these two coupled phenotype-specific equations can be reduced to a single mixed-phenotype equation under cyclic hypoxia. We establish a connection between the minimal go-or-grow model with distinct phenotypic populations and a reduced model describing a single-cell population with oxygen-dependent diffusion and proliferation in the fast-phenotypic-switching regime. This theoretical reduction is validated through numerical simulations.

Branching processes are models used to describe populations that reproduce and die over time. In the classical setting, an individual's reproductive capacity remains constant throughout its lifetime. However, in real-world situations, reproductive capacity typically undergoes ageing - that is, after reaching a peak, it decreases over time. In this work, we study the influence of ageing on the behaviour of the process and how modifying its parameters, along with reproduction rates, affects the destiny of the process.

The introduction of stochasticity into continuous ecological models frequently relies on phenomenological, diagonal diffusion terms that lack a rigorous microscopic basis. We demonstrate that this standard practice fundamentally misrepresents the geometry of demographic fluctuations. By deriving a stochastic Rosenzweig--MacArthur model directly from an integer-valued, Bernoulli-coupled continuous-time Markov chain, we isolate the exact diffusion covariance structure dictated by event stoichiometry. We mathematically prove that coupled predation--conversion events inherently generate a structurally negative predator--prey cross-covariance, exposing the severe mathematical and biological limitations of standard diagonal-noise approximations. Furthermore, we resolve a persistent ambiguity in stochastic population modeling by explicitly formalizing the bifurcation between open-domain formulations (for survival-conditioned interior dynamics) and absorbed formulations (for extinction-permitting dynamics). To rigorously support this distinction, we develop a tailored two-stage Lyapunov well-posedness architecture that separates non-explosion criteria from boundary-barrier positivity invariance. By bridging microscopic event stoichiometry with macroscopic boundary-degenerate diffusions, this work replaces ad hoc noise constructs with a definitive, mathematically exact template for covariance-consistent and boundary-aware ecological modeling.

Cultures of neurons grown on multi-electrode arrays have become a common experimental preparation for investigating developing neural networks. Experiment and simulation have shown that these developing networks eventually exhibit bursting behavior in which the entire culture participates for short periods of time, with inter-burst intervals in which the network is comparatively quiescent. This paper extends previous simulation results by examining the spatiotemporal patterns of such bursting. We show that these bursts originate at a small number of network locations and propagate as waves of activity. We demonstrate that this type of activity does not require fine tuning of neuron or network parameters. We also examine how this activity changes during development and the dependence of such activity and its triggering on both local and global network properties.

The maximum agreement forest (MAF) problem in phylogenetics takes as input a set t >=2 of binary phylogenetic trees T on the same set of taxa X. It asks for a partition X into the smallest number of blocks such that the subtrees induced by these blocks are disjoint and have common topology across all the trees in T. We produce a modified version of the well-known chain reduction rule in order to prove the existence of a kernel of size O( t * r * k ) where k is the natural parameter (the number of blocks) and r=min{max{k,3},t+1}}. We prove this bound for both the unrooted and rooted version of the problem, and demonstrate that the bound r, the length to which common chains are truncated, is tight. Our results constitute the first kernels for MAF in the t > 2 regime.

Microwave sounding is the leading driver of global numerical weather forecasting, but is limited by the scalability of such instruments. With modern machining and commercial microwave components, it is now possible to design low size, weight, power, and cost (SWaP-C) microwave spectrometers while maintaining wide bandwidth performance. Here we report on the status of CubeSounder, a spectrometer tailored for water vapor radiometry that utilizes passive wave guide filter banks. After developing a prototype and high altitude balloon payload, we demonstrated CubeSounder on commercial stratospheric balloon flights. We report on our design process, especially the simulation and fabrication of the custom millimeter-wave filter banks. We also report the initial results of the data collected from the balloon flights.

The signal bandwidth of Digital to Analog Converters based on Sigma Delta Modulation is limited by speed constrains. Time-Interleaving allows coping with complexity vs. speed by replacing the original architecture by M parallel paths. These path are clocked at a frequency M times smaller and their digital outputs time multiplexed. This is then converted to analog by means of a Digital to Analog Converter clocked at the high rate. This preprint proposes that time multiplexing be performed in the analog domain. As a result robustness against dynamic effects is achieved.

We derive a scaling law relating ADC bit depth to effective bandwidth for signals with $1/f^α$ power spectra. Quantization introduces a flat noise floor whose intersection with the declining signal spectrum defines an effective cutoff frequency $f_c$. We show that each additional bit extends this cutoff by a factor of $2^{2/α}$, approximately doubling bandwidth per bit for $α= 2$. The law requires that quantization noise be approximately white, a condition whose minimum bit depth $N_{\min}$ we show to be $α$-dependent. Validation on synthetic $1/f^α$ signals for $α\in \{1.5, 2.0, 2.5\}$ yields prediction errors below 3\% using the theoretical noise floor $Δ^2/(6f_s)$, and approximately 14\% when the noise floor is estimated empirically from the quantized signal's spectrum. We illustrate practical implications on real EEG data.

Beamforming in multiple-input multiple-output (MIMO) systems should take interference mitigation into account. However, for beamform design, accurate channel state information (CSI) is needed, which is often difficult to obtain due to channel variability, feedback overhead, or hardware constraints. For example, amplify-and-forward (AF) relays passively forward signals without measurement, precluding full CSI acquisition to and from the relay. To address these issues, this paper introduces a novel prediction-assisted optimization (PAO) framework for beamform design in AF relay-assisted multiuser MIMO systems. The proposed solution in the AF relay aims at maximizing the signal-plus-interference-to-noise ratio (SINR). Unlike other methods, PAO relies solely on received power measurements, making it suitable for scenarios where CSI is unreliable or unavailable. PAO consists of two stages: a supervised-learning-based neural network (NN) that predicts the positions of transmitters using signal observations, and an optimization algorithm, guided by a digital twin (DT), that iteratively refines the beam direction of the relay in a simulated radio environment. As a key contribution, we validate the proposed framework using realistic measurements collected on a custom-built experimental millimeter wave (mmWave) platform, which enables training of the NN model under practical wireless conditions. The estimated information is then used to update the digital twin with knowledge of the surrounding environment, enabling online optimization. Numerical results show the trade-off between localization accuracy and beamforming performance and confirm that PAO maintains robustness even in the presence of localization errors while reducing the need for real-world measurements.

Counting immunopositive cells on biological tissues generally requires either manual annotation or (when available) automatic rough systems, for scanning signal surface and intensity in whole slide imaging. In this work, we tackle the problem of counting microglial cells in lumbar spinal cord cross-sections of rats by omitting cell detection and focusing only on the counting task. Manual cell counting is, however, a time-consuming task and additionally entails extensive personnel training. The classic automatic color-based methods roughly inform about the total labeled area and intensity (protein quantification) but do not specifically provide information on cell number. Since the images to be analyzed have a high resolution but a huge amount of pixels contain just noise or artifacts, we first perform a pre-processing generating several filtered images {(providing a tailored, efficient feature extraction)}. Then, we design an automatic kernel counter that is a non-parametric and non-linear method. The proposed scheme can be easily trained in small datasets since, in its basic version, it relies only on one hyper-parameter. However, being non-parametric and non-linear, the proposed algorithm is flexible enough to express all the information contained in rich and heterogeneous datasets as well (providing the maximum overfit if required). Furthermore, the proposed kernel counter also provides uncertainty estimation of the given prediction, and can directly tackle the case of receiving several expert opinions over the same image. Different numerical experiments with artificial and real datasets show very promising results. Related Matlab code is also provided.

Improper priors are not allowed for the computation of the Bayesian evidence $Z=p({\bf y})$ (a.k.a., marginal likelihood), since in this case $Z$ is not completely specified due to an arbitrary constant involved in the computation. However, in this work, we remark that they can be employed in a specific type of model selection problem: when we have several (possibly infinite) models belonging to the same parametric family (i.e., for tuning parameters of a parametric model). However, the quantities involved in this type of selection cannot be considered as Bayesian evidences: we suggest to use the name ``fake evidences'' (or ``areas under the likelihood'' in the case of uniform improper priors). We also show that, in this model selection scenario, using a diffuse prior and increasing its scale parameter asymptotically to infinity, we cannot recover the value of the area under the likelihood, obtained with a uniform improper prior. We first discuss it from a general point of view. Then we provide, as an applicative example, all the details for Bayesian regression models with nonlinear bases, considering two cases: the use of a uniform improper prior and the use of a Gaussian prior, respectively. A numerical experiment is also provided confirming and checking all the previous statements.

Parameter estimation of nonlinear state-space models from input-output data typically requires solving a highly non-convex optimization problem prone to slow convergence and suboptimal solutions. This work improves the reliability and efficiency of the estimation process by decomposing the overall optimization problem into a sequence of tractable subproblems. Based on an initial linear model, nonlinear residual dynamics are first estimated via a guided residual search and subsequently refined using multiple-shooting optimization. Experimental results on two benchmarks demonstrate competitive performance relative to state-of-the-art black-box methods and improved convergence compared to naive initialization.

In this work, we analyze alternative effective sample size (ESS) metrics for importance sampling algorithms, and discuss a possible extended range of applications. We show the relationship between the ESS expressions used in the literature and two entropy families, the Rényi and Tsallis entropy. The Rényi entropy is connected to the Huggins-Roy's ESS family introduced in \cite{Huggins15}. We prove that that all the ESS functions included in the Huggins-Roy's family fulfill all the desirable theoretical conditions. We analyzed and remark the connections with several other fields, such as the Hill numbers introduced in ecology, the Gini inequality coefficient employed in economics, and the Gini impurity index used mainly in machine learning, to name a few. Finally, by numerical simulations, we study the performance of different ESS expressions contained in the previous ESS families in terms of approximation of the theoretical ESS definition, and show the application of ESS formulas in a variable selection problem.

Soil texture is a foundational attribute that governs water availability and erosion in agriculture, as well as load bearing capacity, deformation response, and shrink-swell risk in geotechnical engineering. Yet texture is still typically determined by slow and labour intensive laboratory particle size tests, while many sensing alternatives are either costly or too coarse to support routine field scale deployment. This paper proposes a robust and field deployable multispectral imaging (MSI) system and machine learning framework for predicting soil composition and the United States Department of Agriculture (USDA) texture classes. The proposed system uses a cost effective in-house MSI device operating from 365 nm to 940 nm to capture thirteen spectral bands, which effectively capture the spectral properties of soil texture. Regression models use the captured spectral properties to estimate clay, silt, and sand percentages, while a direct classifier predicts one of the twelve USDA textural classes. Indirect classification is obtained by mapping the regressed compositions to texture classes via the USDA soil texture triangle. The framework is evaluated on mixture data by mixing clay, silt, and sand in varying proportions, using the USDA classification triangle as a basis. Experimental results show that the proposed approach achieves a coefficient of determination R^2 up to 0.99 for composition prediction and over 99% accuracy for texture classification. These findings indicate that MSI combined with data-driven modeling can provide accurate, non-destructive, and field deployable soil texture characterization suitable for geotechnical screening and precision agriculture.

Channel knowledge map (CKM) exploits environ-ment information to assist channel estimation during communi-cation. For clustered channels, which represent a typical type ofwireless propagation environment, there has been no researchdevoted to designing an appropriate CKM to enhance theirestimation. To exploit environment information for clusteredchannel, improve channel estimation accuracy and reduce pilotoverhead, we propose ClusterCKM, a CKM providing the rangeof clustered multipath parameters for any pair of transmitter-receiver links in the region of interest. Firstly, we construct Clus-terCKM through estimating the spatial range of scatterer clustersfrom historical channel information. From these spatial range ofscatterer clusters, ClusterCKM infers the range of multipathparameters for the target link. Furthermore, a ClusterCKM-based channel estimation algorithm is developed to utilize theparameter range provided by ClusterCKM. Simulation resultsshow that, more accurate channel estimation can be achievedand pilot overhead can also be reduced by ClusterCKM and theClusterCKM-based estimation algorithm.

This paper introduces CSI-RFF, a new framework that leverages micro-signals embedded within Channel State Information (CSI) curves to realize Radio-Frequency Fingerprinting of commodity off-the-shelf (COTS) WiFi devices for open-set authentication. The micro-signals that serve as RF fingerprints are termed ``micro-CSI''. Through experimentation, we have found that the presence of micro-CSI can primarily be attributed to imperfections in the RF circuitry. Furthermore, this characteristic signal is detectable in WiFi 4/5/6 network interface cards (NICs). We have conducted further experiments to determine the most effective CSI collection configurations to stabilize micro-CSI. Yet, extracting micro-CSI for authentication purposes poses a significant challenge. This complexity arises from the fact that CSI measurements inherently include both micro-CSI and the distortions introduced by wireless channels. These two elements are intricately intertwined, making their separation non-trivial. To tackle this challenge, we have developed a signal space-based extraction technique for line-of-sight (LoS) scenarios, which can effectively separate the distortions caused by wireless channels and micro-CSI. Over the course of our comprehensive CSI data collection period extending beyond one year, we found that the extracted micro-CSI displays unique characteristics specific to each WiFi device and remains invariant over time. This establishes micro-CSI as a suitable candidate for device fingerprinting. Finally, we conduct a case study focusing on area access control for mobile robots. Our experimental results demonstrate that the micro-CSI-based authentication algorithm can achieve an average attack detection rate close to 99% with a false alarm rate of 0% in both static and mobile conditions when using 20 CSI measurements to construct one fingerprint.

We investigated the relationship among neural representations of vocalized, mimed, and imagined speech recorded using publicly available stereotactic EEG recordings. Most prior studies have focused on decoding speech responses within each condition separately. Here, instead, we explore how responses across conditions relate by training linear spectrogram reconstruction models for each condition and evaluate their generalization across conditions. We demonstrate that linear decoders trained on one condition generally transfer successfully to others, implying shared speech representations. This commonality was assessed with stimulus-level discriminability by performing a rank-based analysis demonstrating preservation of stimulus-specific structure in both within- and across-conditions. Finally, we compared linear reconstructions to those from a nonlinear neural network. While both exhibited cross-condition transfer, linear models achieve superior stimulus-level discriminability.

Ultrasonic imaging methods often assume linear direct models, while in reality, many nonlinear phenomena are present, e.g. multiple reflections. A family of imaging methods called Full Waveform Inversion (FWI), which has been developed in the field of seismic imaging, uses full acoustic wave simulations as direct models, taking into account virtually all nonlinearities, which can ultimately enhance the accuracy of ultrasonic imaging. However, the problem of cycle skipping -- the existence of many local minima of the Least Squares (L2) misfit function due to the oscillatory nature of the signals -- is worsened when FWI is applied to ultrasound data because of a lack of low-frequency components. In this paper, we explore the use of the squared Wasserstein (W2) Optimal Transport Distance as the metric for the misfit between the acquired and the synthetic data, applying the method to Nondestructive Evaluation with ultrasonic phased arrays. An analytical continuous time-domain derivation of the adjoint acoustic field related to the W2 misfit is presented and used for the computation of the gradients. To cope with the computational burden of FWI, we apply a low-memory strategy that allows for the computation of the gradients without the storage of the full simulated fields. The GPU implementation of the method (in CUDA language) is detailed, and the source code is made available. Six prototypical cases are presented, and the corresponding sound speed maps are reconstructed with FWI using both the L2 and the W2 misfit functionals. In five of the six cases, the pixel-wise sum of squared errors obtained with W2 was at least one order of magnitude lower than that obtained with W2, with an increase in the gradient computation time not exceeding 2\%. The results highlight both the adequacy of the W2 misfit for ultrasonic FWI with phased arrays and its computational feasibility.

A novel non-orthogonal multiple access (NOMA) based low-delay service framework is proposed for fog radio access networks (F-RANs). Fog access points (FAPs) leverage NOMA for local delivery of cached content, while the cloud access point employs NOMA to simultaneously push content to FAPs and directly serve users. Based on this model, a delay minimization problem is formulated by jointly optimizing user association, cache placement, and power allocation. To address this non-convex mixed-integer nonlinear programming problem, an alternating optimization (AO) algorithm is developed, which decomposes the original problem into two subproblems, namely joint user association and cache placement, and power allocation. In particular, a low-complexity algorithm is designed to optimizing the user association and cache placement strategy using the McCormick envelope theory and Lagrangian partial relaxation. The power allocation is optimized by invoking the successive convex approximation. Simulation results reveal that: 1) the proposed AO-based algorithm effectively balances between the achieved performance and computational efficiency, and 2) the proposed NOMA-based F-RANs framework significantly outperforms orthogonal multiple access-based F-RANs systems in terms of average transmission delay in different scenarios.

The extreme sensitivity of chaotic systems to external perturbations makes them natural candidates for sensing applications. We propose a single-shot echo-based protocol for estimating small rotations about an unknown axis that leverages random symmetric probe states prepared via chaotic dynamics. In contrast to previous protocols for this axis-agnostic rotation sensing problem that depend on difficult-to-prepare anticoherent states, the random probe states used in our protocol can be prepared via constant-depth chaotic circuits composed of random one-axis twisting pulses. We demonstrate analytically that our protocol achieves Heisenberg scaling relative to an arbitrary rotation axis that need not be a priori known. We also investigate the effects of collective and single-particle dephasing in our protocol using analytical and numerical tools. While the requirements on dephasing rates to maintain Heisenberg sensitivity are strict, they are achievable in near-term experiments, for instance, for magnetometric rotosensing with high-spin lanthanide atoms such as dysprosium-164.

Operational probes of the interface between quantum mechanics and general relativity in the Newtonian regime -- via mass-energy equivalence in clocks or spatial superpositions in interferometers -- share a common description in terms of an effective qubit-qubit Ising coupling. Here we generalize both paradigms to interacting $(N+1)$-level effective qudits made of atomic ensembles with particle number, $N$. The many-body enhancement boosts the signal-to-noise and increases the effective interaction rate, facilitating the observation of gravitationally induced entanglement and decoherence, certified by metrological witnesses based on local and collective measurements. Furthermore, we show that quantum effects induced by gravitational interaction can be simulated by trapped bimodal Bose-Einstein condensates with long-range (e.g. dipolar) coupling, providing a programmable analogue platform to explore gravitating quantum dynamics at accessible time and energy scales. Finally, extending the protocol to a sensor network broadens the entanglement-detection window.

This manuscript is the first in a series of instalments that investigate spherically symmetric solutions within the effective dynamics program of Loop Quantum Gravity. The choice of lattice is adapted such that it remains invariant under a set of symmetry transformations maximally mapping spherical symmetry to the discrete setting. The conditions for symmetry restriction of the dynamics are investigated and a subspace is identified to make computations feasible. Afterwards symplectic structure and scalar constraint are explicitly computed on this subspace. This lays the groundwork to target several particular solutions, such $k=1$ cosmology and black holes, which will serve as the subjects of forthcoming follow-up papers.

We determine conditions on classical group algebra codes so that they have pre-orientation for cup products and copy-cup gates. This defines quantum codes that have constant-depth $\operatorname{CZ}$ and $\operatorname{CCZ}$ gates constructed via tensor products of classical group algebra codes, including hypergraph and balanced products. We show that determining the conditions relies on solving the perfect matching problem in graph theory. Conditions are fully determined for the 2- and 3-copy-cup gates, for group algebra codes up to weight 4, including for codes with odd check weight. These include the bivariate bicycle codes, which we show do not have the pre-orientation for either type of copy-cup gate. We show that abelian weight 4 group algebra codes satisfying the non-associative 3-copy-cup gate necessarily have a code distance of 2, whereas codes that satisfy conditions for the symmetric 3-copy-cup gate can have higher distances, and in fact also satisfy conditions for the 2-copy-cup gate. Finally we find examples of quantum codes from the product of abelian group algebra codes that have inter-code constant-depth $\operatorname{CZ}$ and $\operatorname{CCZ}$ gates.

Generalized master equations (GMEs) -- time-local but generally neither trace-preserving nor Hermiticity-preserving -- are convenient tools to compute properties of the environment of an open or continuously monitored quantum system. A two-sided master equation yields the fidelity and quantum Fisher information (QFI) of environment states, thereby setting fundamental limits for hypothesis testing and parameter estimation under continuous monitoring. For unmonitored noise or inefficient detection, the QFI of the detectable part of the environment may be obtained from a recently derived GME acting on multiple system replicas. Tilted master equations provide the full counting statistics of quantum jumps and diffusive measurements, enabling, e.g., studies of quantum thermodynamics beyond average values. Here we focus on bosonic linear systems, governed by a quadratic Hamiltonian and linear jump operators, whose dynamics preserves Gaussianity. For Gaussian initial states, we recast a generic GME as a compact set of ordinary differential equations for the covariance matrix (a Riccati-type equation), first moments, and normalization. These equations can be integrated efficiently without Hilbert-space truncation, and admit analytical results in simple settings. We also provide specialized forms for fidelity/QFI and full counting statistics. We illustrate the formalism with a continuously monitored optical parametric oscillator, using it to determine sensitivity limits for frequency estimation and to benchmark Hasegawa's thermodynamic uncertainty relations.

Quantum metrology promises measurement precision beyond classical limits by exploiting large-scale quantum states, yet realizing this advantage faces two fundamental challenges: the deterministic preparation of non-trivial quantum probes and the efficient extraction of metrological information in high-dimensional Hilbert spaces. Here, we introduce quantum confocal microscopy in Fock space that simultaneously resolves both challenges. Drawing a direct analogy between classical wave optics and quantum state evolution in a bosonic mode, we construct a confocal system with two Fock-space lenses. The first lens deterministically focuses a coherent state into a quantum probe with a tightly concentrated photon-number distribution, while the second lens maps the metrological information back to the vacuum state for efficient readout. Using a superconducting circuit QED platform, we prepare focused probe states with mean photon numbers up to ${N} = 500$, achieving a 21.5$\pm$1.1 dB compression of the photon-number uncertainty relative to a coherent state, with a scalable quantum circuit of $\mathcal{O}(1)$ operational depth. We demonstrate a displacement sensitivity scaling as $N^{-0.416}$, approaching the Heisenberg scaling ($N^{-0.5}$), and achieve a record metrological gain of 19.06$\pm$0.13 dB beyond the standard quantum limit. This work establishes quantum confocal microscopy as a scalable and practical framework for quantum-enhanced precision measurement, readily extendable to other bosonic platforms and high-dimensional quantum many-body systems.

We present an analytical treatment of Axion-Like-Particle (ALP)--photon mixing with extragalactic background light (EBL) attenuation for constant, Gaussian-stochastic, and non-Gaussian magnetic field configurations--with direct implications for Very High Energy (VHE) gamma-ray observations such as LHAASO, HAWC, and CTA experiments. For constant fields, we derive exact probabilities and identify a perturbative plateau regime where photon survival scales as quartic order of magnetic field, isolating the four-point magnetic correlation as a sensitive probe of non-Gaussianity. For Gaussian stochastic fields, we obtain--for the first time--analytical formulas for non-exponential-decay components in the strong-attenuation regime. Contrary to the widely used domain-like model, photon survival is suppressed by 4-6 orders of magnitude, while both conversion and survival probabilities exhibit distinct multi-peak structures from mass-equal resonance, stochastic resonance, and EBL attenuation. Extending to non-Gaussian fields, we show that non-Gaussianity can enhance photon survival by several orders of magnitude relative to the Gaussian case, potentially explaining the unexpectedly VHE photon event observed by LHAASO. Our results demonstrate that stochastic magnetic fields cannot be reduced to domain-like coherence without losing essential physics, and that VHE gamma-ray spectra encode observable information about both the power spectrum and non-Gaussian structure of intergalactic magnetic fields--critical as next-generation observatories push toward PeV sensitivities.

Photon sources based on spontaneous parametric down-conversion (SPDC) are central to experimental quantum optics and quantum technologies. Their performance is commonly quantified by three metrics: pair-collection probability, heralding efficiency, and spectral purity. In bulk-crystal SPDC, these metrics are known to be mutually constrained, yet the physical origin of the resulting trade-offs is often obscured. We show that these trade-offs originate from the frequency-dependent population of discrete spatial modes in the SPDC emission. By performing a Laguerre-Gauss mode decomposition at each frequency component, we show how spectral-spatial non-separability impacts collection probability, heralding efficiency, and purity. We apply this framework to two widely used quasi-phase-matching configurations: collinear degenerate type-0 and type-II SPDC in periodically poled bulk crystals, and quantify how different phase-matching functions shape the spectral-spatial mode structure. In particular, for type-II SPDC we compare standard periodically poled and aperiodically poled Gaussian phase matching. We experimentally validate some of our theoretical results using spatial- and spectral-projection measurements. This spectral-spatial mode analysis provides a quantitative and predictive framework for understanding and engineering bulk-crystal photon sources, enabling systematic multi-parameter optimization beyond qualitative design guidelines.

