Despite the success of reinforcement learning for large language models, a common failure mode is reduced sampling diversity, where the policy repeatedly generates similar erroneous behaviors. Classical entropy regularization encourages randomness under the current policy, but does not explicitly discourage recurrent failure patterns across rollouts. We propose MEDS, a Memory-Enhanced Dynamic reward Shaping framework that incorporates historical behavioral signals into reward design. By storing and leveraging intermediate model representations, we capture features of past rollouts and use density-based clustering to identify frequently recurring error patterns. Rollouts assigned to more prevalent error clusters are penalized more heavily, encouraging broader exploration while reducing repeated mistakes. Across five datasets and three base models, MEDS consistently improves average performance over existing baselines, achieving gains of up to 4.13 pass@1 points and 4.37 pass@128 points. Additional analyses using both LLM-based annotations and quantitative diversity metrics show that MEDS increases behavioral diversity during sampling.

Explaining why dense retrievers assign high relevance scores remains challenging because retrieval decisions are made through opaque high-dimensional embeddings. Existing explanations often focus on surface signals, such as lexical matches, token alignments, or post-hoc textual rationales, and thus provide limited insight into the latent factors that shape dense retrieval behavior at the embedding level. We propose Xetrieval, an embedding-level mechanistic framework for explaining dense retrieval. Xetrieval first introduces a lightweight reasoning internalizer that approximates Chain-of-Thought reasoning directly in the embedding space with a single forward pass, enriching sentence embeddings with reasoning-oriented information while avoiding expensive autoregressive generation. It then decomposes these reasoning-enhanced embeddings into sparse, human-interpretable features, each associated with a coherent natural language description. By aggregating sparse feature overlaps across multiple document-side views, Xetrieval provides feature-level explanations of individual retrieval decisions. Experiments on diverse retrievers and benchmarks show that Xetrieval uncovers coherent interpretable features, yields stronger pair-level intervention effects, and supports task-level feature steering. The project page and source code are available at https://hihiczx.github.io/Xetrieval .

Today's large language models (LLMs) are capable of supporting multilingual scenarios, allowing users to interact with LLMs in their native languages. When LLMs respond to subjective questions posed by users, they are expected to align with the views of specific demographic groups or historical periods, shaped by the language in which the user interacts with the model. Existing studies mainly focus on researching the opinions represented by LLMs among demographic groups in the United States or a few countries, lacking worldwide country samples and studies on human opinions in different historical periods, as well as lacking discussion on using language to steer LLMs. Moreover, they also overlook the potential influence of prompt language on the alignment of LLMs' opinions. In this study, our goal is to fill these gaps. To this end, we create an evaluation framework based on the World Values Survey (WVS) to systematically assess the alignment of LLMs with human opinions across different countries, languages, and historical periods around the world. We find that LLMs appropriately or over-align the opinions with only a few countries while under-aligning the opinions with most countries. Furthermore, changing the language of the prompt to match the language used in the questionnaire can effectively steer LLMs to align with the opinions of the corresponding country more effectively than existing steering methods. At the same time, LLMs are more aligned with the opinions of the contemporary population. To our knowledge, our study is the first comprehensive investigation of the topic of opinion alignment in LLMs across global, language, and temporal dimensions.

Spoken Language Models (SLMs) are increasingly central to modern speech-driven applications, but performance degrades under acoustic shift - real-world noise, reverberation, and microphone variation. Prior solutions rely on offline domain adaptation, which is post-hoc, data-intensive, and slow. We introduce the first test-time adaptation (TTA) framework for generative SLMs that process interleaved audio-text prompts. Our method updates a small, targeted subset of parameters during inference using only the incoming utterance, requiring no source data or labels. This stabilizes token distributions and improves robustness to acoustic variability without degrading core task accuracy. Evaluated on automatic speech recognition, speech translation, and 19 audio understanding tasks from AIR-Bench, our approach yields consistent gains under diverse corruptions. Because adaptation touches only a small fraction of weights, it is both compute- and memory-efficient, supporting deployment on resource-constrained platforms. This work enhances the robustness and adaptability of generative SLMs for real-world speech-driven applications.

