论文检索

Mengting Zhang, Gaofeng Pan, Zhixiong Zhang, Yang Li, Guangyin Zhang

Scientific opinion classification based on discourse functions provides a structured semantic basis for analytical tasks such as gap identification and hypothesis generation. However, this task is uniquely challenged by the multi-label nature of scientific expressions and AIMRaD structural constraints. Existing LLM-based methods typically rely on direct label generation, which obscures decision logic, or treat discourse information as passive context rather than a structural prior. We propose OPINE, a multi-stage framework that reformulates classification as a controllable *scoring-calibration-refinement* pipeline. By decoupling textual evidence from decision logic, OPINE generates independent label-wise affinity scores calibrated by AIMRaD priors. To resolve the multi-label challenge, we introduce a quantile-based decoding rule to naturally capture co-existing roles, alongside a pairwise refinement mechanism to mitigate confusion between similar categories. We contribute a new benchmark of 18 discourse functions across diverse sections. Experimental results show that OPINE generally outperforms strong baselines, reaching F1 scores of 63.20%, 53.68%, and 63.22% under Micro, Macro, and Example settings, respectively. Our analysis reveals that integrating discourse structures as explicit priors is superior to conventional passive context integration, while pairwise refinement successfully mitigates confusion between functionally similar categories. The code and dataset are available at https://github.com/znoodle63/OPINE.

Weiwei Qi, Zefeng Wu, Tianhang Zheng, Zikang Zhang, Xiaojun Jia, Zhan Qin, Kui Ren

Ensuring Large Language Model (LLM) safety is crucial, yet the lack of a clear understanding about safety mechanisms hinders the development of precise and reliable methodologies for safety intervention across diverse tasks. To better understand and control LLM safety, we propose the Expected Safety Impact (ESI) framework for quantifying how different parameters affect LLM safety. Based on ESI, we reveal distinct safety-critical patterns across different LLM architectures: In dense LLMs, many safety-critical parameters are located in value matrices (V) and MLPs in middle layers, whereas in Mixture-of-Experts (MoE) models, they shift to late-layer MLPs. Leveraging ESI, we further introduce two targeted intervention paradigms for safety enhancement and preservation, i.e., Safety Enhancement Tuning (SET) and Safety Preserving Adaptation (SPA). SET can align unsafe LLMs by updating only a few safety-critical parameters, effectively enhancing safety while preserving original performance. SPA safeguards well-aligned LLMs during capability-oriented intervention (e.g., instruction tuning) by preventing disruption of safety-critical weights, allowing the LLM to acquire new abilities while maintaining safety capabilities. Extensive evaluations on different LLMs demonstrate that SET can reduce the attack success rates of unaligned LLMs by over 50% with only a 100-iteration update on 1% of model weights. SPA can limit the safety degradation of aligned LLMs within 1% after a 1,000-iteration instruction fine-tuning on different tasks. Our code is available at: https://github.com/ZJU-LLM-Safety/SafeWeights-ACL

Poon Tsz Nok, Antonio Valerio Miceli Barone

We introduce a self-play framework for semantic equivalence in Haskell, utilizing formal verification to guide adversarial training between a generator and an evaluator. The framework leverages Liquid Haskell proofs for validating equivalence and execution-based counterexamples for inequivalence, organized via a difficulty-aware curriculum. To facilitate this, we release OpInstruct-HSx, a synthetic dataset of \approx 28k validated Haskell programs. Empirical experiments show that our evaluator transfers effectively to downstream tasks, achieving up to 13.3pp accuracy gain on EquiBench and consistent gains on PySecDB. Ablation studies on the SEQ-SINQ regimes indicate that while inequivalence supervision provides data volume, equivalence proofs are uniquely responsible for the model’s reasoning capabilities. The entire training pipeline and dataset are publicly released on GitHub and Hugging Face respectively.

Jongyoon Kim, Minseong Hwang, Seung-won Hwang

Unsupervised domain adaptation generalizes neural retrievers to an unseen domain by generating pseudo queries on target domain documents. The quality and efficiency of this adaptation critically depend on which documents are selected for pseudo query generation. The existing document sampling method focuses on diversity but fails to capture model uncertainty. In contrast, we propose Uncertainty-based Iterative Document Sampling (UnIte), addressing these limitations by (1) filtering documents with high aleatoric uncertainty and (2) prioritizing those with high epistemic uncertainty, maximizing the learning utility of the current model. We conducted extensive experiments on a large corpus of BEIR with small and large models, showing significant gains of +2.45 and +3.49 nDCG@10 with a smaller training sample size, 4k on average.

Dylan Xinming Hou, Juntian Zhang, Xu Gu, Yichen Wu, Nils Lukas, Gus Xia, Xiuying Chen, Yuhan Liu

The evolving realism of AI-generated Videos (AIGC-V) is rapidly rendering traditional artifact-centric detection insufficient, necessitating a paradigm shift from low-level inspection to high-level semantic verification. This paper presents a comprehensive survey of AIGC-V detection, reframing the task as Factual Fidelity Verification, which asks whether the events, entities, and physical processes depicted in a video are consistent with real-world facts. To systematize this rapidly evolving field, we propose a Vision–Language Dual-View taxonomy that organizes existing methods into a hierarchical, four-layer landscape, spanning intrinsic cue analysis, spatiotemporal consistency modeling, cross-modal consistency reasoning, and language-guided world-level reasoning. This dual-view framing highlights a fundamental transition from artifact matching to evidence-based semantic verification enabled by vision–language models and agentic reasoning pipelines. Based on a systematic review of 195 papers, we synthesize AIGC-V generation paradigms, survey the landscape of detection methods, and review evaluation metrics and benchmarks in line with proposed views. Finally, we discuss current challenges and identify promising directions toward robust, explainable, and trustworthy detection.