We investigate a non-Abelian SU$(2)$ quantum link model in 2+1 dimensions on a hexagonal lattice using tensor network methods. We determine the static quark potential for a wide range of bare coupling values and find that the theory is confining. We also probe the existence of a Luescher term and find a clear signal, however, the value of the dimensionless constant $γ$ strongly deviates from the expected universal value $-π/24$ for almost all values of the coupling $g^2$ we investigated. The width of the strings scales logarithmically with the string length again for all $g^2$-values, providing evidence for a rough string, with no indication for a roughening transition.

We derive a Lorentz-covariant spectral universality for stationary stochastic fields in Minkowski spacetime. We show that no covariant local mapping can relate temporal and spatial power spectra in more than one spatial dimension. For Lorentz homogeneous spectra, the temporal index is symmetry protected, observer invariant, and offset from the spatial index by a universal geometric factor set by effective momentum space dimensionality. We show how spectral universality breaks down for anisotropic scaling and dispersion dominated spectra, establishing the necessity of a Lorentz-covariant formulation of relativistic spectral inference.

We construct a probability distribution, induced by the Perron--Frobenius eigenvector of an exponentially large graph, which cannot be efficiently sampled by any classical algorithm, even when provided with the best-possible warm-start distribution. In the quantum setting, this problem can be viewed as preparing the ground state of a stoquastic Hamiltonian given a guiding state as input, and is known to be efficiently solvable on a quantum computer. Our result suggests that no efficient classical algorithm can solve a broad class of stoquastic ground-state problems. Our graph is constructed from a class of high-degree, high-girth spectral expanders to which self-similar trees are attached. This builds on and extends prior work of Gilyén, Hastings, and Vazirani [Quantum 2021, STOC 2021], which ruled out dequantization for a specific stoquastic adiabatic path algorithm. We strengthen their result by ruling out any classical algorithm for guided ground-state preparation.

Excited-state quantum phase transitions (ESQPTs) have been extensively studied in two-level models, but their characterization remains challenging in systems displaying mixed regular and chaotic dynamics. In this work, we investigate ESQPTs within the three-level Lipkin-Meshkov-Glick model, where an enlarged Hilbert space and multiple separatrices give rise to rich spectral structures strongly influenced by chaos. To investigate the different dynamical regions, we have calculated Poincaré sections and Peres lattices. In addition, by combining chaos-sensitive measures with standard ESQPT diagnostics, we provide a static analysis of ESQPT signatures in this model and establish a robust framework for future studies of its dynamical behavior. The degree of chaos and the Kullback-Leibler divergence are found to be very effective chaos-sensitive measures, which are complementary to ESQPT diagnostics such as the mean field limit and the participation ratio. Hence we provide a standard framework to work with ESQPTs in chaotic three-level systems.

We study perturbations of charged scalar and Dirac fields around Reissner-Nordström-de Sitter black holes with a global monopole. To this end, we first derive the equations of motion for both fields on the aforementioned background; these equations are then reformulated uniformly into the Teukolsky equation. Since the Teukolsky equation in asymptotically de Sitter spacetimes can be mapped into the Heun equation, we are able to solve quasinormal spectra by employing the Heun function method, not only for photon sphere modes but also for de Sitter and near-extremal modes. We analyze the spectra of all three types for both fields and, in particular, ascertain the effects of the global monopole. In the near-extremal regime, we find that the presence of a global monopole, on the one hand, leaves the strong cosmic censorship conjecture unaffected for scalar perturbations, while on the other hand, it enhance the violation of strong cosmic censorship for Dirac perturbations. Furthermore, we identify that the impact of the global monopole on both the spectra and the strong cosmic censorship is achieved by a shift in the modified multipole number. Our work demonstrates that the Heun function method is an efficient and robust approach for exploring the interaction between asymptotically de Sitter black holes and perturbing fields.

Bound states in the continuum (BiCs) convert dissipative open systems into effectively closed quantum subspaces through destructive interference. We show that two identical giant atoms coupled to a one-dimensional waveguide support BICs that coincide with maximally entangled atomic states. Most importantly, entanglement is predominantly determined by the geometric design; the ratio of intra-atomic connection lengths fixes the concurrence, while the propagation phase between atoms selects a family of Bell-like states. We further analyze the dynamical stability of these maximally entangled BICs under exact time evolution, revealing a clear hierarchy of robustness against parameter perturbations. Our results establish an analytical bridge between symmetry, geometry, entanglement, and BICs in giant-atom waveguide platforms.

The security of the proposed quantum Internet relies on repeater protocols designed under the assumption of stochastic, characterizable noise. We demonstrate that in adversarial environments this assumption induces performance vulnerabilities for computationally bounded repeater nodes. We show that the standard BBPSSW distillation protocol recursively purifies error syndromes rather than entanglement. This leads to a state of low fidelity despite diagnostic metrics indicating perfect convergence. Moreover, we show that the verifier cannot check the adversarial influence via the maximum likelihood estimation algorithm since it is blind to computationally bounded observers. To address these vulnerabilities, we propose a Cryptographic Network Stack centered on a trapdoor verification protocol. The protocol exploits private randomness to restore operational stability without requiring channel characterization.

We develop a general theory of spin-dependent density-density correlations, that is valid for any temperature, interactions, dimensions and mass or population status of Fermi gases with two internal states. We use gauge invariance and the Pauli principle to establish constraints on the spin-dependent density-density correlations that are consistent with the fluctuation-dissipation and Wick's theorem. As an example, we study the spin-dependent density-density correlations from the BCS to the Bose regime in two dimensions at zero temperature, inspired by experiments in 6Li. We show that two-particle irreducible contributions involving collective excitations, many-particle scattering and vertex corrections, are essential to describe experiments. In particular they turn out to be responsible for the emergence of an experimentally observed minimum in the opposite-spin density-density correlations.

We study the minimum cut problem in weighted undirected graphs using variational quantum algorithms in which only feasible cut configurations are explored. Although minimum cut admits efficient classical solutions, it is a fundamental component of more complex network optimization problems such as multicut and network interdiction. Our objective is to examine quantum models in which feasibility is preserved by the mixer dynamics, without introducing penalty terms in the cost Hamiltonian. We employ a ring structured XY mixer that restricts the quantum evolution to the subspace of valid cut configurations, ensuring that all sampled states correspond to feasible solutions. To address scalability limitations, we suggest an iterative metaheuristic strategy that decomposes large instances into smaller subproblems solved sequentially using the same quantum model. The results obtained using the mixer indicate that the initial probability distribution can be systematically controlled, thereby enabling the development of warm start techniques within variational quantum based algorithms.

We investigate the interplay of the Weak Gravity Conjecture (WGC) and the Weak Cosmic Censorship Conjecture (WCCC) in $F(R)$-Euler-Heisenberg black holes in Anti-de Sitter and de Sitter backgrounds. The solution is characterized by the electric charge $q$, the $F(R)$ deviation $f_{R_0}$, the Euler--Heisenberg coupling $λ$, and the constant scalar curvature $R_0$. We establish a universal entropy--extremality relation that provides thermodynamic evidence for the WGC independently of $f_{R_0}$ and $R_0$. Photon sphere analysis from both geodesic and topological perspectives confirms the simultaneous compatibility of the WGC and WCCC, with the Euler--Heisenberg coupling restoring photon spheres in the naked singularity regime. Gravitational lensing in the strong- and weak-deflection limits reveals that the photon sphere radius is independent of the cosmological background while the critical impact parameter nearly doubles in de Sitter. Black hole shadow images under isotropic accretion are constructed. Within the Barrow entropy framework, we uncover van der Waals-type phase transitions and analyze Joule-Thomson expansion, identifying the small black hole phase as the WGC-compatible thermodynamic regime accessible via isenthalpic cooling.

In this study, we examine quantum Oppenheimer-Snyder black holes (BHs) embedded within a cloud of strings and immersed in perfect fluid dark matter. Also, beginning with the underlying spacetime geometry, we determine how quantum corrections, string cloud contributions, and dark matter effects alter the geometrical structure and physical characteristics of the BH. Also, the optical behavior is investigated via a systematic analysis of the photon sphere and the associated BH shadow, emphasizing possible observational features capable of differentiating this configuration from classical models. We also analyze the motion of test particles, focusing on how surrounding matter components affect trajectories, stability conditions, and effective potentials. Scalar field perturbations are considered to investigate the BH response to external excitations and to extract information regarding its dynamical properties. In this case, the thermodynamic behavior of the system is studied, including the role of string clouds and dark matter in modifying BH thermodynamic quantities. Also, the obtained results present a unified description of the combined effects of quantum corrections, nonstandard matter sources, and BH physics, with potential relevance for both observational constraints and theoretical modeling of compact objects.

We introduce the concept of optical coherence squeezing in double-slit interference. We construct Hermitian operators that characterize the coherence at the slits, leading to coherence uncertainty relations and a corresponding squeezing condition. We also analyze states exhibiting such squeezing and show its manifestations in the uncertainty of the magnitudes and positions of the intensity fringes. Our work identifies coherence as a fundamental degree of freedom for squeezing, complementing phase, amplitude, and polarization, which could benefit quantum-enhanced interferometry.

Tailoring error correcting codes to the structure of the physical noise can reduce the overhead of fault-tolerant quantum computation. Hyperbolic Floquet codes use only weight-2 measurements and can be implemented directly on hardware with native pair measurements. We construct hyperbolic and semi-hyperbolic Floquet codes from $\{8,3\}$, $\{10,3\}$, and $\{12,3\}$ tessellations via the Wythoff kaleidoscopic construction with the Low-Index Normal Subgroups (LINS) algorithm. The $\{10,3\}$ and $\{12,3\}$ families are new to hyperbolic Floquet codes. We evaluate these codes under four noise models: phenomenological, ancilla Entangling Measurement (EM3), Single-step Depolarizing EM3 (SDEM3), and erasure. Under phenomenological noise, specific-logical threshold crossings occur near $p_e \approx 0.3$--$0.5\%$ for $\{8,3\}$ ($k=6$--$56$) and $0.15$--$0.2\%$ for $\{10,3\}$ ($k=12$--$146$). EM3 ancilla noise yields a threshold of ${\sim}1.5\%$ for all three families. SDEM3 is a depolarizing noise model motivated by Majorana tetron architectures; fine-grained codes achieve thresholds of ${\sim}1.0$--$1.2\%$ for all three families. The erasure model captures detected photon loss on spin-optical links; fine-grained codes achieve erasure thresholds of ${\sim}8.5$--$9\%$ for $\{8,3\}$, ${\sim}7$--$8\%$ for $\{10,3\}$, and ${\sim}6.5$--$8\%$ for $\{12,3\}$. Photon loss is the dominant error source in photon-mediated quantum computing. Under the full three-parameter SPOQC-2 noise model, the $\{8,3\}$ codes achieve a 2D fault-tolerant area $2.2\times$ that of the surface code compiled to pair measurements while encoding $k = 10$ logical qubits. In a companion paper, we evaluate the same code families in a distributed setting.

In this work, we theoretically investigate the deflection of light for strong- and weak-field regimes in the background of an electrically charged BH described in Kalb-Ramond gravity, which introduces the Lorentz symmetry violation parameter $l$, as well as the control of the degree of nonlinearity incorporated by electrodynamics through the parameter $γ$. We analytically constructed the expansion coefficients in both limits and used them as a basis to investigate gravitational lensing effects through observables, taking into account the variation of the parameters involved in the model, both for the canonical field and the phantom case.

The amount of extractable work from a physical system is fundamentally connected to the information available about its state, as illustrated by Maxwell's demon and the Gibbs paradox. In standard thermodynamic protocols involving system--bath interactions, the maximum work is given by the free-energy difference between the initial state and the corresponding Gibbs state at the bath temperature. This motivates a natural question: does information also limit work extraction in closed quantum systems that do not involve a heat bath and where work is obtained through unitary operations generated by a time-dependent Hamiltonian? While ergotropy quantifies the maximum work extractable via unitary operations, it assumes complete knowledge of the quantum state, typically requiring full state tomography. In realistic scenarios, however, only partial information is accessible. In this case, the relevant figure of merit is observational ergotropy, which depends explicitly on the measurement used to probe the system. We show that observational ergotropy decreases under classical post-processing of measurement outcomes, implying that fine-grained measurements allow greater work extraction than coarse-grained ones. Moreover, maximizing observational ergotropy over all possible measurements recovers standard ergotropy, which decomposes into incoherent (classical) and coherent (quantum) contributions. Our results demonstrate that coherence in the measurement projectors constitutes the key resource, enabling work extraction beyond the incoherent limit and establishing coherence as the origin of quantum advantage in observational ergotropy extraction.

Confinement and string breaking are two fundamental phenomena in gauge theories. Signatures of both are currently pursued in quantum-simulator experiments, opening a new angle on strongly interacting dynamics of gauge fields out of equilibrium, complementary to traditional particle-physics settings. In this work, we report the emergence of metastable confinement dynamics in a U(1) lattice gauge theory, originating from the competition between string tension and four-Fermi coupling - a competition that naturally arises in Rydberg atom arrays. We show that the initial string state can be resonantly melted through controlled energy matching, a phenomenon we identify as resonant string breaking. We demonstrate this mechanism for both static and Floquet-driven systems, where periodic modulation generates a spectrum of tunable sideband resonances. Our work provides new insights into the mechanisms of confinement and string breaking driven by long-range interactions and time-dependent fields, which are available in current quantum simulators on a variety of platforms.

Bias-tailored quantum error correcting codes (QECCs) offer a higher error threshold than standard QECCs and have the potential to achieve lower logical errors with less space overhead. The spin-cat qubit, encoded in a large nuclear spin-$F$ system, is a promising candidate for bias-tailored QECCs. Yet its feasibility is hindered by the difficulty of performing fast covariant SU(2) rotation with arbitrary rotation angles for nuclear spins and by a lack of noise characterization for gate operations in neutral atom platforms. Here we demonstrate single-qubit controls of ${}^{173}\mathrm{Yb}$ spin-cat qubits with nuclear spin $I=5/2$ in an optical tweezer array. We implement a covariant SU(2) rotation and non-linear rotations by optical beams and achieve an averaged single-Clifford gate fidelity of $0.961_{-5}^{+5}$. The measurement of the coherence time and spin relaxation time shows that the idling error becomes increasingly biased toward dephasing errors as the magnitude of the encoded sublevel $|m_F|$ increases. Furthermore, we benchmark the noise bias of rank-preserving gates on spin-cat qubits, demonstrating a finite bias of $18_{-11}^{+132}$, in contrast to the case of the two-level system in ${}^{171}\mathrm{Yb}$, which shows no bias within the experimental uncertainty. Our work demonstrates the feasibility of spin-cat qubits for realizing bias-tailored QECCs, paving the way for achieving hardware-efficient quantum error correction.

Variational quantum algorithms are a very promising tool for near-term quantum computing. However, despite their flexibility and wide applicability, their performance is fundamentally limited by Barren Plateaus (BP), where gradients vanish and optimization becomes intractable. Noise-Induced Barren Plateaus (NIBP) are particularly interesting, as they are predicted to arise due to noise accumulation independent of circuit structure. We experimentally study NIBP on IBM quantum hardware and demonstrate their absence under non-unital amplitude damping characterized by the qubit's $T_1$ coherence times. We use Information Content Landscape Analysis (ICLA) to efficiently estimate gradient norms for circuits ranging from 8 to 102 qubits, with hundreds of parameters and circuit runtimes of hundreds of microseconds. Classical simulations of the 8-qubit case under noiseless, depolarizing, amplitude damping, and dephasing noise models serve as a baseline comparison. We thoroughly analyze the experimental results considering calibration data, shot-noise, and circuit structure. We robustly observe that the gradient magnitude saturates beyond a characteristic circuit runtime, in contrast with the exponential decay expected from NIBP. Using recent theoretical results, we corroborate that under $T_1$-dominated noise NIBP do not occur and extract an effective $T_1^\text{eff}$ that is significantly shorter than suggested by standard calibration data. Our results experimentally confirm recent predictions on the absence of NIBP under non-unital noise. These findings also indicate that conventional benchmarking metrics based on average values for device characteristics may be insufficient to predict variational algorithm performance, but full distributions need to be considered.

We develop a modified interacting dark energy (IDE) model to study the redshift evolution of the Hubble constant ($H_0$), in light of the Hubble tension. In this framework, the energy exchange between dark energy and dark matter induces a redshift dependence of $H_0$. We evaluate the model against a comprehensive suite of observations, including baryon acoustic oscillations (BAO) from DESI DR2 and SDSS, cosmic chronometers, type Ia supernovae from the Pantheon sample, and Planck CMB distance priors. Analysis of late-Universe data yields $α= 0.0107^{+0.0032}_{-0.011}$, with the best-fit value on the order of $10^{-2}$, revealing a decreasing trend of $H_0$ with redshift. This supports a power-law evolution beyond $Λ$CDM. Incorporating CMB data further tightens the constraint to the order of $10^{-5}$, which we attribute to the suppression of dark-sector interactions at high redshifts, a consequence of the strong baryon--photon coupling. These results indicate that the IDE framework provides a theoretically consistent and observationally viable mechanism for describing the redshift evolution of $H_0$, offering a promising avenue toward alleviating the Hubble tension.

Sympathetic laser cooling is a key concept in precision spectroscopy and quantum state control of charged particles. Significant challenges arise in the metrologically relevant case where the effective interaction between the particles is weak and the particle to be cooled exhibits significant initial motional energy. Here we specifically address the most generally applicable case where the laser-cooled ion and the particle of interest are confined to two spatially separate potential wells with equal motional frequency for resonant enhancement of the cooling dynamics. We analyze the latter through numerical simulations and find that anharmonicities of the potential wells can prevent maintaining the resonance condition throughout the cooling process and thus inhibit a significant reduction in motional energy. We propose a cooling scheme that sweeps the trapping frequency of the potential wells. We show that this scheme enables efficient cooling from cryogenic temperatures all the way to the quantum regime of motion. As a specific application scenario, we analyze the sympathetic cooling of (anti-)protons into the quantum regime of motion for quantum-logic-spectroscopy-based tests of CPT invariance at the quantum limit in Penning traps. Nevertheless, our results and cooling strategies are generally applicable to other laser-inaccessible ion species.

Relativistic jets from black holes can extract energy not only from accretion but also directly from the black hole's spin, as described by the Blandford-Znajek mechanism. A longstanding question is whether magnetic flux can accumulate near the event horizon to such an extent that it halts accretion entirely, enabling energy extraction purely from spin. Previous studies have shown that accretion persists through instabilities and that jet power only modestly exceeds the accretion energy budget, yet some observational results suggest much higher efficiencies. Here we present state-of-the-art general relativistic magnetohydrodynamic (GRMHD) simulations that sustain a quasi-steady magnetically arrested disk state for approximately 10,000 dynamical times, during which accretion is globally suppressed across the full azimuthal extent. In this regime, jet power exceeds the accretion energy input by more than two orders of magnitude, demonstrating a previously unachieved level of efficiency. These results challenge conventional assumptions about the limits of black hole energy extraction and suggest a new framework for interpreting powerful jet systems. Our findings raise important questions about the long-term stability of such states and the fundamental limits of the Blandford-Znajek process.

Preserving multipartite entanglement in open many-body quantum systems is fundamentally limited by unavoidable environmental noise. We study the open-system dynamics of multipartite entanglement in an anisotropic XXZ spin chain interacting with a thermal spin bath, focusing on two states with distinct types of multipartite entanglement: the generalized GHZ and the generalized W state. Using a master-equation approach combined with the Bethe ansatz technique, we show analytically that robustness of multipartite entanglement at low temperatures can be enhanced by suitably tuning the anisotropy of the system. Our results highlight interaction-induced spectral control as a mechanism for stabilizing multipartite entanglement in quantum computing platforms.

The BASE collaboration aims to measure antimatter systems with the highest precision in order to perform a rigorous test of CPT symmetry and search for physics beyond the Standard Model. As part of the BASE collaboration, we pursue the development of quantum logic inspired cooling and detection techniques for g-factor measurements of (anti-)protons. Implementing these methods requires full quantum-level control of individual antimatter particles confined in cryogenic Penning traps. By mapping the (anti-)proton's internal state onto a co-trapped 9Be+ "logic" ion via free Coulomb coupling in a double-well potential, we can accelerate measurement cycles and push g-factor precision measurements on (anti-)protons toward the quantum limit. Here, we present an overview of the proposed method and the current status of the project, with special emphasis on the new cryogenic multi-Penning-trap stack and the proton detection system.

Bohr's principle of complementarity, prohibiting simultaneous access to certain physical properties within a single experimental arrangement, is considered to be a defining feature of quantum mechanics. It is commonly viewed as inducing an intrinsic hierarchy among incompatible observables: some sets of quantum properties are fundamentally more incompatible than others, as quantified by the maximal sharpness permitting their joint measurement. We show that this hierarchy ceases to be absolute in the multi-copy regime. Analyzing qubit spin observables, we prove a No-Comparison Theorem establishing that no global ordering of incompatible observable sets is preserved across all finite-copy configurations. In particular, two sets of observables can exhibit reversed complementarity ordering depending solely on whether the available resources are arranged as identical copies or as parallel-antiparallel pairs. Thus, the degree of quantum incompatibility is not an intrinsic property of observables alone but depends on the global configuration of the prepared quantum probes. Our results uncover a configuration-dependent structure of complementarity, reveal a subtle role of entanglement in shaping the structure of measurement limitations, and call for a reassessment of quantum information protocols under finite resources.

We investigate the steady state of two single-mode uniform spatially separated polaritonic conden- sates exposed to resonant pumping with entangled pairs of photons. We demonstrate the principal possibility of driving the system to an entangled state despite its exposure to noises arising from the excitonic reservoir and photon leakage through the microcavity mirrors. Estimates are provided for the flux of entangled particles required to drive the system into a steady state that violates the partial-transpose criterion for entanglement. Furthermore, we trace the evolution of the system after a sudden disappearance of the entangled pumping. Our analysis provides estimates for the entanglement lifetime in a system of two exciton-polariton condensates

The discovery of new quantum error-correcting codes that encode several logical qubits into relatively few physical qubits motivates the development of efficient and accurate methods of decoding these systems. Here, we adopt the minimum-weight perfect matching algorithm, a subroutine invaluable to decoding topological codes, to decode bivariate bicycle codes. Using the equivalence of bivariate bicycle codes to copies of the toric code, we propose a method we call the 'cylinder trick' to rapidly find a correction using matching on code symmetries. We benchmark our decoder on the gross code family, cyclic hypergraph-product codes, generalized toric codes, and recently proposed directional codes, demonstrating the general applicability of our protocol. For a subset of these codes, we find that our decoder can be significantly improved by augmenting matching with strategies including belief propagation and 'over-matching', thus achieving performance competitive with state-of-the-art approaches.

We derive the geodesic equation for point particles propagating in Moyal-type noncommutative spacetimes using a field-theoretic approach based on the quasi-classical limit of the noncommutative Klein-Gordon equation. Starting from a twisted-geometric construction of the covariant Laplace-Beltrami operator, we obtain the noncommutative Hamilton-Jacobi equation and show that all noncommutative effects are absorbed into an effective, position-dependent mass function $M(x)$ appearing in an otherwise standard relativistic dispersion relation. The corresponding particle dynamics then acquires an additional term in the geodesic equation that takes the form of a fixed external force $F_{\text{NC}}^μ= -\frac{1}{2} g^{μν}\partial_νM^2(x)$, sourced entirely by the quantum nature of spacetime. We compute this effective mass perturbatively up to fourth order in the noncommutativity parameter for a general metric, proving that all odd-order corrections vanish identically. For the specific case of an $(r-θ)$ twist applied to spherically symmetric backgrounds, we obtain explicit expressions demonstrating that the leading correction to geodesic motion appears at $Θ^2$ order and is proportional to the probe particle's mass, while massless particles remain unaffected.