Memory is essential as language agents move from isolated tasks to long-horizon, stateful workflows, yet existing evaluations often reduce it to retrieval or question answering. We introduce ContextWeave, a longitudinal benchmark that evaluates whether recalled experience improves downstream agent performance in realistic office-work streams. ContextWeave reconstructs privacy-preserved, multi-month workflows of 14 participants into 1,005 executable tasks, including 568 core evaluation tasks, with instructions, containerized environments, trajectories, and task-specific rubrics. It measures workspace quality and alignment with participant-specific preferences, complemented by diagnostics of relevance, continuity, solvability, and robustness to misleading recall. Across six memory components under a fixed model, the strongest configuration raises Workspace Score from 68.08 to 78.20 and Preference Score from 41.50 to 70.60. With a fixed memory component, recall improves both outcomes for all five tested base models, although gains vary substantially. Our analysis shows that actionable, experience-rich memory supports workflow continuation and reduces redundant exploration more effectively than compact summaries, while it can also be more susceptible to misleading recall. These findings motivate memory systems that optimize not only retrieval relevance but also reliable use during execution.

As real-world tasks grow increasingly complex, long-context reasoning has become a core capability for Large Language Models (LLMs). However, few studies explore which data types are effective for long-context reasoning and why. We find that structured table data with periodic structures shows strong potential for long-context reasoning. Motivated by this observation, we mathematically analyze tabular dependency structures using mutual information, revealing periodic non-vanishing dependencies in table data. Furthermore, we systematically analyze the capabilities of structured table data, conduct relevant scaling experiments, and validate its underlying mechanisms for enhancing long-context reasoning, yielding several meaningful insights. Leveraging these insights, we propose a simple yet scalable pipeline(TableLong) for synthesizing high-quality, diverse, and verifiable structured table data to boost long-context reasoning via RL. Extensive experimental results demonstrate that table data significantly enhances the long-context reasoning capability of LLMs across multiple long-context benchmarks (+8.24\% on average), and even improves performance on out-of-domain benchmarks (+8.06\% on average). We hope that our insights provide practical guidance for effective post-training data to enhance long-context reasoning in LLMs.

Warp knitting machines operate under high-speed and continuous working conditions, where minor abnormalities in yarn tension or vibration may quickly evolve into severe defects. Conventional vision-based inspection methods usually detect faults only after visible defects appear, resulting in substantial latency and limited capability for early warning and predictive maintenance. To address this issue, we propose KnitLLM, a multimodal semantic learning framework for early fault diagnosis in warp knitting machines, which integrates sensor sequences and fabric images for proactive anomaly detection. KnitLLM first employs a vector-quantized variational autoencoder to map heterogeneous temporal signals and image features into a unified discrete semantic space, enabling aligned fault primitive representation across modalities. A Transformer-based masked fault modeling strategy is then introduced to learn long-range dependencies and structural priors of the knitting process by reconstructing masked process states. In addition, a diffusion model is used to augment rare fault samples, alleviating class imbalance and improving recognition performance for minority fault categories. Theoretical analysis from the perspectives of manifold representation, quantization error, and attention structure further supports the semantic abstraction and information preservation capabilities of the proposed framework. Experiments on a real-world industrial dataset containing 10,000 h of warp knitting production data demonstrate that KnitLLM consistently outperforms competitive baselines across multiple metrics, achieving an F1-score of 0.961 and a mean detection latency of 124 ms. The results indicate that the proposed method can identify early faults before visible defects emerge, offering a practical solution for intelligent maintenance and fault warning in high-speed textile manufacturing.

Improving reasoning LLMs requires the ability to judge the quality of long reasoning traces for effective reasoning data curation, strong training signals during reinforcement learning, and an in-depth understanding of reasoning behaviors during model performance evaluation. Additionally, surfacing reasoning mistakes that the model makes would enable improving the model's performance at runtime through providing feedback. Due to the difficulty of this complex task on long reasoning traces, single-model judges (even frontier models) do not do well at identifying reasoning defects. Additionally, leveraging frontier models during online training of reasoning LLMs is generally prohibited due to guardrails in terms of use. In this work, we introduce Reasoning Jury, a system that replaces the single judge with a jury of LLMs and a moderated consensus mechanism, to improve the fidelity of judgments for identifying reasoning defects. In reasoning jury, defects of a reasoning trace and their severity are surfaced through a deliberation where a moderator conducts a discussion amongst the jury where the jurors critique each other's judgments and get to modify their initial votes. The moderator derives a consensus through deliberation amongst jurors or consolidation of judgements. We show that Reasoning Jury with a jury of open-weight models (e.g., gpt-oss-120b) is able to significantly outperform frontier models (opus-4.6, sonnet-4.6, and gemini-3.1-pro) at correctly identifying reasoning defects. Besides accuracy performance improvements, the aggregated cost of the jury (initial verdicts, deliberations, consolidation, etc.) is a fraction (8 to 15