A classic feature of gravity is that it is an attractive force. If a source mass is prepared in a localized (classical- like) state, it will cause another probe mass to move towards it. Here we consider the situation in which a source mass is prepared in a quantum superposition of distinct spatial states while a probe mass interacts with it. Conditional on the detection of the source mass in a specific state, the probe mass will be found to move away from the source mass (repulsion). This signifies the quantum superposition of gravitational forces acting on the probe mass and thereby the fact that spacetime can exist in quantum superpositions. The technique used is the repulsive effect arising from an anomalous negative weak value. A potential experimental implementation with spin bearing nanocrystals is outlined.

We show that noiseless random quantum circuits pass the linear cross-entropy benchmark (LXEB) test with high probability. If the circuits are linear depth, and thus form unitary 4-designs, the LXEB test is passed with probability $1-O(1/\sqrt{k})$, where $k$ is the number of independently drawn samples from the output distribution of the random circuit. If the circuits are of depth $\tilde O(n^2)$, and thus form unitary $n$-designs, the LXEB test is passed with probability $1-O(e^{-k \log(n)/n})$. In proving our results, we show strong concentration of the random circuit collision probability at linear depth and establish that the tails of the distribution of random circuit output probabilities start to resemble Porter-Thomas at near-quadratic depths. Our analysis employs higher moments and high-degree approximate designs.

An experimental implementation of the Coherent One-Way Quantum Key Distribution (COW-QKD) protocol is reported under realistic conditions, and a clean and easy-to-use framework for performing finite key analysis of the COW-QKD protocol is provided by extending a set of existing results. The framework provided here is used to perform finite key rate analysis of the COW-QKD protocol with respect to the actual parameters used in the experimental realization reported here. The system is kept running for several hours with different experimental parameters and stable secure key rates between 1.2 to 1.6 kbps are observed. In addition, QBER, phase error rate and secure key rate are obtained under finite key analysis, and it is shown that COW-QKD is secure for medium-range transmissions (up-to ~ 156 (171) km of optical fiber with 0.2 dB loss per km if detector efficiency is 0.1 (0.2) and other parameters are same as those used in this experiment).

We study the spectra of collective low excitations of two atomic ion ensembles which are confined in a liner trap by addressing lases. When the left ensemble is driven by an external optical field, its corresponding response spectrum to the incident optical light shows a vibration-induced transparency phenomenon when the detuning of the laser addressing the ion is tuned to the first red sideband. In the case of the detuning tuned to the first blue sideband, the response spectrum shows a conversion from the absorption peak to the transparency window. Furthermore, we investigate the fluctuation spectra of the collective excitation modes of ion ensemble and show the similar phenomena.

Estimating thermal expectations of local observables is a natural target for quantum advantage. We give a simple classical algorithm that approximates thermal expectations, and we show it has quasi-polynomial cost $n^{O(\log n/ε)}$ for all temperatures above a phase transition in the free energy. For many natural models, this coincides with the entire fast-mixing, quantumly easy phase. Our results apply to the Sachdev-Ye-Kitaev (SYK) model at any constant temperature -- including when the thermal state is highly entangled and satisfies polynomial quantum circuit lower bounds, a sign problem, and nontrivial instance-to-instance fluctuations. Our analysis of the SYK model relies on the replica trick to control the complex zeros of the partition function.

Stupendously large astrophysical black holes (SLABs) are hypothetical black holes with masses of more than a trillion Suns. Because observable consequences of their existence have only recently been seriously considered, there have been relatively few constraints on their abundance. This work motivates a simple yet powerful constraint on SLABs: their huge shadows are visible against the cosmic microwave background. SLABs could thus appear as negative sources in microwave data. In fact, the shadow of a SLAB with a fixed mass becomes easier to detect with increasing redshifts past $1.6$. The limits are powerful enough to rule out SLABs of mass $\gtrsim 10^{17}\ M_{\odot}$ within the last scattering surface, and imply $Ω_{BH} \lesssim 10^{-5}$ for masses $10^{15}$--$10^{18}\ M_{\odot}$. I also discuss the effects of accretion and their implications for the limits: SLAB growth, positive accretion luminosity, and obscuring material.

A quantum amplifier usually adds extra noise inevitably through coupling to internal degrees of freedom while amplifying the signal. The introduction of quantum correlations can effectively suppress this extra noise. In this work, we utilize the established quantum correlation between the Stokes field and atomic spin waves in the Raman amplification process to feedback a portion of the Stokes field into the amplifier. This leads to a reduction in quantum noise that is independent of the feedback loss at high gain. A maximum of 6 dB noise reduction is observed. The single-path feedback amplifier is found to be sensitive to the feedback phase, a property that expands its potential for applications in quantum precision measurement, and the general concept can be extended to integrated optics and fiber optic systems.

We study Lorentzian vacuum transition probabilities between two minima of a scalar field potential within the framework of $f(R)$ gravity. The analysis extends the previously considered WKB expansion of the Wheeler-DeWitt equation to modified gravity theories, up to second order. We apply the general method for homogeneous and isotropic FLRW universes, with zero and positive spatial curvature, for any $f(R)$ model. For the flat case we obtain analytic expressions for the transition probabilities for any model if we assume a constant Ricci scalar; this assumption has been considered in previous studies, in the Euclidean approach, from symmetry arguments. On the other hand, we also obtain explicit solutions without this assumption for power-law $f(R)=R^{1+n}$ models. Moreover, in the positive curvature scenario, we obtain that the assumption of a constant Ricci scalar is not consistent, but we are able to find analytical solutions in approximated regimes. In all cases we have found that the general behavior of the probabilities already found for Einstein Gravity is preserved, including the prediction of a non-singular initial state due to quantum corrections, even though the probabilities increase or decrease in a model dependent way.

In this letter, we develop a unified semiclassical framework for the thermodynamics and quantization of the Reissner--Nordström (RN) black hole (BH) based on the Misner--Sharp--Hernandez (MSH) mass. Treating the quasi-local horizon energies as the relevant thermodynamic variables, we formulate a horizon-by-horizon first law and Smarr relation. Using a reduced phase-space quantization, we obtain a discrete MSH mass spectrum for both horizons, which reproduces the minimal entropy spacing. Quantum transitions between adjacent levels yield Planck-scale corrections to the Hawking temperatures and a universal logarithmic contribution to the entropy, consistent with independent approaches to quantum gravity. We encode these corrections into a quantum-deformed RN geometry via a simple multiplicative factor that preserves the classical horizon positions while reproducing the corrected surface gravities. The associated effective stress tensor behaves as a conserved vacuum-polarization source with characteristic $r^{-4}$ falloff and a small trace, providing a compact representation of semiclassical backreaction. The deformation slightly lowers both horizon temperatures, weakens the inner-horizon instability, and induces tiny shifts in photon-sphere and shadow observables for macroscopic BHs.

Large, 3D trapped ion crystals offer improved sensitivity in quantum sensing protocols, and are expected to be implemented as platforms in near-future experiments. However, numerical techniques used to study the laser cooling of such crystals are inefficient as the number of ions, $N$, in the crystal increases. Here we develop a powerful numerical framework to simulate laser cooling of up to $10^5$ ions stored in a Penning trap. We apply this framework to characterize and optimize the cooling of ellipsoidal 3D crystals. We document new pathways to enhanced cooling based on the addition of an axial component to the potential energy-dominated $\boldsymbol{E}\times\boldsymbol{B}$ modes. Furthermore, we observe greatly enhanced cooling of the perpendicular kinetic energy to below 1 mK in prolate ion crystals, enabling a simplified cooling beam setup for such crystals. We propose specific values of trap and laser beam parameters which lead to optimal cooling in a variety of examples. This work illustrates the feasibility of preparing large 3D crystals for high-sensitivity quantum science protocols, motivating their use in future experiments.

In this work we study the Unruh temperature as arising from tunneling through a barrier for an observer in flat Minkowski spacetime with arbitrary acceleration $a(t)$. For the defining case of constant acceleration $a(t) = a_0$, the Unruh temperature (W. Unruh, Phys. Rev. D 14, 870 (1976)) is given by $k_b\,T_U =\tfrac{\hbar\,a_0}{2\,π\,c}$. Extending the work of de Gill et al. (A. de Gill, D. Singleton, V. Akhmedova and T. Pilling, Am. J. Phys. 78, 685 (2010)) we generalize the gravitational WKB approach to derive the Unruh temperature for arbitrary acceleration. We show that the often employed Schwarzschild-like form of the flat metric is not appropriate for the WKB calculation with an arbitrary $a(t)$, and instead derive a generalized Unruh temperature for the generalized Rindler metric where $a_0\to a(t)$. We derive a generalization of the Rindler coordinates appropriate for arbitrary $a(t)$, and stress the importance of the role of the integrated acceleration $χ(t) = \int^t dt'\,a(t')$, which can also act as a temporal coordinate. We explore several non-trivial examples of $a(t)$ and their generalized Unruh temperatures. We additionally develop an approximation to the Unruh temperature for small deviations away from constant acceleration by the standard approach of considering the negative frequency content of a purely positive frequency plane wave of an inertial observer, as measured by the co-moving arbitrarily accelerated observer. Lastly, we develop and explicit coordinate transformation between the arbitrarily accelerated observer and conformal coordinates, where the plane wave structure of the solutions of the wave equation is readily transparent, and analogous to the form for the inertial observer.

The fact that gravitational environments cannot be shielded (since gravity is universal) makes them of great theoretical interest to decoherence mechanisms and to the quantum-to-classical transition. While past results seemed to indicate that graviton-induced decoherence of spatial superpositions happens only for macroscopic systems, recently it was shown that this mechanism can be enhanced through the system's own dynamical internal structure. In this work, we extend this analysis by including the interaction with a classical Newtonian potential. We show that, although the graviton bath alone dominates the mechanism for short times compared to a timescale established by the size of the quantum spatial superposition, the interplay between the gravitons and the internal degrees of freedom of the system renders decoherence inevitable in the long-time limit, even for microscopic masses. We also show that this mechanism is slightly slowed down by the interplay with the classical Newtonian potential, which, for systems without dynamical internal degrees of freedom, can even lead to recoherence, at least in principle.

In arXiv:2503.00188 we introduced broadband pulsed (BBP) homodyne measurements as a generalization of standard pulsed homodyne quadrature measurements. BBP can take advantage of detectors such as calorimeters that have the potential for high efficiency over a broad spectral range. BBP homodyne retains the advantages of standard pulsed homodyne, enabling measurement of arbitrary quadratures in the limit of large amplitude local oscillators (LO). Here we quantify the convergence of standard and BBP homodyne quadrature measurements to those of the quadrature of interest. We obtain lower bounds on the fidelity of the post-measurement classical-quantum state of outcomes and unmeasured modes, and the fidelity of the states obtained after applying operations conditional on measurement outcomes. The bounds depend on the LO amplitude and the moments of number operators. We demonstrate the practical relevance of these bounds by evaluating them for standard pulsed homodyne used for estimating values of the characteristic function of the Wigner distribution, expectations of moments, for quantum teleportation and for continuous variable error correction with GKP codes.

In this article, we extend the numerical studies developed in [arXiv:2210.12898] to construct periodic stationary axisymmetric solutions containing multiple horizons in each fundamental domain. As a direct application, we consider periodic stationary axisymmetric solutions with two identical equidistant counter-rotating horizons. These solutions can be parametrized by the period $L$, the mass $M$, and the absolute value of the angular momentum $|J| >0$. We provide strong numerical evidence for the existence of such configurations, without any restriction in terms of the distance between horizons. This is in sharp contrast with the non-zero total angular momentum case, as it was recently established in \cite{Peraza:2024uto} that static single-horizon periodic solutions cannot be put into rotation if $L < 4M$. It is shown that these solutions do not have any struts on the axis, and it is also explicitly shown that, by taking non-equidistant horizons, struts develop between the black holes. Other global properties of the solutions are also presented.

Stabilization of laser frequencies is critical for sensitive Rydberg measurements, including in applications such as dissipative time crystal (DTC) dynamics, yet conventional approaches often require complex or costly hardware. We demonstrate a compact, low cost stabilization method using a scanning Fabry Perot interferometer (SFPI) to transfer lock a 960nm coupler laser to an 852nm probe. The lock suppresses coupler multi MHz free running drift and improves the Allan deviation by up to an order of magnitude, reaching <75kHz at 66s. Applied to DTC oscillations using a Rb 2 photon D2 transition, the second harmonic generated 480nm (from 960nm lock) reduces DTC frequency drift from >20kHz to a few kHz and lowers instability by more than an order of magnitude with a minimum Allan deviation of 0.2kHz at <10s. These results establish SFPI-based transfer locking as a practical and accurate approach for scalable multi laser Rydberg experiments requiring long-term stability in a compact and low cost system.

A protocol for identifying controlled-NOT (CNOT) gates versus single-qubit-only gates in universal quantum circuits using randomized input states was recently shown to be intimately connected to the quantum resource of state texture. Here we revisit this gate identification protocol and demonstrate that a more general fidelity-based formulation succeeds for nearly all laboratory bases. We then examine a broader family of quantum resource theories, where a distinct resource theory can be defined for each choice of reference pure state, establishing core resource-theoretic requirements without the computational shortcut offered by the "grand sum" employed in the original formulation of state texture. By extending from single "resourceless" states to convex sets via a convex-roof construction, we recover single-qubit measures of known resource theories such as imaginarity and coherence. Finally, we introduce a family of "fixed-point resource theories" that includes fixed-point instances of the theories of state texture, genuine coherence, purity, and athermality. For these fixed-point resource theories we show that, under free operations, the fidelity-based lower bound is weakly monotonic, while specific violations of strong monotonicity are found for the convex-roof logarithmic measure.

Long-range anisotropic dipole-dipole interactions between ultracold polar molecules are predicted to drive exotic quantum phases, yet direct many-body signatures of these interactions in degenerate Fermi gases have remained elusive. Here, we report the observation of an interaction-induced controlled deformation of the Fermi surface, providing a clear many-body signature in a deeply degenerate Fermi gas of $^{23}\text{Na}^{40}\text{K}$ molecules. Using double microwave (MW) shielding, we prepare $8 \times 10^3$ molecules at $0.23(1)$ times the Fermi temperature, achieving a three-fold suppression of inelastic losses compared to single MW shielding while preserving strong elastic dipolar scattering. We observe Fermi surface deformations of up to $7\,\%$, more than two times larger than those observed in magnetic atoms, despite operating at two orders of magnitude lower densities. Crucially, we demonstrate continuous tuning of the interaction potential from axial U(1) to biaxial C$_{2}$ symmetry, directly imprinting this geometry onto the Fermi surface. We find excellent agreement between our experimental results and parameter-free Hartree-Fock theory. These results establish MW-shielded polar molecules as a highly tunable platform for exploring strongly correlated dipolar Fermi matter and offer a promising path towards topological superfluidity.

We present a first-principles derivation of a non-equilibrium quantum master equation for a continuously driven open quantum system interacting with a structured electromagnetic environment. Starting from a driven Caldeira-Leggett model in which an external classical field couples simultaneously to the system and reservoir degrees of freedom, we proceed without assuming that the standard equilibrium fluctuation-dissipation theorem holds. The bath statistics acquire explicit dependence on the two-time autocorrelation function of the applied field, leading to drive-biased noise correlations and intrinsically non-Markovian dynamics. By eliminating the reservoir exactly at the operator level, we obtain a driven quantum generalized Langevin equation whose noise and dissipation kernels depend on two independent times. Exploiting the Gaussian nature of the driven bath, we derive a modified Hu-Paz-Zhang master equation in which the diffusion coefficients and coherent forces inherit explicit memory of the external field. We demonstrate that the physically observable oscillation frequency remains encoded in the homogeneous Green's function of the Langevin equation, while the drive-induced corrections manifest exclusively through modified diffusion and drift terms. Our results provide a unified microscopic framework for understanding field-biased fluctuation relations with direct relevance to cavity and circuit quantum electrodynamics experiments operating far from equilibrium.

We compute the entanglement entropy of a real massive scalar field near a non-rotating BTZ black hole using Thermo Field Dynamics. Modeling the black hole as a collapsing dust shell in AdS3, we derive the shell trajectory R(t) as seen by a Fiducial Observer (FIDO). From the Hartle-Hawking and Killing-Boulware vacua, we obtain the Wightman function difference and compute energy density, revealing a sharply localized energy density just outside the horizon, consistent with the brick wall picture. A full thermodynamic analysis yields an entanglement entropy proportional to the horizon area, numerically matching the Bekenstein-Hawking entropy. All intermediate steps, including junction conditions, Kruskal extension, WKB modes, and UV regularization, are explicitly detailed.

Absolutely stabilizer states are those that remain convex mixtures of stabilizer states after conjugation by any unitary. Here we give a characterization of such states for multiple qudits of all prime dimensions by introducing a polytope of their allowed spectra. We illustrate this through the examples of one qubit, two qubits, and one qutrit. In particular, the set of absolutely stabilizer states for a single qubit is a ball inscribed in the stabilizer octahedron, but for higher dimensions the geometry is more complicated. For odd-prime-dimensional qudits, we also give a complete characterization of absolutely Wigner-positive states, i.e., states whose Wigner function remains nonnegative after conjugation by any unitary. In so doing, we show there are absolutely Wigner-positive states that are not absolutely stabilizer, which can be seen as a unitarily-invariant version of bound magic. We then study the radii of the largest balls contained in the sets of absolutely stabilizer states and absolutely Wigner-positive states. These radii respectively tell us the lowest possible purity of nonstabilizer and Wigner-negative states. Conversely, we also find the radius of the smallest ball containing the set of absolutely Wigner-positive states, giving a tight purity-based necessary condition thereof.

Identifying the boundary between classical and quantum computation is a central challenge in quantum information. In multi-qubit systems, entanglement and magic are the key resources underlying genuinely quantum behaviour. While entanglement is well understood, magic -- essential for universal quantum computation -- remains relatively poorly characterised. Here we show that determining membership in the stabilizer polytope, which defines the free states of magic-state resource theory, requires super-exponential time $\exp( n^2)$ in the number of qubits $n$, even approximately. We reduce the problem to solving a $3$-SAT instance on $n^2$ variables and, by invoking the exponential time hypothesis, the result follows. As a consequence, both quantifying and certifying magic are fundamentally intractable: any magic monotone for general states must be super-exponentially hard to compute, and deciding whether an operator is a valid magic witness is equally difficult. As a corollary, we establish the robustness of magic as computationally optimal among monotones. This barrier extends even to classically simulable regimes: deciding whether a state lies in the convex hull of states generated by a logarithmic number of non-Clifford gates is also super-exponentially hard. Together, these results reveal intrinsic computational limits on assessing classical simulability, distilling pathological magic states, and ultimately probing and exploiting magic as a quantum resource.

We introduce a family of many-body systems of distinguishable continuous-variable particles in which interparticle interactions are set by the adjacency matrix of a graph. The ground-state wavefunction of such systems is of a generalized Jastrow form involving the product of pair-correlation functions over the edge set of the graph. These systems describe quantum fluids when the graph is complete, and the pair function has a well-defined permutation symmetry. In general, they provide the continuous-variable generalization of spin systems on graphs, with broken permutation symmetry. The corresponding parent Hamiltonian is shown to include (a) two-body interactions determined by the graph adjacency matrix and (b) three-body interactions over all possible 2-paths on the graph. Employing elements of graph theory, we chart the landscape of models, recovering known instances in the literature and providing numerous new examples of ground-state solvable models for which the system Hamiltonian, ground-state wavefunction, and corresponding energy eigenvalue are specified.

We study the spacelike Kasner singularity of spherically-symmetric, static and asymptotically flat black holes in Einstein gravity minimally coupled to a massless scalar with a suitable self-interacting scalar potential. We focus on how the asymptotic information such as the mass and scalar charge affect the properties of the Kasner singularity, including the Kasner exponents. We show how a nontrivial integration constant can be extracted from the near-singularity geometry and find a general pattern that this integration constant asymptotes to a linear combination of the mass and scalar charge at large mass limit. We also find that there may be a black hole upper bound on the maximum surviving time of a massive particle inside such a black hole before it falls into the Kasner singularity, and the Schwarzschild black hole saturate this bound.

The study of fermionic quantum field theories is an important problem for realizing the standard model of particle physics on a quantum computer. As a step towards this goal, we consider the massive Thirring and Gross--Neveu models with arbitrary number of fermion flavors, $N_f$, discretized on a spatial one-dimensional lattice of size $L$ in the Hamiltonian formulation. We compute the gate complexity using the higher-order product formula and using block-encoding/qubitization and quantum singular value transformations in the limit of large $N_f$ and $L$. We also prepare the ground states of both models with excellent fidelity for system sizes up to 20 qubits with $N_f = 1,2,3,4$ using the adaptive-variational quantum imaginary time algorithm. In addition, we also classify the dynamical Lie algebras of these relativistic fermionic models and show that they belong to the same isomorphism class. Our work is a concrete step towards the quantum simulation of real-time dynamics of large $N_f$ fermionic quantum field theories models relevant for chiral symmetry breaking, understanding dimensional transmutation, and exploring the conformal window of field theories on near-term and early fault-tolerant quantum computers.

Quantum state transfer is the primitive of transporting an unknown state on one site of a lattice to another. Using power-law interactions, recent state transfer protocols achieve speedup by utilizing the intermediate ancilla sites. However, these protocols require the ancillas to be in a perfectly initialized state, which, due to noise or imperfect control, may not be the case. In this work we introduce the $\textit{robustness}$ of a state transfer protocol, which quantifies the protocol's tolerance to error in the initial ancilla state. In the Heisenberg picture, state transfer grows operators supported on the final site such that they no longer commute with all operators on the starting site. We prove that this robustness tightly bounds the Schatten $p$-norms of these commutators between initial and final-site operators. This generalizes the known cases of $p=\infty$ and $p=2$, which govern completely state-dependent and state-independent state transfer respectively, demonstrating that intermediate values of $p$ govern partially state-dependent state transfer. In conjunction with existing power-law light cones, our result gives new minimum runtimes for partially state-dependent protocols which, in certain regimes, are parametrically better than existing bounds. We introduce new robust state transfer protocols, charting the landscape between complete state-dependence and state-independence.

Single-shot fidelity is the standard benchmark for superconducting qubit readout, but it does not directly minimize the total wall-clock time required to certify a quantum state. We develop an information-theoretic description of dispersive readout by treating the measurement record as a stochastic communication channel. Within a trajectory model that incorporates T1 relaxation with full cavity memory, we compute the classical Chernoff information governing the multi-shot error exponent. We find a consistent separation between the integration time that maximizes single-shot fidelity and the time that minimizes total certification time. For representative transmon parameters and hardware overheads, the throughput-optimal integration window is longer than the fidelity-optimal one, yielding certification speedups of approximately 9 to 11 percent, with the gain saturating near 1.13x in the high-readout-power and high-overhead regime. Comparing the extracted classical information to the unit-efficiency Gaussian Chernoff benchmark defines an information-extraction efficiency metric. Typical dispersive schemes are limited to about 45 percent capture at short integration times by detection efficiency, decreasing to approximately 12 percent at a throughput-optimal integration time of about 1.22 microseconds due to T1-induced trajectory smearing. This formulation connects readout calibration to the operational objective of minimizing certification time in high-throughput superconducting processors.

The Dymnikova black hole represents a regular spacetime solution interpolating between a de Sitter core and an asymptotically Schwarzschild geometry. In this work, we investigate a generalized Dymnikova black hole surrounded by perfect fluid dark matter (PFDM) and immersed in a cloud of strings (CS). We analyze how these additional matter sources modify the thermodynamic, optical, and dynamical properties of the spacetime. We derive the Hawking temperature and specific heat capacity and examine the thermal stability and phase structure of the black hole. The results reveal non-monotonic temperature behavior and parameter-dependent phase transitions. We further study photon dynamics, including the photon sphere and black hole shadow, and show that both PFDM and string cloud parameters significantly affect the shadow radius and strong-field structure. Additionally, we investigate the motion of massive test particles, circular orbits, and stability conditions. The corresponding effective potentials, specific energy, and angular momentum are analyzed. Finally, we explore quasi-periodic oscillations (QPOs) by computing the fundamental epicyclic frequencies and discuss how the model parameters encode observable astrophysical signatures.

It is shown that conformal symmetry exists in force-free electrodynamics (FFE) in Minkowski spacetime, a foundational framework for describing magnetospheres around astronomical objects. In force-free magnetospheres, charges are constrained to move along magnetic field lines and experience zero Lorentz force, due to the everywhere perpendicular orientation of electric and magnetic fields. However, a general angle-preserving conformal mapping of force-free fields does not necessarily produce another physically admissible force-free configuration when sources are present. In this work, we demonstrate that such invariance can nevertheless arise for certain choices of the free functions. Specifically, the governing stream equation is shown to be invariant under Möbius transformations. This symmetry reveals a structural linkage between known solutions and, notably, maps the region inside a magnetospheric horizon (the lightsurface) of one solution to the exterior of its dual counterpart, and vice versa.

In the early 1970s, Jacob Bekenstein discovered that black holes have entropy, which became one of the greatest scientific revolutions of the second half of the 20th century. The objective of this paper is to present a simple derivation -- partly heuristic and partly geometric -- of the equation for the entropy of a black hole, which we now know as the Bekenstein-Hawking entropy. We will also briefly explore the physical implications of this equation and its relationship to the work of Stephen Hawking.

The dark matter (DM) density profile in the innermost region of the Galaxy remains an open question. In particular, while adiabatic growth of the supermassive black hole Sgr A$^\ast$ at the Galactic Center (GC) can induce a 'spike' in central DM density, the existence of such a spike is still under debate. Here we present new constraints on the spike slope $γ_{\rm sp}$ using conventional DM indirect detection searches. We first recast existing photon and neutrino line searches, which include the contribution from the GC region, into constraints on the thermally-averaged DM annihilation cross section $\langleσv\rangle$ in the presence of a DM spike. We then derive new bounds on the spike profile for a generic Weakly Interacting Massive Particle (WIMP) DM scenario, where the thermal freeze-out mechanism fixes the annihilation cross-section at $\langleσv\rangle\sim (2-3) \times 10^{-26}~{\rm cm}^3~{\rm s}^{-1}$. We find that for DM annihilation to photons, constraints from Fermi-LAT and MAGIC rule out spike profiles at the GC for a broad range of WIMP DM masses from 10 GeV to 100 TeV. Our result holds even if the photon channel constitutes only $1\%$ of the total annihilation rate. For the neutrino channel, we use the IceCube data to constrain the existence of an extremely steep spike in the $\mathscr{O}(1-10)$ TeV DM mass range. Our analysis can be easily extended to other annihilation channels.

Machine learning algorithms are being used more frequently in the first-level triggers in collider experiments, with Graph Neural Networks pushing the hardware requirements of FPGA-based triggers beyond the current state of the art. To meet the stringent demands of high-throughput and low-latency environments, we propose a concept for latency-optimized preprocessing of sparse sensor data, enabling efficient GNN hardware acceleration by removing dynamic input sparsity. Our approach rearranges data coming from a large number of First-In-First-Out interfaces, typically sensor frontends, to a smaller number of FIFO interfaces connected to a machine learning hardware accelerator. In order to achieve high throughput while minimizing the hardware utilization, we developed a hierarchical sparsity compression pipeline optimized for FPGAs. We implemented our concept in the Chisel design language as an open-source hardware generator. For demonstration, we implemented one configuration of our module as preprocessing stage in a GNN-based first-level trigger for the Electromagnetic Calorimeter inside the Belle II detector. Additionally we evaluate latency, throughput, resource utilization, and scalability for a wide range of parameters, to enable broader use for other large scale scientific experiments.

Primordial magnetic fields (PMFs) generated in the early Universe may leave observable imprints in the present-day large-scale structure. However, it remains unclear on which spatial scales primordial signatures can survive the nonlinear processes accompanying structure formation. The aim of this study is to investigate the evolution of PMFs during gravitational collapse and to determine the spatial scales on which primordial signatures can persist. We perform a suite of high-resolution direct numerical simulations of self-gravitating, magnetized halos. By varying the viscosity, we probe different Reynolds-number regimes and follow the coupled evolution of gravitational collapse and magnetohydrodynamic turbulence. At sufficiently high Reynolds numbers, turbulence generated during collapse triggers the onset of a small-scale dynamo, which amplifies magnetic energy below the Jeans scale and modifies the magnetic energy spectrum significantly. Whether dynamo amplification dominates the magnetic field evolution is determined by the competition between the dynamo growth time and the free-fall time. Our results highlight the importance of resolving the Jeans scale and the associated turbulent inertial range in cosmological MHD simulations to accurately capture the interplay between gravitational compression and dynamo amplification and to assess which structures retain memory of primordial fields.

We state a conjecture about the zeta function of crepant resolutions of Berglund--Hübsch orbifold hypersurfaces over a finite field. In addition to numerical evidence, we show that our conjectural zeta function satisfies the Weil conjectures and we elucidate its connection with Monsky--Washnitzer cohomology.

We investigate a three-dimensional lattice model of two flavors of massless staggered fermions coupled through two independent four-fermion interactions, $U_I$ and $U_B$. Using large-scale fermion-bag Monte Carlo simulations, we map out the phase diagram in the $(U_I, U_B)$ parameter space and identify three distinct phases: a massless fermion phase, a symmetry-broken massive phase, and a symmetric massive phase. When one of the interactions is absent ($U_B=0$), the system undergoes a single continuous transition directly connecting the massless and symmetric massive phases, a feature previously associated with unconventional fermion mass generation. We find that turning on a nonzero $U_B$ separates this direct transition into two successive transitions with an intermediate symmetry-broken phase. The transition from the massless to the broken phase belongs to the Gross-Neveu universality class, while the transition from the broken to the symmetric massive phase falls into the three-dimensional XY universality class. Our results indicate that the special point at vanishing coupling, where the direct transition occurs, plays the role of a multicritical point organizing the surrounding phase structure. These findings provide a unified lattice perspective on conventional and unconventional mechanisms of fermion mass generation within a single model.

A fast simulation method is presented for a depleted monolithic active pixel sensor, which uses a data driven parameterization of the charge collection and propagation. This approach provides an efficient alternative to TCAD simulations, particularly for sensors whose proprietary process details - such as doping profiles or implant geometries - are unavailable. Data was obtained with a MALTA2 sensor fabricated in a 180 nm CMOS imaging technology on 30 μm epitaxial silicon using the MALTA beam telescope at CERN SPS. The model reproduces the measured inpixel efficiency with high accuracy and enables a realistic yet computationally lightweight analog pixel simulation. This method will be further employed in optimizing the digital sensor design for applications in high-rate particle tracking and high-granularity calorimetry.

In this proceeding, we discuss the finite-size scaling analysis of the order parameter related to the chiral phase transition in QCD with two massless quarks. We use data obtained in lattice QCD calculations performed with highly improved staggered quarks (HISQ) for a range of light quark masses, $1/240 \leq m_\ell/m_s \leq 1/27$ for different spatial volumes ($N_σ$) on Euclidean lattices with temporal extent $N_τ=8$, satisfying $3\,N_τ\leq N_σ\leq 10\,N_τ$. We observe that infinite volume extrapolated data for the order parameter agree reasonably well with the expected $O(2)$ scaling behavior even for physical ratios of the light-to-strange quark mass ratio. We quantify deviations from asymptotic scaling and perform a detailed analysis of the influence of finite-size effects in terms of temperature and quark masses at a fixed lattice cutoff. This is crucial for improving the reliability of the infinite-volume extrapolated estimate of the chiral order parameter and for a more precise determination of chiral phase transition temperature from direct Lattice QCD simulations.

Using first-principles numerical simulations, we find a new spatially inhomogeneous phase in a rotating gluon plasma. This mixed phase simultaneously contains regions of both confining and deconfining states in thermal equilibrium, separated by a spatial transition. The position of the boundary between the two phases is determined by the local critical temperature. We calculate the critical temperature of the local transition as a function of angular velocity and radius for a full (imaginary) rotating system and within a local thermalization approximation, and find an excellent agreement between these approaches. An analytic continuation of the results to the domain of real angular frequencies indicates that the confinement phase localizes at the periphery of the rotating system and the deconfinement phase appears closer to the rotation axis. We argue that the anisotropy of the gluon action in the curved co-rotating background can quantitatively explain the remarkable property that the spatial structure of this inhomogeneous phase disobeys the picture based on a straightforward implementation of the Tolman-Ehrenfest law. We also perform the first lattice simulation of rotating $N_f=2$ QCD which confirms that a similar picture is expected for theory with dynamical quarks.

We present an exhaustive BRST lagrangian description of partially massless bosonic fields in four-dimensional space. The basic fields are formulated in terms of two-component spin-tensors in (A)dS space where the tracelessness conditions are automatically fulfilled. The mass shell of partially massless fields is reformulated in terms of constraints on Fock space vectors including the second-class constraints. A conversion procedure for transforming second-class constraints into first-class ones is developed, allowing one to construct a Hermitian and nilpotent BRST charge in the Fock space under consideration. It is proven that the hermiticity and nilpotency restrict the conditions on the theory parameters, which are fulfilled only in dS space. The hermiticity of the BRST charge is incompatible with AdS space. The gauge invariant Lagrangian is constructed on the basis of the BRST charge, and for spin $s$ and depth $t$ the allowed states in the Lagrangian include only $(s-t)$ Stückelberg fields. Their exclusion leads to gauge transformations of degree $(s-t)$ for the physical fields. The Lagrangian equations of motion exactly reproduce the mass shell conditions. The Lagrangian in terms of conventional spin-tensor fields is also presented.

Proton decay searches in the next generation of water Cherenkov detectors, such as Hyper-Kamiokande, are expected to probe the $10^{35}$-year lifetime regime where atmospheric neutrino backgrounds and systematic uncertainties begin to play an increasingly important role. In this study, we employ the GiBUU framework and reevaluate the proton decay search sensitivity for the $\textrm{p}\rightarrow\textrm{e}^{+}π^{0}$ channel by incorporating a typical event reconstruction performance in water Cherenkov detectors. Using sophisticated models implemented in GiBUU -- most notably the mean-field potential and Boltzmann transport -- which have been benchmarked against accelerator neutrino scattering data, in particular pion production, we find that the resulting proton decay signal detection efficiency and atmospheric neutrino background rate are comparable to those previously evaluated for the current and near future water Cherenkov experiments using $\textit{ad hoc}$ nuclear models. In addition to pion final-state interactions, we evaluate the impact of differences in the Fermi momentum distribution of nucleons in the nucleus, as a source of systematic uncertainty, on the signal detection efficiency and the expected background event rate. We find that the uncertainty associated with pion final-state interactions is moderate, whereas the choice of Fermi momentum distribution can significantly affect the estimated atmospheric neutrino background rate and constitutes the dominant contribution. Our study provides an independent and complementary characterisation of nuclear effects on proton decay searches and helps to refine sensitivity estimates in the regime where systematic uncertainties become more relevant.

The calculation of resonance form factors in effective field theory as well as on the lattice is a highly challenging task. In a recent paper, we proposed a novel method based on the introduction of a background field and the Feynman-Hellmann theorem to address the problem, and applied it to a toy model. In the present work we use this method for the electromagnetic form factors of the $ρ$-meson. By matching the results to Chiral Perturbation Theory, we provide a first, crude estimate of all three form factors of the $ρ$-meson within the effective field theory. Contact contributions to these form factors turn out to be substantial. A procedure for lattice calculations is outlined, paving the way for an ab initio approach to the problem.

We study a realization of neutrino masses within the linear seesaw mechanism based on non-holomorphic modular $A_4$ symmetry, extending modular-invariant flavor models beyond the conventional holomorphic framework. The model is constructed in a non-supersymmetric setting and involves six heavy $SU(2)_L$ singlet fermions, $N_R$ and $S_L$, together with a single flavon field, thereby significantly reducing the field content. The modular transformation properties of the Yukawa couplings under $A_4$ symmetry lead to a highly constrained neutrino mass matrix with a distinctive flavor structure. After presenting the general theoretical framework, we perform a systematic numerical analysis of neutrino phenomenology by restricting the modulus parameter $τ$ to the fundamental domain and scanning the allowed parameter space. We identify regions consistent with current neutrino oscillation data at the $3σ$ level and obtain predictions for currently unknown observables, including the absolute neutrino mass scale and leptonic CP-violating phases. We further examine the implications for neutrinoless double beta decay, highlighting testable signatures in upcoming precision oscillation and rare-process experiments. These results demonstrate the phenomenological viability and predictive power of non-holomorphic modular symmetry in linear seesaw neutrino mass models.

The thermodynamics of the $SU(4)$ gauge theory with a single flavor of fundamental quarks is analyzed on the lattice with dynamical fermion simulations, which is the low-energy sector of a realistic, strongly-interacting dark matter model -- the Hyper Stealth Dark Matter. The gravitational wave spectrum from the first-order confinement transition in the early universe is further calculated, where the effect of the dark sea quarks, which decrease the interface tension in the effective potential of the Polyakov loop, is shown numerically to lower the gravitational wave amplitude.

We investigate {the arrival time and the Faraday rotation} of extragalactic electromagnetic signals from fast radio bursts (FRBs) propagating through chiral cosmic media within the framework of Maxwell-Carroll-Field-Jackiw {(MCFJ)} electrodynamics. By treating the interstellar medium as a cold, ionized chiral plasma, {we derive} the time delay between two traveling signals, expressing it in terms of modified dispersion measures (DMs) {containing chiral contributions}. {The Faraday rotation angle is then} written in terms of modified rotation measures (RMs). By combining the DMs and redshift data from a set of FRBs, {we obtain} constraints on the chiral parameter magnitude at the order of $10^{-26}$--$10^{-24}$ GeV. {Using the Faraday} rotation formulae and RM measurements, {upper bounds as stringent as $10^{-43}$ GeV on the MCFJ parameters are also obtained.}

In this paper, we discuss the entanglement phase transition of pseudo entropy in CFTs. We focus on the case where the in-state and the out-state are different boundary states related by boundary condition changing operators. We compute the pseudo entropy with BCFT methods and find a phase transition with respect to the conformal weight of the boundary condition changing operators. For holographic CFTs, we confirm that the CFT results match that evaluated in AdS.

We present a holomorphic quantization scheme for free point particles on two-dimensional constant curvature Riemannian backgrounds. The procedure is based on a Lagrangian embedding of the particle configuration space into a product of coadjoint orbits of the background isometry group. Examples are provided by particles on the plane, torus, sphere, and hyperbolic plane, with or without a monopole field. We elaborate the method by recovering the Hamiltonian spectrum and the wave functions on such spaces. As a by-product, we obtain a geometric and physical interpretation of Repka's result on the decomposition of tensor products of $\mathbf{SL}(2,\mathbb{R})$ discrete series representations.

We present the first complete calculation of hadron production in deep-inelastic scattering (DIS) at finite transverse momentum to next-to-next-to-leading order (NNLO) in perturbative QCD. To overcome the long-standing challenge of infrared divergences in semi-inclusive processes with identified final state hadrons at finite transverse momentum, we implement the recently developed $q_T$-subtraction framework based on the recoil-free jet definition. By utilizing the winner-take-all recombination scheme, we achieve a consistent factorization for hadron-jet associated production, enabling the inclusion of $\mathcal{O}(α_s^3)$ corrections. Our results demonstrate a significantly improved stabilization of the perturbative expansion and a reduction in scale uncertainties compared to previous next-to-leading order predictions. We find that the NNLO corrections are essential for a robust description of high precision multiplicity data from the ZEUS collaborations. This work provides a high precision theoretical foundation for the upcoming Electron-Ion Collider era and establishes a new benchmark for the exploration of the nucleon's three-dimensional structure.

We investigate Bogomolnyi-Prasad-Sommerfield (BPS) topological configurations in the $CP^1$ model minimally coupled to a Maxwell gauge field. Starting from the nonlinear $O(3)$ sigma model, we explicitly construct its classical mapping to the $CP^1$ formulation, emphasizing the emergence of a local $U(1)$ gauge symmetry intrinsically linked to the Fubini-Study target-space geometry. Focusing on the static sector, we analyze magnetized BPS lump configurations. By examining the topological boundary conditions, we show that the magnetic flux is quantized and fully determined by the asymptotic behavior of the gauge field. Within the BPS framework, we identify the specific self-interaction potential required for the existence of self-dual configurations. A detailed analysis of the asymptotic behavior, both near the lump core and at spatial infinity, demonstrates that finite-energy solutions necessarily correspond to lump-like BPS configurations in the $CP^1$ target space, with the scalar field vanishing asymptotically. The resulting BPS equations are solved numerically, and the main physical properties of the solutions are presented. We find that these configurations are regular, magnetically localized, and energetically stable, with their internal structure rigidly controlled by the geometry of the $CP^1$ target space. In contrast to Abelian Higgs vortices, the configurations discussed here do not rely on spontaneous symmetry breaking but arise purely from the geometry of the $CP^1$ target space.

The topology of the ground state of QCD manifests itself through the dependence of its potential energy density on the CP-violating angle $Θ$. This is computed in the present note within a three-flavor linear meson model. It is shown that the isospin symmetry breaking condensate leads to a 5\% shift in the scale of topological fluctuations, due to which a value emerges in close agreement with lattice QCD determinations. Transparent analytical computation makes explicit the physical content of each step significantly correcting the starting crude estimate of topological susceptibility and eventually achieving the quite accurate final result. This clear physical interpretation lends also a certain pedagogical value to the proposed treatment.

Measurements of the production of the W$^{\pm}$ and Z$^0$ bosons at midrapidity in pp collisions at $\sqrt{s} = 13$ TeV with ALICE at the Large Hadron Collider (LHC) are presented. The W$^{\pm}$ and Z$^0$ bosons are detected via their (di)electronic decay channels, with the electron reconstruction performed in the midrapidity region ($|y|< 0.6$). The $p_{\rm T}$-integrated and $p_{\rm T}$-differential production cross sections of electrons from W$^{\pm}$ decays in the interval $30 < p_{\rm T} < 60$ GeV/$c$, as well as the $p_{\rm T}$-integrated production cross section of Z$^0$ bosons, are measured. The results are described by perturbative QCD calculations using different sets of parton distribution functions. The production of W$^{\pm}$ bosons and azimuthally correlated associated hadrons is also measured as a function of the charged-particle multiplicity for the first time at the LHC. The former increases approximately linearly with the charged-particle multiplicity, while for the latter, there are hints of a faster-than-linear increase. These observations are compared with theoretical calculations.

The measurement of the transverse-momentum fraction ($z^{\rm ch}$) carried by prompt and non-prompt J/$ψ$ in charged-particle jets in proton--proton collisions with a center-of-mass energy $\sqrt{s}= 13$ TeV is reported by the ALICE Collaboration at the CERN Large Hadron Collider. The measurement is based on a Transition Radiation Detector triggered data sample corresponding to an integrated luminosity of $\mathcal{L} = 1.63\pm0.03$ pb$^{-1}$. Inclusive J/$ψ$ mesons with transverse momentum ($p_{\rm T}$) above 1 GeV/$c$ are reconstructed at midrapidity through the electron--positron decay channel. The prompt and non-prompt J/$ψ$ contributions are separated using the secondary vertices from beauty-hadron decays. Jets are reconstructed from J/$ψ$-meson candidates and charged particles using the anti-$k_{\rm T}$ algorithm with jet resolution parameter $R = 0.4$ in the pseudorapidity range $|η_{\text{jet}}| < 0.5$. The distribution of the charged-particle jet $p_{\rm T}$ fraction carried by the prompt and non-prompt J/$ψ$ mesons, $z^{\rm ch}$, is measured in the range $0.3 < z^{\rm ch} \leq 1.0$ for the jet-$ p_{\rm T}$ range $7 < p_{\rm T}^{\rm jet} < 15$ GeV/$c$. The measured $z^{\rm ch}$ distributions for prompt and non-prompt J/$ψ$ are compared to pQCD calculations and PYTHIA 8 simulations. Prompt and non-prompt J/$ψ$ production within charged-particle jets are found to be qualitatively well reproduced by PYTHIA 8 for $z^{\rm ch}<0.9$, however, as $z^{\rm ch}$ approaches one, the simulations overshoot the measured data, indicating an overestimation of the fraction of isolated J/$ψ$. This observed tension between data and simulations in the highest $z^{\rm ch}$ bins reflects the challenge of simulating hadronization of low-momentum jets.

We present a comprehensive analysis of the bottom-charmed ($B_c$) meson spectrum within the inverse matrix QCD sum rules formalism. In this framework, conventional QCD sum rules are recast as an inverse problem, allowing for the direct reconstruction of hadronic spectral densities from first principles without invoking phenomenological continuum parametrizations or quark-hadron duality assumptions. We compute the masses and decay constants of conventional $B_c$ mesons with quantum numbers $J^P = 0^-$, $1^-$, $0^+$, and $1^+$. The obtained results are in close agreement with available experimental measurements and are consistent with predictions from various theoretical and phenomenological approaches. The inverse matrix formulation exhibits improved numerical stability and reduced systematic uncertainties relative to standard implementations, highlighting its suitability for precision spectroscopy of heavy quarkonium systems.

We provide a comprehensive lattice formulation of various types of the Dirac operator indices, employing $K$-theory to classify the Wilson Dirac operator via its spectral flow. In contrast to the index of the overlap Dirac operator defined through the Ginsparg-Wilson relation, which is restricted to flat tori in even dimensions, our formulation offers several key advantages: 1) It can be applied straightforwardly to the Atiyah-Patodi-Singer index for manifolds with boundary. 2) The boundary can be curved, allowing for the inclusion of gravitational background effects. 3) The mod-2 index in both even and odd dimensions can be defined as a natural extension of the same formulation. In this talk, we present the mathematical proof and provide numerical evidence supporting the formulation.

In this thesis, we assess the sensitivity of next-generation long-baseline neutrino oscillation experiments, DUNE, T2HK, and T2HKK, to three popular beyond the Standard Model (BSM) scenarios. Within the three BSM studies, we examine: (i) long-range neutrino-matter interactions induced by flavor-dependent, anomaly-free gauged baryon-lepton symmetries mediated by ultra-light vector boson, showing that DUNE and T2HK can constrain, discover, and in some favorable cases distinguish among different symmetries; (ii) Lorentz invariance violation (LIV), where we derive analytical dependencies of CPT-conserving and CPT-violating LIV parameters on baseline and energy, highlighting the superior reach of DUNE in probing all the LIV parameters, in contrast to T2HK, which is essentially blind to the CPT-conserving LIV parameters; and (iii) active-sterile oscillations over a broad range of $Δm^2_{41}$, where we derive sensitivity to CP phases for benchmark choices of $Δm^2_{41}$ and establish exclusion limits, emphasizing the role of near detectors. Together, these studies show that future long-baseline facilities not only resolve the neutrino mass ordering, value of $δ_{\rm CP}$, and $θ_{23}$ octant, but also provide powerful probes of BSM physics.

Inclusive production of $ψ(2S)$ mesons is measured for the first time at midrapidity ($|y| < 0.9$) in pp collisions at $\sqrt{s} = 13$ TeV using the ALICE detector at the LHC. The measurement is performed in the transverse-momentum ($p_{\rm T}$) range of 4 to 16 GeV/$c$ and is based on events triggered by the Transition Radiation Detector, corresponding to an integrated luminosity of $1.7 \pm 1.8\%$ (syst.) pb$^{-1}$. The $p_{\rm T}$-differential production cross section of the $ψ(2S)$ and its ratio to the J/$ψ$ production cross section are presented. The results extend the accessible $p_{\rm T}$ range down to 4 GeV/$c$, and a mild rise of the $ψ(2S)$-to-J/$ψ$ cross section ratio with increasing $p_{\rm T}$ is observed. The measurements are compared with theoretical predictions based on NRQCD and the ICEM model.