Video anomaly detection (VAD) has been extensively researched due to its potential for intelligent video systems. However, most existing methods based on CNNs and transformers still suffer from substantial computational burdens and have room for improvement in learning spatial-temporal normality. Recently, Mamba has shown great potential for modeling long-range dependencies with linear complexity, providing an effective solution to the above dilemma. To this end, we propose a lightweight and effective Mamba-based network named STNMamba, which incorporates carefully designed Mamba modules to enhance the learning of spatial-temporal normality. Firstly, we develop a dual-encoder architecture, where the spatial encoder equipped with Multi-Scale Vision Space State Blocks (MS-VSSB) extracts multi-scale appearance features, and the temporal encoder employs Channel-Aware Vision Space State Blocks (CA-VSSB) to capture significant motion patterns. Secondly, a Spatial-Temporal Interaction Module (STIM) is introduced to integrate spatial and temporal information across multiple levels, enabling effective modeling of intrinsic spatial-temporal consistency. Within this module, the Spatial-Temporal Fusion Block (STFB) is proposed to fuse the spatial and temporal features into a unified feature space, and the memory bank is utilized to store spatial-temporal prototypes of normal patterns, restricting the model's ability to represent anomalies. Extensive experiments on three benchmark datasets demonstrate that our STNMamba achieves competitive performance with fewer parameters and lower computational costs than existing methods.

Visuo-Tactile policies leveraging optical tactile sensors have shown great promise in contact-rich manipulation. These sensors achieve high spatial resolution and multi-dimensional force sensing by utilizing an internal camera to monitor the deformation of their elastic gel surface, thereby indirectly inferring tactile cues. Despite their advantages, extracting fine-grained contact states necessary for contact-rich manipulation remains an open challenge. Existing methods typically use either raw images or cumulative motion fields to represent tactile cues. However, both are prone to perception ambiguity. Raw tactile images mainly capture appearance changes, while cumulative motion fields only reflect the aggregate gel deformation. Consequently, distinct fine-grained contact states can exhibit highly similar patterns, making it difficult to explicitly distinguish subtle contact variations. To address this issue, we explore the dynamic priors of tactile motion and discover that the correlation between transient and cumulative motion can explicitly distinguish fine-grained contact states. Based on this insight, we propose a motion-aware tactile representation to facilitate contact-rich manipulation. Beyond tactile representation, effective fusion of tactile and visual modalities is also critical. Most existing fusion methods either directly concatenate features from each modality or train modality-specific networks separately and fuse their outputs. However, these strategies struggle to simultaneously model cross-modal interactions and preserve modality-specific characteristics. In this work, we take advantage of the Mixture-of-Transformers architecture and propose a unified modality-aware visuo-tactile policy that captures cross-modal complementarity while maintaining modality-specific properties.

As language models (LMs) evolve from chat assistants to long-horizon agents capable of multi-step reasoning and tool use, existing benchmarks remain largely confined to structured or exam-style tasks that fall short of real-world professional demands. To this end, we introduce $OneMillion-Bench $OneMillion-Bench, a benchmark of 400 expert-curated tasks spanning Law, Finance, Industry, Healthcare, and Natural Science, built to evaluate agents across economically consequential scenarios. Unlike prior work, the benchmark requires retrieving authoritative sources, resolving conflicting evidence, applying domain-specific rules, and making constraint decisions, where correctness depends as much on the reasoning process as the final answer. We adopt a rubric-based evaluation protocol scoring factual accuracy, logical coherence, practical feasibility, and professional compliance, focused on expert-level problems to ensure meaningful differentiation across agents. Together, $OneMillion-Bench provides a unified testbed for assessing agentic reliability, professional depth, and practical readiness in domain-intensive scenarios.