Charged kaon--deuteron interactions offer a sensitive probe of the strangeness sector of QCD at low energy, particularly through their scattering parameters. Despite available theoretical predictions, experimental constraints remain absent, particularly for the $\rm K^{-}d$ system. The first measurement of femtoscopic correlation functions for $\rm K^{-}d \oplus K^{+}\overline{d}$ and $\rm K^{+}d \oplus K^{-}\overline{d}$ particle pairs in Pb--Pb collisions at $\sqrt{s_{\rm NN}}=5.02$ TeV recorded by ALICE at the LHC is presented. Correlation functions for both systems are analyzed in three centrality classes and fitted using the Lednický--Lyuboshitz model to extract the source size and strong interaction scattering lengths. The $\rm K^{-}d$ scattering length is found to be $\Re f_0 = -1.44 \pm 0.15\text{ (stat.)}^{+0.10}_{-0.10}$ (syst.) fm and $\Im f_0 = 1.34 \pm 0.33$ (stat.)$^{+0.21}_{-0.15}$ (syst.) fm, while for $\rm K^{+}d$ the value $\Re f_0 = -0.68 \pm 0.16$ (stat.)$^{+0.09}_{-0.09}$ (syst.) fm is obtained. These results provide the first constraints on the $\rm K^{\pm} d$ interaction modeling at low energy, delivering a long-awaited experimental benchmark for testing chiral QCD dynamics.

The phase diagram of two color QCD under influence of baryon density ($μ_B$), isospin ($μ_I$), chiral ($μ_5$ and $ν_5$) imbalances is investigated within the framework NJL model approach. This letter establishes the existence of weak dualities in two color quark matter--symmetries revealed when full thermodynamic potential is projected or restricted to specific condensation channels. These dualities provide a unified understanding of two interesting phenomena in the phase diagram: universal catalysis and chameleon effect of chiral chemical potential.

Based on a data sample corresponding to an integrated luminosity of 6.4~fb$^{-1}$ of $e^+e^-$ annihilation and collected with the BESIII detector at 13 center-of-mass energy points ranging between 4.600~GeV and 4.950~GeV, we report the first observation of the singly Cabibbo-suppressed decay $Λ_c^+ \to Σ^0 K_S^0 π^+$ with a statistical significance of 5.9$σ$. The branching fraction is determined to be $\mathcal{B}(Λ_c^+ \to Σ^0 K_S^0 π^+) = (0.58 \pm 0.14_{\rm stat.} \pm 0.04_{\rm syst.}) \times 10^{-3}$. In addition, the decay $Λ_c^+\toΣ^0K_{S}^{0} K^+$ has also been investigated, and the first evidence for this decay is obtained with a statistical significance of 3.7$σ$.

Four-volumes ($V_4=$ spatial-3-volume$\times$lifetime) are calculated within the model of three-fluid dynamics (3FD) and compared with those of the the JET AA Microscopic Transport Model (JAM). The calculations are performed for central Au+Au collisions at energies $\sqrt{s_{NN}}=$ 3 -- 19.6 GeV. These $V_4$ indicate optimal collision-energy ranges for realizing macroscopic high baryon-density matter. It is found that the 3FD four-volumes noticeably exceed those in the JAM, which indicates a stronger baryon stopping in the 3FD model as compared to that JAM. It is argued that this difference in the baryon stopping correlates with stiffness of the EoS implemented in these models. Contrary to JAM, the four-volume, where a baryon density ($n_B$) exceeds three times the normal nuclear density ($n_0$), does not exhibit a maximum as a function of $\sqrt{s_{NN}}$. It decreases monotonically with increasing $\sqrt{s_{NN}}$, remaining at a fairly macroscopic level (i.e. $V_4\geq 5.5^4$ fm$^4$/c). For higher baryon densities, $V_4$ exhibits maxima in its dependence on $\sqrt{s_{NN}}$. The optimal energy range for densities $n_B/n_0>$ 4 is located at $\sqrt{s_{NN}}=$ 3.2--8 GeV. Even for $n_B/n_0>$ 6, the four-volume remains quite macroscopic ($V_4\geq 4^4$ fm$^4$/c) at $\sqrt{s_{NN}}=$ 4.5--9 GeV contrary to the JAM.

In this work, we have studied the two-particle cumulant in pp, d-Au, and Au-Au collisions. The two-particle cumulant was treated as an event-by-event distribution, and its skewness and kurtosis were analyzed. The non-flow correlations, like jets and decays, constantly produced a skewed distribution, regardless of the model used. On the contrary, HYDJET++ produced a smooth Gaussian distribution at higher $η$ windows in fixed impact parameter collisions. The skewness increased consistently for higher $η$ windows in all non-QGP models like PYTHIA, PHOJET, QGSJET, and DPMJET. The kurtosis of the distribution also increased with $η$ windows in non-QGP models. The skewness and kurtosis of the distributions produced by HYDJET++ decreased with $Δη$ and eventually reached zero at higher $Δη$.

We make predictions for the expected multiplicities of exotic charmed hadrons and charmed nuclei in Au+Au collisions at SIS100 and LHC beam energies, using input on light hadron and charm production from the UrQMD transport model, and applying the Thermal-FIST model. We demonstrate that the CBM experiment has the capability to explore these states with production rates of one per 3 seconds for $χ_{c0}(1P)$ and $χ_{c1}(1P)$, and one every 3 minutes for $X(3872)$ at the expected data taking rates. Due to the higher baryon density at CBM compared to the LHC, charmed nuclei, if they exist, will be equally abundant at CBM as at the LHC, even though the total charm production at CBM is much lower.

We construct simple qubit-regularized Hamiltonian lattice gauge theories formulated in the monomer--dimer--tensor-network (MDTN) basis that are free of sign problems in the pure gauge sector. These models naturally realize both confined and deconfined phases. Using classical Monte Carlo methods, we investigate the associated finite-temperature phase transitions and show that they exhibit the expected universality classes of conventional SU(N) lattice gauge theories in various spacetime dimensions. Furthermore, we argue that second-order quantum phase transitions separating the confined and deconfined phases are likely to exist. Such critical points would provide a nonperturbative route to defining continuum limits of qubit-regularized gauge theories, potentially allowing Yang--Mills theory and related continuum gauge theories to emerge from finite-dimensional lattice constructions.

We present a non-singular, definition-level formulation of F-theory by replacing the traditional shrinking-fiber limit of M-theory with compactification on a tower-completed circle described using perfectoid geometry and condensed mathematics. This construction provides an intrinsic eleven-dimensional carrier for modular data and admits a canonical tilting and comparison procedure that yields elliptic geometry as an output rather than an auxiliary input. Using this framework, we establish a precise M-theory/Type IIB dictionary in the constant-coupling sector, showing how the physical axio-dilaton is fixed by eleven-dimensional geometric and topological data. The correspondence is tested at the level of the ten-dimensional bosonic effective action, including its topological couplings inherited from eleven dimensions. The tower-completed geometry naturally organizes global sectors in generalized cohomology, with charge data governed by K-theory and exhibiting a canonical prime-power torsion structure. We further show how this framework extends to varying-coupling backgrounds and duality defects, admits a natural adelic completion with prime-independence, and generalizes to higher-rank and U-duality geometries. We also discuss holographic aspects and the anomaly-refined extension of the duality group beyond the bosonic truncation. Together, these results provide a coherent, non-singular foundation for F-theory and its extensions.

It has been conjectured that there exists an E11-invariant formulation of eleven-dimensional supergravity in which the propagating fields of the theory are realised through an infinite tower of higher duals. In this work, we prove this conjecture explicitly within E11 exceptional field theory at the linearised level. Starting from the pseudo-Lagrangian, we construct parent actions for all higher gradient dual fields that are associated with the three-form, the six-form, and the dual graviton. We show that the resulting Euler-Lagrange equations constrain the Stueckelberg fields to be pure curls, ensuring that the higher duals propagate the same physical degrees of freedom as the original supergravity fields. The additional Stueckelberg fields, which are not predicted by the tensor hierarchy algebra, are shown to play a specific role as sources for the Labastida tensors of the higher dual fields.

We study the scaled transverse momentum spectra over a wide parameter space of state-of-the-art hydrodynamic simulation models in order to learn what information can be obtained from the shape of identified-particle spectra -- previously observed to be surprisingly universal across centrality and collision systems in both experimental data and hydrodynamic simulations. We study its sensitivity to each of 17 model parameters in the context of 4 different models for particlization when switching from the hydro description to the kinetic theory afterburner. We find that the strongest sensitivity is to parameters relating to bulk viscosity, free-streaming time, and the $\texttt{T$_\mathrm{R}$ENTo}$ nucleon width parameter $w$. However, we find that the model generally has surprisingly little flexibility in describing the scaled spectrum observable, despite the large number of parameters. Within this small range of parameter dependence, we further find significant tension in a simultaneous description of momentum-integrated observables. In particular, while the mean transverse momentum prefers a large value of the nucleon width parameter $w$, a small value is required to obtain scaled spectra that are consistent with experimental measurements. We speculate on the origin of these model tensions and possible missing physics in the commonly-used $\texttt{T$_\mathrm{R}$ENTo}$+free streaming+hydro+afterburner simulation model.

We consider a renormalizable extension of the minimal supersymmetric standard model (MSSM) endowed by an R and a gauged B - L symmetry. The model incorporates chaotic inflation driven by a quartic potential, associated with the Higgs superfields which lead to a spontaneous breaking of U(1)B-L. Consistency with the ACT data is achieved by considering a fractional shift-symmetric Kaehler potential which includes two free parameters (p,N) constrained in the ranges 1.355<p<6.7 and 6x10^-5<N<0.7. An explanation of the mu term of the MSSM is also provided, under the condition that a related parameter in the superpotential is somewhat small. Baryogenesis occurs via non-thermal leptogenesis which is also realized by the inflaton's decay to the lightest and/or next-to-lightest right-handed neutrinos for normal ordered light neutrino masses.

Exchange by two neutrinos (as well as by other fermions) generates a long-range parity-violating (PV) potential of the form $\sim G^2/r^5$, with characteristic range $\hbar/(2m_νc)$. This interaction produces a $-0.8\%$ correction to the effective weak charge of the Cs atom, thereby resolving the $2σ$ discrepancy between the Standard Model prediction and the Cs PV measurement, and yielding $\sin^2θ_W=0.2375(19)$ at $q^2\approx0$ (to be compared with the Standard Model value $\sin^2θ_W=0.2387$). We derive updated limits on an extra $Z'$ boson and obtain a constraint on isospin-conserving oblique radiative corrections characterized by the Peskin--Takeuchi parameter, $S=-0.32(53)$ at $q^2\approx0$. The corresponding relative correction to the proton weak charge is $3\%$, and a similar correction is expected for the electron weak charge.

The Canadian subatomic physics community establishes its scientific, and thus funding, priorities through periodic Long-Range Plans (LRP). The community is now putting together a new LRP, which will be in effect from 2027 through 2034, with its scope extending through 2041. As part of this process, the Canadian Institute of Nuclear Physics (CINP) has put together a strategic report, following an extensive consultation process. The report describes the broad and ambitious research program undertaken by the Canadian nuclear physics research community, both onshore and abroad, touching on key questions regarding the origin, evolution, and structure of visible matter in the universe. This document provides a grid of different Canadian nuclear physics projects undertaken now and in the future, and their associated timelines. It concludes with specific recommendations for maximizing Canadian scientific output in nuclear physics.

Doubly Special Relativity (DSR) augments special relativity by introducing, alongside the invariant speed of light $c$, a second observer-independent scale typically associated with the Planck regime. At the level of effective wave equations this principle manifests itself through deformed dispersion relations and energy-dependent spatial operators. Here we quantify such effects in a prototypical exactly solvable bound-state problem: the three-dimensional Klein--Gordon oscillator generated by a non-minimal momentum coupling that yields isotropic harmonic confinement while preserving rotational symmetry. We analyze two standard DSR realizations (Amelino--Camelia and Magueijo--Smolin, parametrized by an invariant energy scale $k$) as well as a generalized DSR framework based on a first-order expansion in the Planck length $l_p$. After stationary reduction and separation in spherical coordinates, the eigenfunctions retain the generalized-Laguerre and spherical-harmonic structure of the undeformed oscillator, whereas DSR deforms the algebraic quantization condition that relates the principal oscillator number $N=2n+\ell\in\mathbb{N}_0$ to the relativistic energy. Closed-form spectra are obtained for the standard DSR cases, and perturbative Planck-suppressed shifts are derived for the generalized model. In all realizations the deformation induces branch-dependent shifts of both positive- and negative-energy solutions, which increase with excitation and vanish smoothly in the limits $k\to\infty$ or $l_p\to0$. The main goal of this paper is to extract analytic spectra and Planck-suppressed shifts that enable a direct comparison between different DSR prescriptions in a fully three-dimensional setting.

The action for self-dual gauge fields that emerges from the recently constructed superstring field theory is found. The new superstring field theory reduces to that of Sen in a certain limit, and in this limit the new action for self-dual gauge fields reduces to Sen's action for such fields. The theory describes two decoupled self-dual gauge fields that couple to two metrics, with each gauge field coupling to only one of the metrics. The action also features a background metric, and the non-linear coupling to these three metrics is non-standard. There are two spin-two gauge invariances, and diffeomorphisms arise from the diagonal subgroup.

This paper investigates the physical and kinematic properties of dust-rich regions in a small sample of group-centered elliptical galaxies, emphasizing their connection with the hot X-ray emitting gas and detailed dust grain characteristics. Comprehensive multi-wavelength data, including H-alpha and CO emission detected by MUSE and ALMA, demonstrate the presence of dust clouds embedded within complex, hot X-ray atmospheres shaped by AGN feedback. X-ray images show bubbles and cavities surrounded by bright rims. We find that dust regions containing molecular gas traced by CO are preferentially located at the rims of these X-ray cavities, suggesting that AGN-driven outflows enhance the condensation of cold, dusty gas at these compressive interfaces. Kinematic measurements indicate that molecular and ionized gas phases are dynamically and spatially linked, supporting the framework of a multiphase medium arising from the top-down condensation rain in the hot plasma and related chaotic cold accretion. Crucially, spatial variations in the total-to-selective extinction ratio Rv show that regions where dust, CO, and H-alpha emission coincide exhibit notably smaller Rv values, implying steeper extinction curves and the predominance of smaller or less evolved dust grains within these mixed-phase environments. This contrasts with larger Rv values found elsewhere in the dust clouds, suggesting grain growth or survival mechanisms within shielded cold gas.

Cosmic strings are topological defects arising in a variety of cosmological scenarios, as the Universe undergoes symmetry-breaking phase transitions, whose discovery would offer valuable insight into the high-energy physics that shaped the early Universe. To interpret such a detection, robust theoretical models are essential. The Velocity-dependent One-Scale (VOS) model is particularly prominent: it self-consistently treats the network as a thermodynamic system, characterizing its key properties, and predicting its large-scale evolution. This has recently been extended to include superconducting cosmic strings, which carry additional degrees of freedom, giving rise to the Charge-Velocity dependent One-Scale (CVOS) model. One limitation of the latter model is that it only included loss mechanisms for the charge and current. Here, we extend this model by phenomenologically including a possible energy source mechanism, specifically by allowing current-carrying strings interactions with an external magnetic field. We discuss how this coupling impacts the network's evolution, present and classify the physically allowed scaling solutions for this extended CVOS model, and comment on the different impacts of the external magnetic field on the evolution of the network's charge and current. Under our modeling assumptions, one of the ten physically plausible scaling solution enables the network to gain energy by this mechanism.

We study holographic charge measurement for continuous internal symmetries. Charged boundary operators are characterized by Wilson lines of bulk gauge fields ending on the boundary, while charge measurement is performed using U-shaped defects hanging from the boundary. We derive universal features of this process from a low-energy point of view, and show how the hanging defect picture mimics the thickening regularization of continuous symmetry operators in field theory. Furthermore, we provide explicit top-down realizations in AdS/CFT setups in Type IIB string theory and M-theory, featuring Abelian as well as non-Abelian symmetries. In the case of Type IIB constructions, we analyze the brane dynamics underlying the charge measurement process. Along the way, we also characterize how hanging brane configurations can be regarded as being topological, and demonstrate how tachyon dynamics account for their fusion rules.

We explore the prospects of detecting flavorful lepton number violation at the Electron-Ion Collider (EIC) through resonant production of heavy neutral leptons (HNLs), resulting in $e^- p \to \ell^+_α+ k\, j+X$, where $α\in \{e, μ, τ\}$ and $k$ denotes the number of jets. We work in the $ν$SMEFT framework of the Standard Model Effective Field Theory augmented with $n$ singlet HNLs, one of which is in the mass range $10-100$~GeV, within kinematic reach of the EIC. To explore the EIC sensitivity, we focus on the HNL production mechanism induced by mixing with light neutrinos. We study kinematic distributions for signal and backgrounds, including hadronization and detector effects, and suggest a set of cuts to minimize backgrounds. In the mass range considered, we find that the EIC with muon detection capabilities and an integrated luminosity of $100~\mathrm{fb}^{-1}$ can reach sensitivities comparable to the strongest direct (LHC) and indirect constraints, and is especially relevant in the $ν$SMEFT framework beyond dimension four. Our study motivates further assessment of muon detection capabilities at the EIC and $τ$ hadronic reconstruction, as well as a more general theoretical analysis involving production mechanisms mediated by higher-dimensional operators in the effective theory.

In this dissertation, we study various aspects of type IIB string theory compactified on freely acting orbifolds. We focus particularly on asymmetric orbifolds, which are examples of non-geometric string compactifications and constitute an intriguing corner in the string landscape. We describe the orbifolds using two complementary approaches; the duality twists and the lattice approach. First, we discuss the general construction of freely acting (asymmetric) orbifolds, focusing on the closed string sector. Afterwards, we present explicit examples of orbifolds preserving $\mathcal{N} = 6,4,2$ or $0$ supersymmetry in five dimensions and we demonstrate the connection between freely acting orbifolds and Scherk-Schwarz reductions by matching the lightest untwisted orbifold states with those arising form the effective supergravity theory. Regarding the orbifold twisted sectors, we show that at special points in the moduli space, generically massive states can become massless. Furthermore, we show that the spectrum of non-supersymmetric orbifolds can become tachyon-free away from special loci in the moduli space. Next, we turn our attention to the orbifold moduli space, and we focus on orbifolds preserving $\mathcal{N} = 2$ supersymmetry in five dimensions, for which we determine the classical hypermultiplet and vector multiplet moduli spaces. Moreover, by constructing dual orbifold pairs, we argue that no quantum corrections to the metric on the vector multiplet moduli space arise. Then, we discuss aspects of the swampland program in the context of freely acting asymmetric orbifolds and we verify that the swampland distance conjecture is valid in the non-geometric compactifications of string theory studied in this thesis. Finally, we focus on the D-brane spectrum of freely acting orbifolds and we determine the conditions under which the various D-branes survive the orbifolding.

We derive a fluctuation dissipation relation connecting the drag and diffusion jet transport coefficients for an energetic light quark traversing a non-perturbative thermalized gluon plasma. The hard quark is taken to be close to on-shell, with an energy scale parametrically larger than the medium temperature. We introduce a general complex-valued function for each transport coefficient. Evaluating these in the deep Euclidean momentum region enables their expression in terms of local operators. Using contour-integration techniques, we relate these local operators, after vacuum subtraction, to the physical transport coefficients that arise along a branch cut, close to light-like dispersion. The derived relation relates the longitudinal drag coefficient to the longitudinal and transverse diffusion coefficients, a $4^{\rm th}$ order fluctuation, and the thermal gluon condensate.

We apply, for the first time, Bayes Linear Emulation and History Matching to the calibration of non-perturbative models in Monte Carlo event generators. In contrast to the usual approach of "Monte Carlo tuning", History Matching does not result in best-fit plus ellipsoidal parameter uncertainty estimates but instead identifies all parameter space regions that are consistent with data. This approach leads to a systematic and robust quantification of parametric uncertainties in the models, especially in those challenging cases where different, possibly disjoint, regions of parameter space deliver similar results, which are usually not properly treated with current methodology. We highlight the power of this method with the hadronisation models available through Sherpa: the built-in cluster fragmentation Ahadic and string fragmentation through an interface to Pythia.

We show that within the Standard Model any system of hadronic weak charm decays related by $U$-spin satisfies the following rate sum rule: (sum of CF and DCS CKM-free rates) divided by (sum of SCS CKM-free rates) = 1, which holds up to second order in $U$-spin breaking. We test this sum rule against available data and find that it is well satisfied in all cases. For systems in which some decay rates have not yet been measured, we use this sum rule to predict the missing rates.

In this note we initiate the study of ``twisted holography'' for the dualities involving the BFSS matrix quantum mechanics and the IKKT matrix model in their $N \rightarrow \infty$ limits. We identify the admissible twists of each model, compute their cohomology in the BV-BRST formalism, and identify them -- in the planar limit and in perturbation theory around the trivial background -- with corresponding twists of IIA and IIB string theories, respectively. The twisted gravitational duals make manifest certain infinite dimensional symmetry algebras. In the BFSS example, the dual IIA supergravity twists are also obtained as certain zero mode truncations of the minimal (1/16-BPS) and maximal (1/4-BPS) twists of eleven-dimensional supergravity.

We show that when an anomalous and a non-anomalous global symmetries are spontaneously and simultaneously broken, the resulting Nambu--Goldstone boson is associated with the non-anomalous one. Applied to the Majoron, this implies that its underlying symmetry is $B-L$ rather than $L$, naturally resolving the domain wall problem. This result further demonstrates that effective couplings of the form $a \, F \, \tilde{F}$ do not uniquely indicate an anomalous origin, allowing the detection of non-anomalous Nambu--Goldstone bosons, such as the Majoron, in searches for axion--gauge boson interactions, potentially serving as evidence for the existence of new charged fermions.

The dark photon is a well-motivated ultralight dark-matter candidate that may couple to the Standard Model through kinetic mixing. We search for dark-photon dark matter in the mass range $m_{A'}c^2 = 4\times10^{-12}$-$2\times10^{-9}\,\mathrm{eV}$ (1-500 kHz) using a three-axis magnetometer inside a large conductive shielded room. We set new laboratory limits on the kinetic-mixing parameter $ε$, improving upon previous laboratory bounds by up to three orders of magnitude. Our search exploits a geometry-defined null response along one axis as a noise reference; a subtraction procedure reduces the noise floor and improves sensitivity. These results establish the strongest laboratory constraints in this mass range and illustrate how null-axis magnetometry can broaden terrestrial searches for ultralight vector dark matter.

High-energy partons lose energy while propagating through the hot, strongly interacting medium produced in ultrarelativistic nucleus-nucleus collisions, leading to a suppression of particle production at high transverse momentum ($p_\mathrm{T}$). The dependence of this energy loss on the size of the colliding nuclear system has yet to be firmly established experimentally. This Letter presents a systematic study of charged-particle suppression across four different nucleus-nucleus collision systems using nuclear modification factors ($R_\mathrm{AA}$) measured by the CMS Collaboration at the CERN LHC. Previous CMS measurements of $R_\mathrm{AA}$ in oxygen-oxygen, xenon-xenon, and lead-lead collisions are recast with identical $p_\mathrm{T}$ intervals and are complemented by the first measurement of the charged-particle $R_\mathrm{AA}$ in neon-neon collisions at $\sqrt{s_\mathrm{NN}}$ = 5.36 TeV. The neon-neon data correspond to an integrated luminosity of 0.76 nb$^{-1}$. The $R_\mathrm{AA}$ in all collision systems examined show similar qualitative trends, but have a magnitude which is ordered with the nucleon number A. The $R_\mathrm{AA}$ feature a downward slope at low $p_\mathrm{T}$, a local minimum at around 5$-$7 GeV, and an upward slope with increasing $p_\mathrm{T}$. The $R_\mathrm{AA}$ are also compared in terms of A$^{1/3}$, which is proportional to the nuclear radius. Models including only initial-state nuclear effects fail to reproduce the observed trends, whereas energy loss models reproduce the trends in the region $p_\mathrm{T}$ $\gt$ 9.6 GeV.

This document presents a synthesis of the theory contributions and discussions from the 2nd EIC-France Workshop, held at IJCLab (Orsay) on 1-3 December 2025. The workshop brought together members of the French hadron-physics community to review recent theoretical developments relevant to the future Electron-Ion Collider (EIC) and to coordinate national efforts in preparation for its early physics program. The report first summarizes the collider's initial running conditions and luminosity performance, as outlined in the EIC Early Science Matrix. It then provides concise overviews of the theoretical presentations on inclusive, semi-inclusive, exclusive, heavy-flavor, and small-x physics. Based on these discussions, two measurements emerged as especially well suited for early EIC operation and strongly aligned with areas of established French expertise: inclusive diffraction and inclusive quarkonium production. These channels offer clean signatures, robust theoretical interpretability, and direct sensitivity to fundamental QCD phenomena such as gluon saturation, heavy-quark dynamics, and the small-x structure of hadrons and nuclei. In addition, the workshop identified longer-term physics opportunities that will benefit from the full capabilities of the EIC after its ramp-up phase. These include accessing the three-dimensional structure of the pion through the Sullivan process and a broader program of exclusive three-body final states, both of which represent high-impact avenues for exploring hadronic structure and non-perturbative QCD. Together, the elements summarized in this report provide a coherent overview of the strategic priorities and scientific ambitions shaping the French community's contribution to the EIC physics program.

We study the three-body systems $DNN$ and $D^{*}NN$ within a hadronic molecular framework by combining a realistic nucleon-nucleon interaction with a $D^{(*)}N$ potential constrained by heavy-quark symmetry. The three-body Schrödinger equation is solved with the Gaussian Expansion Method, and the analytic structure of the spectrum is investigated using the Complex Scaling Method. We find that the $DNN$ system supports a robust and compact bound state in the $I(J^{P})=\tfrac{1}{2}(1^-)$ channel over a broad range of cutoff values, even when the corresponding $DN$ subsystem is weakly bound or unbound. For $D^{*}NN$, the spin-$1$ nature of the heavy meson and the associated spin-dependent forces generate a clear spin hierarchy: deeply bound states appear in both $0^-$ and $2^-$ channels, while the $1^-$ channel exhibits a characteristic two-branch pattern with a strongly bound compact branch and a more weakly bound, spatially extended branch. The root-mean-square radii indicate pronounced spatial compression compared with the deuteron scale, highlighting the cooperative roles of realistic $NN$ correlations, the $D^{(*)}N$ interactions, and heavy-quark symmetry in forming compact heavy-flavor few-body bound states. No three-body resonances under complex scaling are found in the explored parameter space. Our results provide quantitative benchmarks for future experimental searches for such charmed-meson-nuclear bound states.

The Perseus Cluster has been precisely measured by the legacy Hitomi telescope on the Si-group (Si, S, Ar, Ca) and Fe-group elements (Cr, Mn, Ni). These element abundance ratios provide insight into the typical behaviour of supernovae. In Paper II, we presented new massive star explosion models at various metallicity, assuming spherical explosions. We show that while the fitting is improved, some features (e.g., Ni/Fe) remain to be improved. In this article, we extend our calculation to an aspherical explosion using the jet-induced explosion mechanism. The detailed pre- and post-explosion chemical profiles are calculated with a large post-processing network to capture the production of odd-number elements (V, Mn, Cu) and iron-group elements. We further explore how the jet-driven explosions create the diversity of models which could be compatible with the observed diversity in terms of $^{56}$Ni-mass vs ejecta mass, Ti-V relation, and stellar abundances. Finally, we apply the new collapsar models in the Galactic Chemical Evolution context. We study how the galactic stars, including the Zn-enriched star HE 1327-2326, can put constraints on the relative rates of collapsar and some of its model parameters. We show that collapsar could lead to significant changes in some elements, e.g., Zn. Our study shows that the collapsar is a necessary component to explain multiple elemental trends observed in the Milky Way Galaxy.

The legacy Hitomi telescope has delivered the precise measurements of the chemical abundances in the Perseus Cluster, covering the Si-group (Si, S, Ar, Ca) and Fe-group elements (Cr, Mn, Ni). In Paper I (Leung et al., ApJ 2025), we examined the role of convection parameters and presented new core-collapse supernova (CCSN) explosion models at solar metallicity, which fit the observed abundance pattern. In this article, we extend our calculation for the stellar evolutionary models and CCSN models of the initial mass $15 - 60M_{\odot}$ and the metallicity $Z = 0 - Z_{\odot}$. The detailed pre- and post-explosion chemical profiles are calculated with a large post-processing network to capture the production of $α$-chain elements (e.g., Si, S, Ar), odd-number elements (e.g., P, K, Cl), and iron-group elements (e.g., Mn, Ni). We study the role of CCSNe in the production of these elements. We compare the galactic chemical evolution model based on the nucleosynthesis yield of the new massive stars and other yield tables from the literature. For each supernova yield, we perform parameter surveys and search for configurations that produce the best-fit model and best-rate model using the Perseus Cluster as the reference. From the survey, we study how individual chemical elements affect the contributions of massive stars and Type Ia supernovae in the cosmic chemical enrichment

We present a spectroscopic study of photometrically identified extreme emission-line galaxies (EELGs) with observed-frame equivalent widths (EWs) >5000 A of either H alpha or H beta + [OIII] in the CEERS legacy deep field utilizing JWST NIRSpec spectroscopy from the CAPERS, RUBIES, THRILS and CEERS surveys. This master sample allows for performance tests of photometric selections and unveils what types of sources, either AGN or young star formation, were producing excessive ionizing radiation in the early Universe. We identify AGN through broad H alpha emission-lines and report 6 new broad-line AGN at 3.5<z<7 identified by the deep (~8 hr) G395M THRILS survey. We investigate the photometrically selected EELGs in a color-color plot designed for ``Little Red Dot'' selection and demonstrate that it effectively removes AGN with non-extreme lines from the sample. EELGs with and without broad lines show similar optical line ratios. We compare emission-line morphology to EWs and continuum morphologies and find that [OIII] morphology is more compact at higher EW. ~10% of photometrically selected EELGs have broad Balmer lines, jumping to 35% in deep spectroscopy which indicates a significant fraction of photometrically selected EELGs may host AGN. However, many AGN selected as EELGs have incorrectly high photometric EWs. For sources with extreme emission-line EWs that pass our photometric criteria and host an AGN, we find that the narrow H alpha component dominates over the broad, especially in the highest-EW sources. This implies that even when an AGN is present, it does not dominate the extreme emission.

The bright gamma-ray blazar S5 1044+71 has been identified as showing very significant quasi-periodic oscillations in the Fermi-LAT data in recent studies, with a periodicity of about 3 years. With the completion of a new gamma-ray cycle, we aim to revisit the periodicity in Fermi-LAT data, and analyze all available multi-wavelength (MWL) data to search for possible correlations and time-lags. These observations will be used to test for the compatibility of the observed periodicity with a precessing jet from the supermassive black hole. We analyze data from Fermi-LAT, NuSTAR, Swift, AstroSat, ASAS-SN, ZTF, Pan-STARRS, and NEOWISE. In addition we present an analysis from historical observations from Palomar and Pulkovo. Single-band spectral variability, MWL correlations, and cross-correlations are computed. We then model the Fermi-LAT light curve with a precessing jet model, providing constraints on the geometry of the system and providing the evolution of the Doppler factor with time. The latter is used as input for MWL fitting of the spectral energy distribution. We confirm previous claims on the existence of a periodic gamma-ray signal. We detect significant spectral variability in gamma-ray, X-rays, and optical/UV data. We detect significant correlation between low-energy (infrared/optical/ultraviolet) data and gamma-rays, with a correlation index of about 1; the correlation between X-rays and gamma-ray is milder, with a correlation index of about 0.3. We do not detect any significant time-lag between bands. The Fermi-LAT light curve is successfully fit by a precessing jet model. The fit to the spectral energy distributions indicate that S5 1044+71 is a typical blazar, in which the gamma-ray emission is located beyond the broad-line region. All MWL observations we present in this work are consistent with the existence of a precessing relativistic jet from the supermassive black hole.

The recent detection of GRB 250702B, the longest gamma-ray burst observed to date with prompt emission lasting $\sim 2.5\times 10^4$ seconds, challenges the conventional collapsar model. Its remarkable features--including an extraordinary X-ray flare at $\sim 1.3$ days post-detection, a late-time transition from steep to shallow decay in the X-ray afterglow, and hard spectra extending from keV to MeV energies--point to a novel progenitor. Here we show that these multiwavelength signatures can be consistently explained by a relativistic jet powered by successive partial tidal disruptions of a white dwarf (WD) by an intermediate-mass black hole (IMBH). By modeling the time-dependent accretion rate from repeated partial disruptions and the resulting jet evolution, we show that the external forward and reverse shocks account for the long-term X-ray, near-infrared, and radio afterglow, whereas the luminous X-ray flare originates from internal energy dissipation caused by collisions between fast and slow relativistic ejecta associated with the final complete disruption. Our findings establish IMBH-WD tidal disruption events as a viable engine for ultra-long GRBs.

The construction of Dyson spheres, megastructures designed to capture the total radiative output of stars, can be one of the most compelling techno-signature scenarios for advanced extraterrestrial civilizations. By considering equilibrium temperatures, we investigate the luminosities and fluxes of Dyson spheres built around two promising classes of host stars: white dwarfs and red M-dwarfs. Using radiative balance arguments and representative stellar parameters, we compute the temperature-radius relationship for full energy interception and place these hypothetical structures on the Hertzsprung-Russell (H-R) diagram to assess their observational signatures. Our results show that Dyson spheres around white dwarfs produce cooler and fainter blackbody emissions, peaking in the near- to mid-infrared, while those around M-dwarfs radiate more strongly but at longer wavelengths. In both cases, the equilibrium temperature decreases as R_ D^-1/2, while the total luminosity and observed bolometric flux remain fixed by the stellar output. These findings highlight the astrophysical suitability of low-luminosity stars as Dyson sphere hosts and provide practical constraints for future techno-signature searches using infrared surveys.

Transient X-ray obscuration in Seyfert 1 galaxies is thought to arise from clumpy accretion-disk winds near the broad-line region (BLR), but the wind structure and its short-timescale variability are difficult to measure because high-resolution spectra are often suppressed during deep low states. We analyse a coordinated XMM-Newton/NuSTAR campaign on Mrk 335 in June 2021, complemented by long-term Swift monitoring, which captured the source in an intermediate-flux state that preserves strong RGS absorption features. We first model the broadband spectral energy distribution to determine the ionising continuum and then perform self-consistent photoionisation modelling of the RGS spectra. The stacked RGS spectrum requires three photoionised absorbers with time-averaged log xi approx 3.63, 3.10, and 2.01 and outflow velocities |v_out| approx 5820, 3210, and 2140 km/s. Their properties are broadly consistent with the three-phase obscurer reported in the 2009 intermediate state, indicating recurring multi-phase obscuration over decade timescales. Using five consecutive RGS observations, we track the wind evolution on day timescales and find strong variability in column density and ionisation in all phases, together with smaller but coherent changes in outflow velocity. During a flare, the low-ionisation phase shows an extreme drop in N_H, and the subsequent epoch exhibits an increase in outflow velocity in all phases, consistent with rapid restructuring and possible radiative acceleration in a clumpy wind. The high-ionisation phase responds most directly to changes in the ionising luminosity, while the lowest-ionisation phase shows at most a delayed response. Order-of-magnitude constraints place the obscurer at BLR scales (approx 10^3-10^5 R_g), and simple continuity arguments suggest kinetic power that can reach the percent level of L_bol for plausible estimates of geometry and clumpiness.

We present the optical follow-up of SNe 2022ywf and 2023zgx, two examples from the Iax subclass of thermonuclear supernova (SN) events. With peak absolute magnitudes of $M_\mathrm{V} = -13.7$ and $-14.4$ mag, respectively, both objects belong to the extremely low-luminosity (EL) population of the class. A common origin of SNe in the Iax subclass is still under debate since the distribution of certain observables may indicate that the extremely low-luminosity explosions form a distinct population. We aim to estimate the physical properties of the two EL objects, including mapping the ejecta structure. We perform spectral tomography on the spectral series of SNe 2022ywf and 2023zgx around their maxima to map the physical properties of the ejecta. Together with the analysis of BgVriz photometry, a wide range of observables can be studied to investigate their distribution against luminosity. The constrained chemical abundances of the ejecta are compared to the predictions of the hydrodynamic simulations with similar peak luminosities. Constant abundances provide a good match for the distribution of chemical elements for both SNe 2022ywf and 2023zgx. The discrepancies compared to the least luminous pure deflagration model N5def_hybrid are minor, especially at post-maximum epochs. The two SNe also share similar characteristics in their constrained density structures, as well as the evolution of the photosphere. The analysis supports the assumption that pure deflagration models can reproduce the main characteristics of SNe Iax, even for the EL population. The presented indirect observational evidence indicates that these objects show similar intrinsic properties to the relatively luminous Iax sample and fit into the velocity distribution of the subclass.

Galaxy models comprising several components (including dark matter) that are bound by the self-consistently generated gravitational field are readily constructed from distribution functions (DFs) that are analytic functions of the action integrals J. We explain why such models have unphysical velocity distributions unless the DFs of hot components satisfy certain conditions as J_φ-> 0. We show how DFs for both isotropic and radially biased spherical systems can be constructed with specified f(J). We show how to construct DFs for flattened systems with significant velocity anisotropy. Construction of self-consistent models rather than populations that are confined by an external potential leads to the conclusion that radially-biased spherical systems are generically unstable to quadrupolar perturbations. Chaos is likely key to maintenance of these constraints during adiabatic disc growth.

We present a computational framework for galactic evolution based on a coupled stochastic nonlinear oscillator, implemented with the \textbf{Stochastic Hopf Engine}. Gas density ($G$) and star formation rate ($S$) co-evolve through a supercritical Hopf bifurcation, capturing the transition from quiescent stability to merger-driven starbursts. Scatter in dark matter halo properties, modeled as multiplicative noise via the \textbf{Euler--Maruyama method}, broadens the bifurcation into a regime where noise-induced bursts occur below the deterministic threshold. Simulations reveal a periodic signature, the \textbf{Galactic Heartbeat}, emerging as a deterministic limit cycle validated by the \textbf{data3} resonance peak in the star-formation spectrum. A radial reduction yields an effective \textbf{Fokker--Planck equation} for burst amplitude; its stationary solution matches numerical PDFs, providing statistical closure. Including differential shear $Ω(r)$ and spatially varying bifurcation fields reproduces spiral morphologies and AGN-driven quenching. Driving the growth parameter sub-critical ($r_{agn} < 0$) yields ``Red and Dead'' cores via attractor collapse. Dark matter halo scatter suppresses mean star formation while enhancing intermittency, offering a minimal yet interpretable framework linking local feedback and global potentials to macroscopic galactic evolution.

Line intensity mapping provides a statistical approach to tracing the large-scale distribution of matter in the Universe. We apply the HI intensity mapping technique to interferometric data from the MeerKAT International GHz-Tiered Extragalactic Explorations (MIGHTEE) Survey, analysing 17.5 hours of a single pointing in the COSMOS field, using a 60 MHz sub-band in the frequency range 1332 - 1392 MHz ($0.02 \lesssim z \lesssim 0.07$). Using a delay-spectrum-based estimator, we measure the HI power spectrum on sub-megaparsec scales and compare it directly to the power spectrum inferred from a catalogue of individually detected HI galaxies in the same field. After mitigating low-level broadband contamination through conservative outlier flagging in the three-dimensional power spectrum, cross-correlation of time-split visibilities yields a statistically significant detection on scales $3 \lesssim k \lesssim 20 \, \mathrm{Mpc}^{-1}$ with a total signal-to-noise ratio of $\sim 13$. Over this range, the power spectra obtained from visibilities and detected galaxies are consistent within uncertainties and have comparable amplitudes of order $10^{-2}$ - $10^{-1}$ $\mathrm{mK}^2 \mathrm{Mpc}^3$. End-to-end validation is performed by propagating detected galaxies through the power spectrum estimator via both direct intensity-field construction and simulated visibilities, demonstrating agreement up to $k \sim 20 \ \mathrm{Mpc}^{-1}$, beyond which measurements become noise-dominated. A statistically significant correlation is also observed between the data and the simulated visibilities from the detected HI galaxies, which should be free of systematics. These results provide a self-consistent validation of interferometric HI intensity mapping at low redshift and demonstrate agreement with galaxy-based measurements within the same cosmological volume.

Enhanced inverse Compton (IC) losses at high redshift steepen diffuse radio spectra in galaxy clusters, making low-frequency (~100 MHz) observations favorable. However, low-frequency studies often lack the resolution needed to locate particle acceleration sites or separate diffuse emission from radio galaxies. In this paper, we unveil the properties of the radio relic in the distant cluster PSZ2G091.83+26.11 (z=0.822) by resolving the acceleration site and inspecting the downstream region. Using the European LOFAR (ILT) at 145 MHz, we study a radio relic at (sub-)arcsecond resolution for the first time below 1 GHz, complemented by arcsecond-resolution VLA data at higher frequencies. We confirm the diffuse emission is not a radio galaxy. A spectral index gradient toward the cluster center matches previous 5'' maps. High-resolution 0.4'' and 1.9'' images reveal emission ahead of the shock, connecting the relic to a radio galaxy. 1.9'' profiles across the downstream at 145 MHz and 3.0 GHz follow a log-normal magnetic field distribution. The 145 MHz shock surface shows a sharp discontinuity at the same location of a change in electron density, Rotation Measure, and fractional polarization, likely tied to magnetic field changes. Finally, we find hints of redshift evolution of the radio power versus cluster mass correlation. The impressive angular resolution achievable by the LOFAR long baselines is opening an unprecedented view of the low energetic plasma in galaxy clusters. This is extremely significant in the case of high-redshift clusters, where radio emission at low frequencies is less affected by energy losses but its detection is strongly limited by poor resolution.

The Tully-Fisher (TF) and Fundamental Plane (FP) relations are widely used to infer extragalactic distances and peculiar velocities, enabling measurements of large-scale velocity statistics and cosmological parameters. Using the Millennium-TNG hydrodynamical simulation, we assess the accuracy of these methods in the presence of realistic galaxy formation physics. We find that, while the 2-point statistics of velocities are reliably inferred on scales larger than $\sim10\,\hMpc$, significant systematic deviations arise on smaller scales. These deviations originate from spatially correlated residuals in the TF and FP relations, driven by correlations between galaxy structural properties, star-formation history, and the local environment. As a result, TF- and FP-inferred velocity fields exhibit spurious correlations with the galaxy density field that cannot be explained by random scatter alone. We show that extending the TF and FP relations to include additional galaxy properties -- such as star formation rate, gas mass, and stellar mass -- mitigate these environmental correlations, particularly for late-type galaxies. Our results demonstrate that galaxy formation physics induces significant systematics in peculiar velocity measurements on non-linear scales, and that neglecting these effects may bias cosmological analyses.

Voids are parts of the cosmic web least affected by non-linearities and baryonic feedback. We thus calculate the sensitivity of voids to the nature of dark matter (DM), using ultra-light axions as a concrete model and the ongoing Dark Energy Spectroscopic Instrument (DESI) and \textit{Euclid} galaxy surveys as observational settings. We simulate axion effects on voids using mass-peak patch simulations and find that: (i) axions suppress the formation of lower-mass halos leading to the merging of smaller (radius $< 25\,\mathrm{Mpc}/h$) voids into fewer larger (radius $> 25\,\mathrm{Mpc}/h$) voids; and (ii) voids in the presence of axions are emptier of halos, thereby suppressing the void-halo correlation function. These effects strengthen as axion particle mass $m_\mathrm{a}$ decreases. We forecast improvements in axion constraints from the void size function (VSF; the void number density as function of their radius). A \textit{Euclid}-like survey (effective volume of $73\,\mathrm{Gpc}^3$ with a prior on the other $Λ$CDM cosmological parameters from the Simons Observatory cosmic microwave background experiment) can limit the axion energy density (for $m_\mathrm{a} = 10^{-25}\,\mathrm{eV}$) to $< 4.6\%$ of the DM (at $95\%$ credibility), about two times stronger than current limits. Conversely, we show that a Universe with a dark sector consisting of axions at the $10\%$ level, as motivated by the string axiverse, can be recovered with $\sim 2 σ$ preference. A DESI-like survey achieves comparable results. Axion and $Λ$CDM parameters have different degeneracies given VSF and galaxy power spectrum data, indicating future combined analyses will be most powerful in disentangling the DM nature. We anticipate our results will extend to other (e.g., warm or interacting) DM models.

We present a Monte Carlo study of galactic cosmic-ray (GCR) energy deposition and its implications for stratospheric chemistry, performed with the Geant4 toolkit. Primary nuclei (protons, $α$, CNO, and Si) were propagated through an atmosphere modeled from 0 to 120~g~cm$^{-2}$, considering both Polar ($R_{\mathrm{c}}=0.1$~GV) and Equatorial ($R_{\mathrm{c}}=15$~GV) geomagnetic cutoff conditions. The simulations resolve the variation of energy deposition with altitude for primary and secondary particles, revealing that $\sim$~96\% of the stratospheric energy budget arises from cascade secondaries within the 15--35~km domain. By converting layer-resolved energy deposition into ion pair production rates, we quantify the resulting formation of odd nitrogen (NO$_{\rm x}$) and odd hydrogen (HO$_{\rm x}$) radicals, which catalyze the destruction of ozone. The modeled production rates peak between 18 and 22~km altitude, leading to an estimated fractional ozone decrease of order $10^{-3}$--$10^{-2}$ under average GCR fluxes, consistent with observed background modulation over the solar cycle. These results establish a physically consistent link between cosmic-ray induced energy deposition and ozone chemistry, providing a benchmark framework for coupling high-energy particle transport to atmospheric photochemical models.

We investigate the conditions under which $Δ(1232)$ isobars trigger a first-order phase transition within purely hadronic neutron-star matter, using the SW4L relativistic mean-field parametrization. For scalar-vector coupling differences $0.15 \lesssim x_{σΔ} - x_{ωΔ} \lesssim 0.2$ and $x_{σΔ} \gtrsim 1.3$, the onset of $Δ^-$ resonances produces a van der Waals-like instability driven by a self-amplifying feedback in the scalar meson sector, in which the $Δ^-$ particle fraction acts as the order parameter of a Landau-type transition. A Maxwell construction yields a sharp density discontinuity at baryon densities $n_b \sim (1.3$-$2)\,n_0$, separating a $Δ$-free outer core from a $Δ$-rich inner core. The resulting neutron-star sequences satisfy all current multimessenger constraints: maximum masses $M_{\rm max} \approx 2.15$-$2.25\,M_\odot$, radii $R_{1.4} \approx 11$-$12$ km, and tidal deformabilities $Λ_{1.4} \approx 190$-$480$, compatible with NICER observations and GW170817. We compute, for the first time for a $Δ$-induced interface, the $\ell = 2$ composition $g$-mode eigenfrequencies, obtaining $ν_g \sim 400$-$1100$ Hz with gravitational-wave damping times $τ_g \sim 10^3$-$10^9$ s. These frequencies overlap quantitatively with those predicted for hadron-quark phase-transition interfaces, demonstrating that the mass-radius ``knee'', reduced tidal deformability, and $g$-mode spectrum conventionally regarded as signatures of quark deconfinement can be reproduced by a purely intrahadronic mechanism. This extends the masquerade problem from static observables to the domain of gravitational-wave asteroseismology, implying that a future detection of a discontinuity $g$-mode alone would not suffice to identify quark matter in neutron-star cores.

Nulling is a phenomenon where the emission from a pulsar becomes undetectable (or significantly weaker) for a relatively short period of time, followed by a return to a normal emission state. The timescale of nulling ranges from a few pulse periods to many hours or even days. The fraction of time a nulling pulsar spends in a null state varies across the population of canonical pulsars, from 0 to 95 per cent. The long-term behaviour of a pulsar's nulling fraction, however, is currently unknown, as published values have typically been obtained through single observations. Here, we present the first long-term analysis of nulling behaviour in eight pulsars observed in the Parkes Multibeam Pulsar Survey over the course of eight to ten years. We also apply a new Bayesian method for pulse-energy analysis, yielding posterior estimates of the nulling fraction per observation. In several cases, the nulling affects only specific components of the pulse profile, rather than the entirety of the emission. Our analysis reveals that, while most pulsars show no significant trend in their nulling fraction over time, a subset exhibit some evidence for non-zero gradients in nulling fraction. In particular, PSRs J1048$-$3832, J1745$-$3040, and J1825$-$0935 show statistically significant trends over the span of the data. Studying the behaviour of nulling over years and decades is valuable as it can provide insights into the physical emission processes within pulsars. Studying how nulling evolves also provides valuable insights into pulsar evolution and the characterisation of the broader pulsar population.