Zero-shot object navigation requires agents to locate unseen target objects in unfamiliar environments without prior maps or task-specific training which remains a significant challenge. Although recent advancements in vision-language models(VLMs) provide promising commonsense reasoning capabilities for this task, these models still suffer from spatial hallucinations, local exploration deadlocks, and a disconnect between high-level semantic intent and low-level control. In this regard, we propose a novel hierarchical navigation framework named ReMemNav, which seamlessly integrates panoramic semantic priors and episodic memory with VLMs. We introduce the Recognize Anything Model to anchor the spatial reasoning process of the VLM. We also design an adaptive dual-modal rethinking mechanism based on an episodic semantic buffer queue. The proposed mechanism actively verifies target visibility and corrects decisions using historical memory to prevent deadlocks. For low-level action execution, ReMemNav extracts a sequence of feasible actions using depth masks, allowing the VLM to select the optimal action for mapping into actual spatial movement. Extensive evaluations on HM3D and MP3D demonstrate that ReMemNav outperforms existing training-free zero-shot baselines in both success rate and exploration efficiency. Specifically, we achieve significant absolute performance improvements, with SR and SPL increasing by 1.7

Today's AI assistants such as OpenClaw are designed to handle context effectively, making context learning an increasingly important capability for models. As these systems move beyond professional settings into everyday life, the nature of the contexts they must handle also shifts. Real-life contexts are often messy, fragmented, and deeply tied to personal and social experience, such as multi-party conversations, personal archives, and behavioral traces. Yet it remains unclear whether current frontier language models can reliably learn from such contexts and solve tasks grounded in them. To this end, we introduce CL-bench Life, a fully human-curated benchmark comprising 405 context-task pairs and 5,348 verification rubrics, covering common real-life scenarios. Solving tasks in CL-bench Life requires models to reason over complex, messy real-life contexts, calling for strong real-life context learning abilities that go far beyond those evaluated in existing benchmarks. We evaluate ten frontier LMs and find that real-life context learning remains highly challenging: even the best-performing model achieves only 19.3

Silent Speech Interface (SSI) systems typically rely on non-streaming inference, where entire utterances must be processed at once, leading to substantial latency, high memory usage, and limited suitability for real-time interaction or edge deployment. To address this limitation, we propose DualSSI, a dual-mode sEMG-based SSI that supports both streaming and non-streaming inference within a single unified architecture. DualSSI integrates three key techniques: (1) adaptation of dynamic chunk convolution and attention, originally proposed in streaming automatic speech recognition, to exploit limited future context during streaming sEMG-to-speech synthesis; (2) feature distillation from a non-streaming teacher model to improve streaming accuracy; and (3) a Conformer-based encoder-decoder for robust local-global feature extraction under chunk-wise processing. Experiments on Mandarin and English datasets show that DualSSI consistently improves over a strong streaming SSI baseline in both objective metrics and subjective listening tests, while approaching the performance of non-streaming models. Moreover, DualSSI achieves an end-to-end latency of 10.71 ms for 11.6 ms chunks on an Ascend 310B NPU, indicating strong potential for real-time and wearable SSI applications.

Code cloning, a widespread practice in software development, involves replicating code fragments to save time but often at the expense of software maintainability and quality. In this paper, we address the specific challenge of detecting "essence clones", a complex subtype of Type-3 clones characterized by sharing critical logic despite different peripheral codes. Traditional techniques often fail to detect essence clones due to their syntactic focus. To overcome this limitation, we introduce ECScan, a novel detection tool that leverages information theory to assess the semantic importance of code lines. By assigning weights to each line based on its information content, ECScan emphasizes core logic over peripheral code differences. Our comprehensive evaluation across various real-world projects shows that ECScan significantly outperforms existing tools in detecting essence clones, achieving an average F1-score of 85 offers exceptional scalability. This study advances clone detection by providing a practical tool for developers to enhance code quality and reduce maintenance burdens, emphasizing the semantic aspects of code through an innovative information-theoretic approach.