Long gamma-ray bursts (GRBs) frequently exhibit complex prompt emission structures with multiple temporally distinct episodes, such as a main emission (ME) phase followed by a weak extended emission (EE) tail. Whether these subcomponents from a common physical origin with similar classification properties, or instead represent fundamentally different emission mechanisms within a single event, remains an open question. Here, we present a systematic, pulse-resolved analysis of 22 \emph{Swift}/BAT long-duration GRBs, each exhibiting a well-separated, bright ME ($G_1$) followed by a fainter EE ($G_2$) after a background-consistent quiescent gap. For each component, we independently measure standard classification diagnostics, including duration ($T_{90}$), spectral hardness ratio (HR), minimum variability timescale (MVT), and spectral lag. We then compare these properties between the ME and EE within individual bursts. We find that the EE is systematically softer (lower HR in 19 of 22 events), smoother (longer MVT in 17 of 22 events), and more diverse in spectral lag than the ME. However, both components still occupy the long-GRB track in the traditional duration-hardness and duration-MVT planes, indicating a common Type~II (collapsar) origin. These results suggest that the EE in long GRBs represents a physically distinct regime of the central engine, characterized by a lower luminosity, longer emission timescales, and evolved spectral properties, rather than a simple continuation of the main burst. This picture is consistent with late-time fallback accretion onto a black hole or proto-magnetar spin-down.

Interactions within galaxy groups are fundamental drivers of galactic evolution, and establishing a direct observational link between the dynamical processes of satellite galaxies and the rejuvenation of massive host galaxies remains challenging. We present a multi-wavelength work of J1611+4026, a unique triple system characterised by a massive early-type host galaxy, Component C and two gas-rich companions, Components A and B, which are currently undergoing a major merger in its near environment. Utilising deep optical imaging from DESI-LS and spectroscopic data from DESI and P200, we employ 2D morphological decomposition using \textsc{GALIGHT} alongside joint spectrophotometric synthesis modelling with \textsc{BAGPIPES} and \textsc{CIGALE} to deconstruct the structural properties and star formation histories of the member galaxies. Crucially, we identify an asymmetric tidal tail extending $\sim$15.15 kpc from Component A, confirming the ongoing interaction between the companions. Although Component C appears quiescent in both morphology and spectroscopy, we reveal a subtle robust signal of ``minor rejuvenation'', characterised by significant internal dust extinction of $E(B-V) \sim 0.53$ and a UV excess. The reconstructed star formation history indicates a recent ($\sim$100 Myr) starburst that contributes a negligible fraction to the total stellar mass ($f_{\rm burst} < 0.1$ per cent). We propose that this activity is fueled by the accretion of metal-enriched gas stripped from the interacting companions. These results strongly suggest group pre-processing, where interactions between satellite galaxies drive low-level star formation in the massive host through gas transfer, providing a quantitative benchmark for interaction-driven evolution in dense environments.

Kilonovae represent key sites of r-process nucleosynthesis, making opacity estimation and spectral analysis crucial for constraining their composition. Since light r-process elements shape the early ($\sim$0.5-1.5$\mathrm{d}$) ejecta opacity, a detailed study of the selenium element with a focus on atomic data calculation, expansion opacity estimation and spectral analysis is presented. Se atomic data are calculated from Se I to Se X using the GRASP2018 code. A systematic analysis and evaluation of their precision is performed through detailed comparison with the NIST ASD, and other works available in the literature. These atomic data are then used to estimate expansion opacity at different temperatures (e.g., T=5000 K, 10000 K, 20000 K, 100000 K) and densities (e.g., $ρ= 10^{-13}\,\mathrm{g\,cm^{-3}},\,3\times10^{-12}\,\mathrm{g\,cm^{-3}}$). Spectral analysis has been performed with radiative transfer code POSSIS with a pre-computed opacity grid calculated with new densities and temperatures, ranging from -19.5 to -4.5 $\mathrm{g\,cm^{-3}}$ in log-scale and from 1 000 to 51 000 K, respectively. Two scenarios are considered: one in which the opacity contribution comes from 100\% Se ejecta, and another in which Se contributes only partially to the total opacity ($\sim$ 10\% of the total mass). Se atomic calculations show a good agreement with NIST ASD, with accurate energy levels and transitions determined alongside atomic data for higher ionisation stages not fully covered by NIST. The expansion opacities calculated with these new Se data exhibit differences in comparison to existing literature works. Se spectral features can only be observed in the KN scenario consisting of 100\% Se. When Se accounts for about 10\% of the total KN mass, these features become undetectable. All Se results are now available in the new open-source MARTINI platform dedicated to element nucleosynthesis.

The relative roles of secular disc processes and galaxy interactions in driving the growth of supermassive black hole (SMBH) remain unclear. We present a time-resolved, per galaxy analysis of SMBH mass assembly that explicitly tracks bar formation, merger events, and the environment along the main progenitor branches using the high resolution IllustrisTNG50 cosmological simulation. We analyze barred and unbarred disc galaxies in isolated and non-isolated environments using physically motivated boundary definitions. We found that SMBH fueling pathways are regulated by the environment through the timing of bar formation relative to mergers. In isolated barred galaxies, stellar bars form early in dynamically cold discs and establish sustained, coherent accretion phases that regulate late-time SMBH growth. In contrast, in non-isolated galaxies, SMBH growth is dominated by early merger-driven accretion episodes, whereas bars form later and contribute weakly to the primary growth phase. Unbarred control samples show that mergers can trigger rapid SMBH growth without bars, but such growth remains episodic, whereas isolated discs without bars lack sustained accretion. These results demonstrate an environmental bifurcation in SMBH fueling: mergers act as efficient triggers of early growth in dynamically active systems, while bars regulate prolonged accretion only when they form in quiescent discs. This study provides a unified time-domain framework linking galaxy environment, disc dynamics, and SMBH growth by resolving the temporal ordering of bars, mergers, and accretion.

Young star-forming clumps (SFCs) emit strongly in the ultraviolet (UV), making UV imaging ideal for detecting them. The Ultraviolet Imaging Telescope (UVIT) onboard AstroSat, with 1.5arcsec resolution, has enabled the characterization of recently formed (up to 300 Myr) SFCs on tens of parsec scales in nearby galaxies. The spatial distribution of SFCs with different ages can provide insights into the hierarchy of star formation. This study presents a semi-novel approach to characterize SFCs in two nearby spiral galaxies, NGC 5457 and NGC 1313, by combining UVIT FUV data with g-band data from the Dark Energy Camera Legacy Survey (DECaLS). We tested and optimized our method on NGC 5457 and showed that after proper background subtraction, the FUV-g color of SFCs can serve as an equally reliable age indicator as the widely-used FUV-NUV color. Next, we parametrized the star formation hierarchy in NGC 5457 using the two-point correlation function (TPCF) and found good agreement between the hierarchy parameters derived using FUV-NUV and FUV-g based ages. Using our FUV-g based SFC ages, we also constrained the global hierarchy parameter of NGC 1313 for the first time. The development of our FUV-g based method is motivated by the fact that the NUV channel of the UVIT is not operational, and there is a wealth of archival UVIT FUV-only observations of nearby galaxies. This work demonstrates the potential of our method in constraining the SFC ages and investigating hierarchical star formation in nearby galaxies using FUV and optical observations.

We investigate the role of contact interactions in the dynamics of fuzzy dark matter (FDM) modeled through the Schrödinger-Poisson equation in one spatial dimension. While the $Λ$CDM paradigm successfully explains structure formation on large scales, its small-scale predictions remain in tension with observations. FDM offers an alternative framework, where local self-interactions can further influence the formation and evolution of structures. We explore both attractive and repulsive contact interactions in static and expanding backgrounds. Using numerical simulations, we examine their impact on three key scenarios: the properties of the lowest-energy stationary solution, the relaxation of localized initial states, and the gravitational collapse of nonlocalized states. Our results show that contact interactions modify the density profile of the stationary solution and affect the onset of characteristic stages of gravitational collapse, particularly the shell-crossing event. In the (1+1) model, we confirm that relaxation does not converge to the lowest-energy stationary solution, even when local self-interactions are included. Taken together, local self-interactions play a relevant role in shaping the nonlinear dynamics of FDM and motivate further studies in higher-dimensional and cosmologically realistic settings.

Fast radio bursts (FRBs) are brief flashes of emission detectable to cosmological distances. Cosmology applications rely on an understanding of how the detected sample relates to the underlying population. To this end we examine the dispersion measure `tolerance' parameter employed by the FRB search tool \textsc{heimdall} and provide the relation between this and minimum search depth. Several FRB samples can be `retro-fitted' using this to more properly account for survey completeness.

The Variables and Slow Transients (VAST) Survey on the Australian SKA Pathfinder (ASKAP) is designed to systematically explore the dynamic radio sky, detecting sources that vary on timescales from minutes to several years. In this paper, we present Data Release 1 of the VAST Extragalactic Survey, which targets slowly evolving synchrotron transients in the southern sky. The observations were carried out between June 2023 and May 2025, comprising 2945 images of 276 fields spanning $\sim 12300\ \mathrm{deg}^2$, observed at 888 MHz with a typical rms sensitivity of 0.24 mJy $\rm{beam}^{-1}$ and 12-20 arcsec resolution. Each field was revisited approximately every two months, yielding 10 or 11 observations per field. The VAST pipeline extracts the light curves for all the observed sources, and additional filters are implemented to improve the reliability of the resulting light curve database. The light curve database contains 0.5 million sources and 6.4 million individual measurements, publicly available through the CSIRO data access portal. An untargeted variability search yields 117 astrophysical variables, including 27 pulsars, 40 radio stars (10 newly detected at radio wavelengths), 44 active galactic nuclei, two optically identified supernovae, one supernova candidate, one brown dwarf, and two sources without multi-wavelength counterparts that are yet to be identified. This data release provides the first large-scale, high-cadence, uniform view of long-term radio variability in the extragalactic sky and lays the groundwork for future population studies of radio transients with ASKAP.

To address the challenge of estimating redshifts when only single-band images are available, this study introduces a deep learning model named ViT-MDNz. Leveraging robust statistical priors learned from large-scale data concerning the correlation between redshift and morphology, the model can directly estimate redshifts and their associated uncertainties from single-band galaxy images. It integrates a Vision Transformer (ViT) to extract deep morphological features and a Mixture Density Network (MDN) to predict the full redshift probability density function. Trained and evaluated on approximately 300,000 single-band images from the DESI Legacy Imaging Surveys (DESI-LS), the model achieves a normalized median absolute deviation $σ_{\rm NMAD} = 0.034$ and an outlier fraction $f_{\rm out} = 2.6\%$ in the $r$-band for redshifts up to $z \lesssim 1$. Evaluations using probability integral transform (PIT) and continuous ranked probability score (CRPS) confirm that the predicted probability density functions are well calibrated and closely match the true distribution. These results demonstrate that competitive redshift estimates can be obtained using morphological features alone, and that incorporating color information further enhances the accuracy and robustness of the estimation. Therefore, ViT-MDNz provides a practical approach for redshift estimation of galaxy samples with limited photometric band coverage, contributing to improved completeness and usability of redshift catalogs for future large-scale surveys such as DESI and LSST.

Planet formation theories suggest the presence of free-floating planets (FFPs) that are ejected from their formation sites. While these planets emit very little light, they can be identified through gravitational microlensing. Here, we report the discovery of a FFP candidate in the microlensing event KMT-2024-BLG-3237. The observed light curve exhibits strong finite-source effects characterized by a small amplitude $(\lesssim 0.9\,{\rm mag})$ and a short timescale $(\lesssim 3\,{\rm days})$. The analysis yields an Einstein timescale of $t_{\rm E} = 0.54\pm0.02\,{\rm days}$ and an angular Einstein radius of $θ_{\rm E} = 6.30\pm0.48\,μ{\rm as}$. The measurements make it possible to estimate the lens mass as $M \simeq 102\,M_{\oplus}\,(π_{\rm rel}/16\,μ{\rm as})^{-1}$, where $π_{\rm rel}$ is the relative lens-source parallax. Depending on the unknown $π_{\rm rel}$, the lens could be a Neptune-mass planet $(π_{\rm rel} \simeq 0.1\,{\rm mas})$ or a Saturn-mass planet $(π_{\rm rel} \simeq 16\,μ{\rm as})$. A Bayesian analysis yields the lens mass $M = {67.3}_{-42.5}^{+103.2}\,M_{\oplus}$ and the lens distance $D_{\rm L} = {7.34}_{-2.11}^{+0.96}\,{\rm kpc}$. This lens is the eleventh isolated microlens with a measurement of $θ_{\rm E} < 10\,μ{\rm as}$. We find that additional searches for possible signatures of a lens host do not show significant evidence for the host.

Finding intermediate-mass black holes (IMBHs) and measuring their masses and spins are key to understanding massive black hole formation. White dwarf (WD)-IMBH binaries provide a unique probe because they emit both electromagnetic radiation and gravitational waves (GWs), thereby conveying richer information. However, such multi-messenger sources often enter the regime of strong gravity, where existing models fail to capture their relativistic dynamics. Here, we develop a fully relativistic model for the tidal response of a WD close to an IMBH and use it to study the secular orbital evolution as well as the GW signal. We find that for IMBHs more massive than 10^5 solar masses, tidal interaction becomes relativistic and sensitive to IMBH spin. The interaction generally dissipates binary orbital energy and angular momentum, but due to relativistic frame rotation, which reduces phase coherence across pericenter passages, the orbit-averaged tidal dissipation rate can be suppressed by up to about 50% relative to Newtonian predictions. Including tidal dissipation leads to more rapid damping of the orbital eccentricity, to the extent that the pericenter distance may even increase over time, potentially explaining quasi-periodic eruptions and secular orbital period growth. Such tidal effects accumulate into measurable phase and amplitude deviations in the GW signal. For typical space-based observations, the GW waveform mismatch can reach values of order 0.1 within 6 months. Our results indicate that relativistic tidal dissipation is both dynamically important and observationally essential for reliably predicting the multi-messenger signals of WD-IMBH systems.

This study introduces an automated approach for identifying the tip of the red giant branch (TRGB) in globular clusters, combining astronomical data with algorithmic methods. Using a dataset of 160 globular clusters and Python scripts, we matched stellar sources with Gaia data. Our script generates color-magnitude diagrams (CMDs), and uses the local outlier factor (LOF) algorithm to remove outliers. Applying a second-degree polynomial to fit red giant branch (RGB), we identify the TRGB as the star closest to the fitted curve's endpoint. By this method, we expanded TRGB samples in global clusters to 91 with newer observational data. Our results show a decreasing trend in I-band luminosity for metallicities greater than $-$1, consistent with previous studies. The results show a robust trend fitting and the $\rm M_{I}$ of TRGB is about $-$4.02 with extremely low metallicity. Our approach enhances TRGB identification efficiency while providing valuable insights for developing automatic tools in astronomical data analysis.

We report milliarcsecond-scale VLBI detections of compact radio nuclei in four spiral-hosted, double-lobed radio-loud AGNs (spiral DRAGNs), a rare class that challenges the traditional association of powerful jets with elliptical hosts. Using public VLBI data archives, we identify compact cores in four sources and resolve parsec-scale jets in two of them. The VLBI components show low brightness temperatures ($T_{\rm b} \approx 10^9$ K in the core) and jet-to-counterjet ratios consistent with only mildly relativistic intrinsic speeds ($β\lesssim 0.6$ for inclinations $θ\lesssim 80^\circ$), indicating weakly powered pc-scale outflows. The low radio-Eddington ratios $\log(L_{\rm R,1.4\,GHz}/L_{\rm Edd}) \approx -5$ to $-8$ support this interpretation. Three objects lie on the fundamental plane of black hole activity, implying that global accretion-jet coupling in spiral DRAGNs is similar to that in other AGNs. Comparison with recent GRMHD simulations of thin-disk jets suggests that the VLBI-scale cores in spiral DRAGNs may trace an early or intermittently magnetized phase of jet launching. The coexistence of weak pc-scale jets and large kpc-scale lobes implies recurrent or long-duty-cycle jet activity in these late-type hosts.

The gravitational wave $ω$ mode spectrum presents a unique window into the dense interior of neutron stars, probing physics inaccessible to electromagnetic observations. This work investigates the $ω$ modes of compact stars composed of quarkyonic matter. The quarkyonic model, which describes a cross-over transition between nucleonic and quark matter treated as quasi-particles, is formulated within the Relativistic Mean-Field (RMF) theory using the G3 and IOPB-I parameterizations. This core is surrounded by a mantle of hadronic matter, creating a multicomponent stellar interior. The overall Equation of State (EOS) is governed by two key parameters: the transition density ($n_t$), the QCD confinement scale ($Λ_{\rm cs}$), which are varied to construct models consistent with current astrophysical constraints on mass and radius. We compute the complex eigenfrequencies (damped oscillations) of the fundamental and first excited $ω$ modes using the phase-amplitude method within a full general relativistic framework. Our simulations reveal that the admixed quarkyonic structure produces a unique $ω$ mode signature, distinctly different from pure hadronic or hybrid stars. The spectrum exhibits a strong, degenerate dependence on the EOS, where the stiffening effect of the quarkyonic matter influences oscillation frequencies and damping times in a characteristic manner. We also demonstrate that $ω$ mode frequencies for quarkyonic stars follow approximate universal relations, largely independent of the EOS.

Li-rich giants serve as valuable tracers of stellar evolution and surface enrichment processes, for which a statistically large and homogeneous sample is crucial. Using the massive low-resolution ($R \sim 1800$) spectroscopic dataset from the Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) survey, we systematically search for Li-rich giants by determining their stellar lithium abundances through template matching of the Li I 6708 Å absorption line with a grid of synthetic spectra generated from ATLAS9 model atmospheres. The derived abundances are validated against previous high-resolution studies, showing good consistency with a mean absolute error of about 0.15 dex. Therefore, we adopt a threshold of $A(\rm{Li}) > 1.65$ dex to select Li-rich candidates, followed by visual inspection to ensure the reliability of each detection. Eventually, more than 20,000 Li-rich giants are identified, corresponding to approximately 2.5% of all giants in LAMOST DR9. We also investigate the occurrence rate of Li-rich giants in different evolutionary stages. This work presents a large and homogeneous catalog of Li-rich giants derived from the LAMOST low-resolution survey, which provides a reliable and valuable dataset for future studies of stellar evolution and lithium enrichment in evolved stars.

The observed lithum (Li) abundance of Galactic halo stars mainly fall within the range of 2.0--2.4 dex. This nearly constant value, known as the Spite plateau, is approximately a factor of three lower than the value predicted from cosmic microwave background measurements and standard Big Bang Nucleosynthesis (BBN) calculations. This discrepancy -- referred to as the cosmological Li problem -- is considered a potential indication of new physics or astrophysical processes. We employed models incorporating gravitational settling, diffusion, rotation, and magnetic fields to explain the Spite plateau. The rotating models predict that Li abundances in stars with ages of roughly 8--13 Gyr and effective temperatures between 6400 and 5900 K generally fall within 2.0--2.4 dex, forming a well-defined Li plateau, followed by a sharp decline in Li abundance down to about 5200 K. The Li plateau results from the combined effects of variations in convection zone depth, gravitational settling, diffusion, rotation, and magnetic fields. For red giant branch stars with $T_{\mathrm{eff}} \lesssim$ 5200 K, the rotating models predict another Li plateau with an abundance of about 1.0 dex. These results are in good agreement with observations. Moreover, the initial Li abundance of 2.72 dex adopted in the models matches the BBN prediction, implying that the Li problem arises from stellar Li depletion. Furthermore, the rotating models also reproduce the Li and Be distributions of the sample that exhibit the Spite plateau meltdown and Be deviation.

Context:Satellite dwarf galaxies provide key constraints on galaxy formation and evolution as their abundance and spatial distribution reflect both host properties and large-scale environment. Aims:This study quantifies the dependence of satellite populations on host stellar mass, morphology and star formation activity across diverse environments and traces their evolution with cosmic time within the LCDM framework. Methods:The Millennium simulation combined with the semi-analytic model is employed to construct consistent samples of host galaxies brighter than Mr < -16 and their satellites (M_* >= 3x10^5 M_sun, Mr < -9) within their virial radius. Satellite abundance and radial profiles are analyzed in cluster, group and void environments and their evolution is traced from z=2 to z=0 across three host stellar mass bins. Results:Satellite abundance correlates strongly with host stellar and bulge mass while host morphology has little additional impact once stellar mass is controlled for. Dense environments suppress satellite populations relative to voids. At z=0 radial profiles reveal strong central concentrations in voids flattened distributions in clusters and intermediate trends in groups. Their redshift evolution shows progressive flattening for low- and intermediate-mass hosts in dense environments stability for massive hosts and increasing central concentration in voids. The cosmic evolution of satellite abundance further highlights distinct pathways: gradual accumulation in voids, mass-dependent trends in groups and strong late-time suppression in clusters. Conclusions:The distribution and abundance of satellite galaxies are governed primarily by environment with host stellar mass and cosmic epoch acting as secondary modulators. Forthcoming wide-field surveys such as LSST, Euclid and the Roman Space Telescope are expected to provide stringent tests of these predictions.

It is widely believed that stellar-mass black holes (sMBHs) exist within the accretion disks of active galactic nuclei (AGN), forming a distinct population termed ``accretion-modified star" (AMS). Gas from the dense disk accretes onto these AMSs, dissipating substantial gravitational energy through a mini-disk around the sMBHs, which drives powerful outflows that interact with the surrounding disk gas. In this study, we investigate two scenarios for AMS accretion: episodic Bondi explosions with hyper-Eddington accretion (Scenario A) and steady Eddington accretion (Scenario B). These outflows generate turbulence, facilitating outward angular momentum transport in the AGN disk via shock interactions and angular momentum exchange. We explore a broad parameter space-spanning the central supermassive black hole (SMBH) mass ($M_{\rm p}$), dimensionless accretion rate ($\dot{\mathscr{M}}_{\rm p}$), sMBH mass function, and spatial distribution-to calculate the effective viscosity parameter $α_{\rm AMS}$. Our analysis reveals the scaling relations $α_{\rm AMS}\proptoζM_{\rm p}^2\dot{\mathscr{M}}_{\rm p}$ for Scenario A and $α_{\rm AMS}\proptoζM_{\rm p}^{1.5}\dot{\mathscr{M}}_{\rm p}^{0.1}$ for Scenario B, where $ζ$ denotes the ratio of total sMBH mass to the SMBH disk mass. For $ζ={0.01}$ and $M_{\rm p}=10^8 M_\odot$, $α_{\rm AMS}$ ranges from $\sim{3\times10^{-4}}$ to ${0.01}$ (Scenario A) and $\sim{3\times10^{-3}}$ to ${0.04}$ (Scenario B) from the inner to outer disk regions. These results demonstrate that AMS feedback provides an efficient mechanism for angular momentum transport in AGN disks.

The supermassive binary system, $η$ Carinae, is experiencing enormous wind-driven mass loss at a rate unparalleled in the rest of the Galaxy. Their wind-wind collision (WWC) continuously produces shock heated, X-ray emitting plasmas. The XRISM X-ray observatory observed the system in 2023 and 2024 when the X-ray emission began to increase toward periastron passage in 2025. This manuscript reports unprecedentedly high-resolution X-ray spectra in the iron K$α$ band between 6.2 and 7.1 keV, obtained with the Resolve X-ray microcalorimeter. The hydrogen-like (Ly$α$) and helium-like (He$α$) lines reveal three velocity components. Two of them are broadened with maximum velocities of 2000-3000 km/s, likely originating from the post-shock companion wind. The other is relatively narrow, with a Gaussian broadening of only ~290 km/s in 1 sigma, which may originate from the post-shock companion wind at the WWC stagnation point or penetrating the primary wind. The iron fluorescent lines exhibit a moderate blueshift and broadening with velocities at 100-200 km/s, consistent with the primary wind's velocity field. The spectra also confirm a Compton shoulder of the He$α$ line complex for the first time. Both fluorescing and scattering spectral profiles indicate that the binary system is seen from the companion side during these observations. The flux ratio of the Compton scattering emission to the fluorescent line suggests substantial hydrogen depletion of the primary wind, expected from CNO-cycled hydrogen nuclear fusion gas.

Radial gradients of cosmic rays are key parameters for understanding the transport of particles in space. Solar Orbiter, launched on 2020 February 10, approaches the Sun approximately every half year, with a closest perihelion distance of 0.29 au after the end of 2022 during the nominal mission phase. The two double-ended high energy telescopes(HET)onboard the Solar Orbiter measure energetic particles in the energy range between a few MeV/nuc and a few hundred MeV/nuc, which are dominated by anomalous cosmic rays (ACRs) and galactic cosmic rays (GCRs) during solar quiet times. By obtaining the radial gradient of the ACR helium in the inner heliosphere, we advance our understanding of how the transport of the cosmic rays is affected by the particle drift effect and the large-scale magnetic field. The helium observations at Solar Orbiter/HET between 11.1 and 49 MeV/nuc are analyzed. Since we focus on quiet time measurements, we remove the periods of solar energetic particle (SEP) events. The intensities are averaged over the Carrington rotation period. The helium observations from the Proton and Helium Instrument(EPHIN)onboard SOHO were utilized as the baseline to correct the long-term variation caused by the solar modulations. We present the first observation of ACR helium at Solar Orbiter/HET between 2020 February and 2022 July in the inner heliosphere before the sun became fully active. We derive the radial gradient of the ACR helium between 0.3 and 1 au. The averaged radial gradient between 11.1 and 49MeV/nuc is about 22$\pm$4%/au and the averaged value between 11.1 and 41.2MeV/nuc is raised to 32$\pm$8%/au after removing the GCR contribution, which is estimated by a GCR model. In addition, the temporal variation of radial gradients indicates that the gradients are increasing with the enhancement of the solar modulation and the increased tilt angle of the heliospheric current sheet.

The $γ$-ray-emitting narrow-line Seyfert-I ($γ$-NLSy1) are among the most interesting systems for studying disk-jet coupling. The soft X-ray properties of these systems suggest the presence of a disc component, which peaks in the optical/UV regime, in addition to the active jet. In this work, we investigate the optical emission from $γ$-NLSy1 using long-term Zwicky Transient Facility (ZTF) observations and discussed in the context of blazars. We have reported the long-term flux and color variability in the g- and r-bands. The fractional variability ($F_{\rm var}$) goes as high as 72\%, with a mean value of 23\%, while the amplitude of variability ($ψ$) values range from 0.24 to 3.20, which is consistent with the long-term Swift-UVOT variability studies. The color-magnitude diagrams exhibit an RWB or BWB trend similar to that of blazars. The $t_{\rm var}$ suggests an emitting region size of $10^{15-17}$ cm, aligned with emissions coming from the inner accretion disk or base of the jet. The PSD analysis using both DRW and CARMA modeling exhibits a characteristic break timescale of a few days to hundreds of days, which is likely linked to fundamental physical timescales in the system, such as thermal or viscous timescales in the accretion disk or timescales for acceleration and energy dissipation in the jet. The existence of these timescales acts as another signature of the disc-jet connection. These time scales are correlated with black hole mass, and the relation is consistent with previous studies.

We present Cepheid-based distances to two canonical AGN: NGC 4303 (M 61) and NGC 1068. Data were obtained using the Hubble Space Telescope with nonredundant time spacing over 12 visits for each target, and observations were made with the F555W and F814W filters. We found 32,694 point sources in NGC 4303, and 130 of these were determined to be strong Cepheid candidates with periods ranging ~$13-93$ days. In NGC 1068, we found 20,207 point sources, where 51 of these were strong Cepheid candidates with periods ~$14-92$ days. We fit the period$-$luminosity relationship, calibrated based on a geometric distance to the LMC by Riess et al. (2019), to our Cepheid candidates in each galaxy and correct for potential effects of metallicity. Using a distance constraint for the LMC given by Pietrzyński et al. (2019), this yields a distance modulus of $μ= 31.083 \pm 0.035$ mag for NGC 4303 and $μ= 30.150 \pm 0.106$ mag for NGC 1068. Thus, we measure distances of $D = 16.47 \pm 0.27$ Mpc to NGC 4303 and $D = 10.72 \pm 0.52$ Mpc to NGC 1068.

We analyzed millimeter-wave data toward the quasar B0218+357 observed with the Atacama Large Millimeter/submillimeter Array and obtained absorption spectra of the $J$=2-1 and $J$=3-2 rotational transitions of HCN, HCO$^{+}$, HNC, H$^{13}$CN, and H$^{13}$CO$^{+}$ at the cosmological redshift of $z\!=\!0.68$. For HCN, HCO$^{+}$, and HNC, we identified two distinct absorption components that are common to both transitions, whereas a single component was detected in the isotopologue spectra. In this paper, we accurately evaluate the excitation temperatures and their uncertainties from the absorption strengths of these components, and use them to determine the CMB temperature. Uncertainties in the continuum covering factor were propagated into the excitation temperature via Monte Carlo sampling. We further corrected the observed optical depths for biases due to column-density nonuniformity by assuming a lognormal column-density distribution. Under the assumption that the rotational levels are in radiative equilibrium with the cosmic microwave background (CMB), we derived excitation temperature profiles in the optically thin regime. Because the excitation of HCO$^+$ is biased by an additional velocity component and partial collisional excitation, this species was excluded from the final determination of the CMB temperature. From a weighted mean of the excitation temperatures obtained from HCN and HNC, we determined the CMB temperature at $z\!=\!0.68$ to be ${4.50\pm0.17\,\mathrm{K}}$. This constitutes the first measurement of the CMB temperature at $z\!=\!0.68$ based on a quasar absorption line system and represents the most precise determination at this redshift, highly consistent with the standard Big Bang cosmological model.

We test whether "Little Red Dots" (LRDs) are dust-reddened, high-inclination counterparts of compact, blue broad-line AGNs ("Little Blue Dots", LBDs) powered by super-Eddington accretion. We model the central engine as a geometrically thick, radiation-pressure supported accretion flow whose funnel yields strongly anisotropic, intrinsically blue ionizing continua, coupled to an equatorially concentrated BLR and a dusty screen with modest covering factor. Using inclination-dependent SEDs as input to Cloudy, we show that the extreme broad Halpha equivalent widths (EWs) of JWST LRDs are reproduced with global BLR covering factors of only 15%, consistent with standard Type 1 AGNs. Large Balmer EWs arise because self-shadowing suppresses the high-inclination optical continuum while the BLR is illuminated by an EUV-rich SED. Weak high-ionization lines follow from orientation-dependent suppression of the XUV/soft X-ray continuum toward equatorial directions, without requiring a fully enclosing gaseous "cocoon". With a gray attenuation law of AV = 2.8 along LRD-selected sightlines, the fiducial model matches the V-shaped UV-optical continua and large Balmer decrements; strong Balmer breaks occur only for the most obscured views. A compact equatorial dust component tied to the BLR and normalized by energy conservation intercepts and reradiates only a small fraction of Lbol, producing a modest hot-dust bump and far-IR/sub-mm emission consistent with current limits and implying small dust masses. The model unifies LRD and LBD observables via orientation, predicting correlated trends in Halpha EW, Balmer decrement, Balmer break, high-ionization line strengths, and IR emission.

We present 3D-Herschel, a new 0.3-350$μ$m photometric catalog that combines deblended Herschel far-infrared (FIR) imaging with the CANDELS/3D-HST legacy fields to probe the dust-obscured universe. Using the 17-parameter fitting code Prospector-$β$, a Bayesian inference framework, we model 41,387 galaxies spanning 0.5 $< z <$ 2.5 to measure stellar and dust properties with realistic error bars. Comparing fits with and without FIR constraints, we find that for the 3.2$\%$ of galaxies with $>3 σ$ detections in at least two Herschel bands, UV-MIR-only models (0.3-24$μ$m) recover robust stellar ages, SFRs, and stellar masses (50-70$\%$ within the median 1$σ$ error). Consequently, the star-forming sequence shows no systematic offset, with an average deviation of 0.1$\pm$0.07 dex at fixed stellar mass for FIR-detected sources at all redshifts. However, the use of rigid log-average IR templates with fixed dust emission parameters ($γ$, $U_{\mathrm{min}}$, $Q_{\mathrm{PAH}}$) in UV-MIR modeling corresponds to an unevolving MIR-to-IR luminosity ratio and cold dust temperatures. By contrast, fits that include Herschel photometry, with added freedom in the FIR, yield dust temperatures that are $\sim$7K warmer at all redshifts, with $\sim$0.2 dex higher IR-to-7.7$μ$m luminosity ratios at the low-mass end of a Herschel-detected sample (log($M_{\star}$) $\sim$9.6 $M_{\odot}$). These results demonstrate that MIR-to-IR conversions depend on stellar mass, cautioning against $L_{\mathrm{IR}}$-independent templates without FIR data. For galaxies with $<10^{11} \ M_{\odot}$ at $z>1.5$, even with state-of-the-art analysis, Herschel can at best provide upper limits due to source confusion; next-generation FIR telescopes will be essential to fully characterize dust emission in distant galaxies.

The evolutionary classification of molecular clumps, crucial for understanding star formation, is commonly based on human-assigned categories derived from infrared (IR) emission and well-established morphological criteria. However, due to ambiguous signatures, distance uncertainties or heavily obscured IR emission, a significant fraction of sources often remains unclassified. This work demonstrates the capability of machine learning (ML) as a complementary, data-driven approach to automate the identification and classification of these clumps using data from the MALT90 survey, complemented by Spitzer IR photometry. We applied unsupervised clustering with HDBSCAN on molecular line intensities, revealing distinct groupings that correspond to evolutionary stages. Using only five molecular lines (HCO$^+$, HNC, N$_2$H$^+$, HCN, C$_2$H), we identified stable clusters of protostars and regions without active star formation, driven primarily by C$_2$H and N$_2$H$^+$ emission. Incorporating H$^{13}$CO$^+$ gave rise to a distinct UV-dominant cluster, tracing more evolved regions. Infrared properties appeared as non-significant features implying that envelopes of clumps with different masses are similar in their global infrared characteristics. We then employed supervised learning to classify clumps with previously uncertain categories and provided classifications for 522 objects, predominantly as regions without active star formation. Our results show that ML techniques can effectively uncover intrinsic evolutionary structures in complex astrochemical data and assign categories to uncertain sources, providing a powerful, data-driven complement to traditional methods.

Using JWST/MIRI observations, we report the detection of CO$_2$ ice in the dusty torus of the planetary nebula NGC 6302, an environment generally considered hostile to fragile molecular species and ices due to intense UV irradiation. This detection accompanies cold (20-50 K) gas-phase CO$_2$ along the same sightlines. The ice absorption profile exhibits a double-peak profile, a characteristic of pure, crystalline CO$_2$ ice. The CO$_2$ gas-to-ice ratio is more than an order of magnitude higher than in young stellar objects, pointing to distinct ice formation or processing mechanisms in evolved stellar environments. This discovery demonstrates that the dusty torus provides sufficient shielding to harbour ice chemistry, and that ice-mediated surface reactions must be incorporated into chemical models of planetary nebulae.

Non-magnetic Cataclysmic Variables (CVs) show two distinct X-ray components: a hard, optically thin component and a soft, optically thick, blackbody-like component, both produced in the boundary layer between the accretion disk and the White Dwarf (WD). An additional soft component originating from a more extended region has been reported in few CVs. In a short Chandra exposure, we identified a tentative X-ray eclipse in UU Aqr, a non-magnetic CV which shows deep optical eclipses. Using observations with the Nuclear Spectroscopic Telescope Array (NuSTAR) and the XMM-Newton, we detect total eclipses in the orbital intensity profiles of this system in the hard X-ray band (3-10 keV with XMM and 3-25 keV with NuSTAR). However, the soft X-ray band (0.3-2.0 keV) shows no evidence of an eclipse. Detailed eclipse modeling, energy-resolved power spectral analysis and broadband spectral modeling indicate that the hard absorbed X-ray emission originates from a compact region near the WD, such as a boundary layer, while the soft, unabsorbed and un-eclipsed X-ray emission originates in an extended region. Neither scattering of hard X-rays nor colliding winds can account for the observed un-eclipsed soft emission. We instead propose that this component is produced by shocks within vertically extended, radiatively driven accretion-disk winds. We also provide new estimates on the emitting regions, mass and radius of the WD and the donor star using eclipse modeling.

The chemical abundances of a stellar population encode information about nucleosynthesis and its astrophysical sites, but this information is confounded by the specific star formation history of the host galaxy. As a result, placing empirical constraints on supernova yields and timing using abundances has been very challenging. We introduce a galactic chemical evolution model DLEIY that uses an observed star formation history and metallicity distribution to reduce these confounding factors. Using a joint statistical model of the dwarf spheroidal galaxies Sculptor and Fornax, simultaneous constraints on population-averaged yields and galactic outflows are achieved with DLEIY, without fixing the absolute scale of nucleosynthetic yields. The Fe yield from core collapse supernovae is consistent with existing theoretical yield models, while the measured Mg yield is a factor of 2-4 higher, corroborating previous suggestions that yield models may under-predict [Mg/Fe]. We also find that the rate of Type Ia supernovae is enhanced by about a factor of 5 relative to field galaxies, and the delay-time distribution goes as $\sim t^{-2}$, a much steeper relationship than that measured from supernova surveys ($\sim t^{-1.1}$). These findings may suggest a metallicity dependence of the Type Ia rate and delay-time distribution.

Large scale phenomena in spiral galaxies such as shear, supernovae, and magnetic fields all contribute to the formation and subsequent evolution of filamentary structure and star formation within them. In this paper, we analyze the properties and dynamics of filaments in a simulated Milky Way-like galaxy from Zhao et al. 2024. Using filament and superbubble structure analysis codes, we investigate the roles of galactic shear, supernovae and superbubbles, and magnetic fields on the stability and fragmentation of filaments. We find that local shear has little effect on filament stability and the largest structures at outer radii of the disk may be more likely to be dissipated by shear than supernovae. Filaments are largely parallel to the magnetic field, which plays a significant role in filament stability. By measuring the ratio of surface pressure on a filament to that on its central spine, $χ_f=P_{surf}/P_{central}$, we find that filaments with $χ_f \le 1$ are dominated by their own self gravity and have a strong tendency to be gravitationally supercritical, whereas those with $χ_f > 1$ are either transitory or in the act of being formed. Finally, we investigate the role of ISM pressure on filament dynamics and stability as a function of galactic radius, finding considerable changes in filament stability and the accompanying star formation rates in the inner versus outer regions of the disk.

We present the first analysis of the X-ray Imaging and Spectroscopy Mission (XRISM) observation of the supernova remnant (SNR) N103B. We fit the X-ray spectrum taken with the Resolve microcalorimeter, which captured emission lines from the predominantly ejecta elements Si, S, Ar, Ca, Cr, Mn, and Fe. Notably, our fits require a previously unidentified high-temperature, highly-ionized, Fe-dominated plasma component with particularly high Cr and Mn abundances, matching a feature also present in the recent XRISM analysis of the SNR N132D. We find that all ejecta in N103B exhibits significant line broadening arising mostly from thermal Doppler broadening: increasing from $σ_{\rm th}\sim1700$ km s$^{-1}$ for intermediate-mass element (IME: Si, S, Ar, and Ca) ejecta to $\sim$2800 km s$^{-1}$ for Fe-rich ejecta. These velocities correspond to reverse shock velocities of $\sim$3500 and $\sim$5900 km s$^{-1}$, respectively, in the ejecta frame of rest. Finally, we find that the IMEs are redshifted with a bulk velocity of $\sim$360 km s$^{-1}$ while the Fe-dominated components are split: one redshifted at $\sim$1560 km s$^{-1}$ and the other blueshifted at $\sim$1020 km s$^{-1}$. Our results provide further support for the double-ring structure of N103B as it expands into the bipolar winds of a non-degenerate companion and highlight the strength of high-resolution spectroscopic observations of SNRs.

We present the discovery and subsequent 3.2 year monitoring campaign of the repeating fast radio burst FRB 20220529A with CHIME/FRB. We observe a gradual dispersion measure (DM) decline of $-0.881\pm0.001~\mathrm{pc}~\mathrm{cm}^{-3}~\mathrm{year}^{-1}$ ($-1.235\pm0.001~\mathrm{pc}~\mathrm{cm}^{-3}~\mathrm{year}^{-1}$ in the rest frame), implying a $\geq3.5\pm0.2$% decrease of the total electron column in the source environment, and we see scattering timescale variations over weeks to years. We observe a short-lived excursion in which the DM rises by $\sim 1~\mathrm{pc}~\mathrm{cm}^{-3}$, immediately preceding a transient $\sim 2000~\mathrm{rad}~\mathrm{m}^{-2}$ Faraday rotation measure (RM) increase previously reported for this source, before returning to its gradual DM decline. We identify a local line-of-sight magnetic field around FRB 20220529A during this DM/RM excursion of $3.4 \pm 0.2~\mathrm{mG}$, corresponding to one of the most strongly magnetized FRB environments. We measure a decrease in the linear polarization fraction of FRB 20220529A bursts with decreasing frequency that we attribute to depolarization from multi-path propagation in the source environment. We also place a $5σ$ upper limit on the spectral luminosity of an associated persistent radio source of $\leq 5\times10^{28}~\mathrm{erg}~\mathrm{s}^{-1}~\mathrm{Hz}^{-1}$ at 1.5 GHz. These observations are consistent with FRB 20220529A originating from a young ($\sim$ years to centuries old) expanding supernova remnant, with short-lived DM and RM variability arising from interactions with the supernova remnant or with a binary companion.

This study provides a detailed examination of the structural, astrophysical, kinematic, and dynamical properties of the open clusters COIN-Gaia 24, Czernik 24, FSR 0893, and UBC 74, which are located in the direction opposite to the Galactic center. Astrometric, photometric, and spectroscopic data from the Gaia Data Release 3 catalog were used to ensure a precise characterization of cluster members and their physical properties. Membership determination was performed using the UPMASK algorithm applied to a five-dimensional parameter space, yielding 116, 179, 238, and 387 likely members for each cluster, respectively. Structural parameters were derived by fitting King profiles to the radial density distributions of high-probability members. Astrophysical parameters were estimated through Bayesian Markov Chain Monte Carlo isochrone fitting based on PARSEC evolutionary models, complemented by spectral energy distribution analysis using ARIADNE. The resulting extinctions, distances, metallicities, and ages indicate that these are moderately reddened, intermediate-age clusters located between 1 and 3.5 kpc from the Sun. Mean radial velocities combined with Galactic orbital integrations computed with galpy show that all four clusters follow nearly circular, low-eccentricity orbits typical of thin-disc populations. All four OCs have dynamical relaxation times of 22-98 Myr. However, their ages exceed these times by several factors, particularly in Czernik 24, indicating that they are dynamically evolved systems, even though the calculated T_E values represent lower limits. The results confirm that these OCs serve as reliable tracers of the chemical and dynamical evolution of the Galactic thin disc.

We identify a fundamental pathology in the likelihood for time delay inference which challenges standard inference methods. By analysing the likelihood for time delay inference with Gaussian process light curve models, we show that it generically develops a boundary-driven "W"-shape with a global maximum at the true delay and gradual rises towards the edges of the observation window. This arises because time delay estimation is intrinsically extrapolative. In practice, global samplers such as nested sampling are steered towards spurious edge modes unless strict convergence criteria are adopted. We demonstrate this with simulations and show that the effect strengthens with higher data density over a fixed time span. To ensure convergence, we provide concrete guidance, notably increasing the number of live points. Further, we show that methods implicitly favouring small delays, for example optimisers and local MCMC, induce a bias towards larger $H_0$. Our results clarify failure modes and offer practical remedies for robust fully Bayesian time delay inference.

We present measurements of morphological parameters from fitting 53,885 galaxies detected to a magnitude limit of F356W$< 28.5$ in the CEERS NIRCam imaging with galfit in six broadband filters: F115W, F150W, F200W, F277W, F356W, and F444W. We provide a public catalog of Sérsic index, effective semi-major axis, axis ratio, integrated magnitude, and position angle for these galaxies in each of the filters. Uncertainties in the measured parameters are estimated from simulated galaxies that have similar noise and background properties as the observed galaxies. We compare our measurements with those in the CANDELS/EGS field measured with HST/WFC3 and find that the sizes agree to within 0.09 dex and the Sérsic indices agree to within 0.13 dex. We further present the evolution in the size-mass relation, and find that the evolution to $z\sim9$ is consistent with previous results derived at lower redshift. Finally, we look at the color gradients of galaxies at $1<z<5$ and find that for late-type galaxies ($n<2.5$), there is a strong dependence on mass, but no apparent evolution with redshift, indicating that the stellar populations and dust attenuation in more massive galaxies vary substantially with radius and contribute to significant morphological $k-$corrections. For early type galaxies ($n>2.5$), the color gradients are nearly flat with no dependence on mass, indicating that the stellar populations are more uniform throughout. The structural measurements presented are accurate to $20\%$ or better for most galaxies with F356W $<27.0$ mag and will enable further studies of galaxy morphology to $z\sim10$.

We revisit black hole mass estimates for high-redshift broad-line AGNs discovered with JWST by jointly analysing their broad emission lines and their systematic non-detections in deep Chandra imaging. Building upon the self-shadowed, super-Eddington accretion flow framework and the coronal over-cooling prescription of Madau & Haardt (2024), we couple funnel-dependent Comptonization physics with slim-disc spectra from Kubota & Done (2019) and explore the resulting parameter space through a full MCMC inference. Using the sample analysed by Lupi et al. (2024) and Maiolino et al. (2025), we show that X-ray weakness - manifested as extreme bolometric corrections, suppressed 2-10 keV luminosities, and non-detections in the 0.5-5 keV Chandra band - naturally arises when the corona is confined and radiatively over-cooled inside a narrow super-Eddington funnel. The combined broad line+X-ray analysis yields strongly bimodal posteriors: either very massive, very low-Eddington black holes (physically disfavoured), or a population of low-mass ($\sim 10^{6}-10^{7} ~M_{\odot}$) black holes accreting at $f_{\rm Edd} \gg 1$. The latter solution is strongly preferred for nearly all objects and returns masses consistent with, or lower than, local $M_{\rm BH}-M_{\rm star}$ relations, mitigating the extreme mass ratios implied by single-epoch virial estimators. The predicted intrinsic spectra are redder and exhibit reduced hard-X-ray output but higher bolometric luminosities, implying bolometric corrections larger than those typical of the local AGN population, yet consistent with low-redshift highly accreting counterparts. These results support a picture in which many JWST broad-line AGNs are powered by rapidly growing, super-Eddington black holes whose suppressed coronal emission and self-shadowed BLR geometry combine to mimic overmassive black holes at $z \gtrsim 6$.

The current Fermi-LAT source catalog (4FGL-DR4: 7194 sources over 14 years) was built incrementally from the 8-year catalog. In a survey mission like Fermi, data accumulate on each source over time, so after 16 years (reached in August 2024) and twice the data for the original 4FGL sources we have more precise localization (by 24% on average). It is thus time to generate a new original catalog, which implies, beyond adding the sources newly detectable after two more years, changing the existing source names (derived from their coordinates) and reviewing the associations. We present an early 16-year list (FL16Y) of 7220 sources, which relocalizes all sources and improves a few aspects of the catalog analysis, but still uses the same model of interstellar diffuse emission as 4FGL-DR4.

In this work, we explore constraints on the emergence and longevity of technologically intelligent civilizations in our Galaxy, considering the Fermi paradox. We argue that under optimistic assumptions about the probability of life and intelligence emerging on Earth-like planets, the absence of contact with extraterrestrial civilizations imposes limits on their lifespan. Our analysis suggests that if intelligent life is common, technological civilizations must be relatively short-lived, with lifetimes constrained to $\lesssim 5\times10^3$ years under our most optimistic scenario. Considering electromagnetic communication, we note that our current light cone encompasses the entire Galactic history over the past $\sim 10^5$ years, making the lack of detected signals particularly puzzling for long-lived civilizations. We emphasize that these results should be interpreted as upper bounds derived from the Fermi paradox, not as predictions of actual lifespans.