The standard Reinforcement Learning from Human Feedback (RLHF) framework primarily focuses on optimizing the performance of large language models using pre-collected prompts. However, collecting prompts that provide comprehensive coverage is both tedious and challenging, and often fails to include scenarios that LLMs need to improve on the most. In this paper, we investigate alignment through the lens of two-agent games, involving iterative interactions between an adversarial and a defensive agent. The adversarial agent's task at each step is to generate prompts that expose the weakness of the defensive agent. In return, the defensive agent seeks to improve its responses to these newly identified prompts it struggled with, based on feedback from the reward model. We theoretically demonstrate that this iterative reinforcement learning optimization converges to a Nash Equilibrium for the game induced by the agents. Experimental results in safety scenarios demonstrate that learning in such a competitive environment not only fully trains agents but also leads to policies with enhanced generalization capabilities for both adversarial and defensive agents.

Large language models (LLMs) possess impressive linguistic capabilities but often fail to faithfully retain factual knowledge, leading to hallucinations and unreliable outputs. Understanding LLMs' knowledge deficiencies by exhaustively evaluating against full-scale knowledge bases is computationally prohibitive, especially for closed-weight models. We propose stochastic error ascent (SEA), a scalable and efficient framework for discovering knowledge deficiencies (errors) in closed-weight LLMs under a strict query budget. Rather than naively probing all knowledge candidates, SEA formulates error discovery as a stochastic optimization process: it iteratively retrieves new high-error candidates by leveraging the semantic similarity to previously observed failures. To further enhance search efficiency and coverage, SEA employs hierarchical retrieval across document and paragraph levels, and constructs a relation directed acyclic graph to model error propagation and identify systematic failure modes. Empirically, SEA uncovers 40.7x more knowledge errors than Automated Capability Discovery and 26.7 reducing the cost-per-error by 599x and 9x, respectively. Human evaluation confirms the high quality of generated questions, while ablation and convergence analyses validate the contribution of each component in SEA. Further analysis on the discovered errors reveals correlated failure patterns across LLM families and recurring deficits, highlighting the need for better data coverage and targeted fine-tuning in future LLM development.

Large language models (LLMs) still struggle with strategic persuasion, largely because existing approaches either neglect information asymmetry or rely on unrealistic pre-commitment assumptions. We introduce a type-induced commitment-communication mechanism that grounds Bayesian Persuasion (BP) in natural language dialogue without pre-commitment: the persuader narrates their potential types (e.g., honest vs. dishonest) to dynamically construct an information schema, enabling the persuadee to perform Bayesian belief updates within the conversation itself. We implement two variants: Semi-Formal-Natural-Language (SFNL) and Fully-Natural-Language (FNL), evaluating them against strong baselines across multiple LLMs and human judges. BP strategies consistently outperform baselines: SFNL excels in logical credibility, while FNL shows superior robustness and emotional resonance. We verify that gains stem from genuine Bayesian reasoning rather than superficial formatting, and we further show that supervised fine-tuning enables small models to match the persuasive performance of much larger ones.

Foundation models have achieved transformative success across biomedical domains by enabling holistic understanding of multimodal data. However, their application in surgery remains underexplored. Surgical intelligence presents unique challenges - requiring surgical visual perception, temporal analysis, and reasoning. Existing general-purpose vision-language models fail to address these needs due to insufficient domain-specific supervision and the lack of a large-scale high-quality surgical database. To bridge this gap, we propose SurgVLM, one of the first large vision-language foundation models for surgical intelligence, where this single universal model can tackle versatile surgical tasks. To enable this, we construct a large-scale multimodal surgical database, SurgVLM-DB, comprising over 1.81 million frames with 7.79 million conversations, spanning more than 16 surgical types and 18 anatomical structures. We unify and reorganize 23 public datasets across 10 surgical tasks, followed by standardizing labels and doing hierarchical vision-language alignment to facilitate comprehensive coverage of gradually finer-grained surgical tasks, from visual perception, temporal analysis, to high-level reasoning. Building upon this comprehensive dataset, we propose SurgVLM, which is built upon Qwen2.5-VL, and undergoes instruction tuning to 10+ surgical tasks. We further construct a surgical multimodal benchmark, SurgVLM-Bench, for method evaluation. SurgVLM-Bench consists of 6 popular and widely-used datasets in surgical domain, covering several crucial downstream tasks. Based on SurgVLM-Bench, we evaluate the performance of our SurgVLM (3 SurgVLM variants: SurgVLM-7B, SurgVLM-32B, and SurgVLM-72B), and conduct comprehensive comparisons with 14 mainstream commercial VLMs (e.g., GPT-4o, Gemini 2.0 Flash, Qwen2.5-Max).