Jiuan Zhou, Yu Cheng, Yuan Xie, Zhaoxia Yin

With the rapid progress of LLMs, high quality generative text has become widely available as a cover for text steganography. However, prevailing methods rely on hand-crafted or pre-specified strategies and struggle to balance efficiency, imperceptibility, and security, particularly at high embedding rates. Accordingly, we propose Auto-Stega, an agent-driven self-evolving framework that is the first to realize self-evolving steganographic strategies by automatically discovering, composing, and adapting strategies at inference time; the framework operates as a closed loop of generating, evaluating, summarizing, and updating that continually curates a structured strategy library and adapts across corpora, styles, and task constraints. A decoding LLM recovers the information under the shared strategy. To handle high embedding rates, we introduce PC-DNTE, a plug-and-play algorithm that maintains alignment with the base model’s conditional distribution at high embedding rates, preserving imperceptibility while enhancing security. Experimental results demonstrate that at higher embedding rates Auto-Stega achieves superior performance with gains of 42.2% in perplexity and 1.6% in anti-steganalysis performance over SOTA methods.

Massimiliano Pronesti, Anya Belz, Yufang Hou

Recent work on reinforcement learning with verifiable rewards (RLVR) has shown that large language models (LLMs) can be substantially improved using outcome-level verification signals, such as unit tests for code or exact-match checks for mathematics. In parallel, process supervision has long been explored as a way to shape the intermediate reasoning behaviour of LLMs, but existing approaches rely on neural judges to score chain-of-thought steps, leaving them vulnerable to opacity, bias, and reward hacking. To address this gap, we introduce Verifiable Process Reward Models (VPRMs), a reinforcement-learning framework in which intermediate reasoning steps are checked by deterministic, rule-based verifiers. We apply VPRMs to risk-of-bias assessment for medical evidence synthesis, a domain where guideline-defined criteria and rule-based decision paths enable programmatic verification of reasoning traces. Across multiple datasets, we find that VPRMs generate reasoning that adheres closely to domain rules and achieve substantially higher coherence between step-level decisions and final labels. Results show that VPRMs achieve up to 20% higher F1 than state-of-the-art models and 6.5% higher than verifiable outcome rewards, with substantial gains in evidence grounding and logical coherence.

Dongwon Jo, Jiwon Song, Yulhwa Kim, Jae-Joon Kim

While large language models (LLMs) excel at handling long-context sequences, they require substantial prefill computation and key-value (KV) cache, which can heavily burden computational efficiency and memory usage in both prefill and decoding stages.Recent works that compress KV caches with prefill acceleration reduce this cost but inadvertently tie the prefill compute reduction to the decoding KV budget. This coupling arises from overlooking the layer-dependent variation of critical context, often leading to accuracy degradation. To address this issue, we introduce FastKV, a KV cache compression framework designed to reduce latency in both prefill and decoding by leveraging the stabilization of token importance in later layers.FastKV performs full-context computation until a Token-Selective Propagation (TSP) layer, which forwards only the most informative tokens to subsequent layers.From these propagated tokens, FastKV independently selects salient KV entries for caching, thereby decoupling KV budget from the prefill compute reduction based on the TSP decision.This independent control of the TSP rate and KV retention rate enables flexible optimization of efficiency and accuracy.Experimental results show that FastKV achieves speedups of up to 1.82\times in prefill and 2.87\times in decoding compared to the full-context baseline, while matching the accuracy of the decoding-only baselines.Our code is available at https://github.com/dongwonjo/FastKV.

Anal Roy Chowdhury, Debarshi Kumar Sanyal

Vision-Language Models (VLMs) have shown strong generalization across multimodal tasks, but their capacity to handle sign language translation (SLT), which requires fine-grained spatiotemporal reasoning and linguistic understanding, remains unclear. In this study, we evaluate whether small VLMs (with \leq3B parameters) can perform SLT effectively. We perform supervised fine-tuning on four publicly available multilingual SLT datasets, including one German (DGS), two American (ASL), and one Indian (ISL), applying parameter-efficient LoRA to the language decoder while keeping the vision encoder frozen and training only the connector. To evaluate translation quality, we propose entity- and semantics-aware metrics tailored for SLT. We highlight the data imbalance issues present in the above widely used SLT datasets. Our analysis highlights the limitations in applying general-purpose VLMs to SLT, unlike their applicability in other tasks, and provides insights to inform future development of VLMs for SLP, which is essential for building inclusive AI applications.

Sikui Zhang, Guangze Gao, Ziyun Gan, Chunfeng Yuan, Zefeng Lin, Houwen Peng, Bing Li, Weiming Hu

Large language models (LLMs) experience significant performance degradation when the input exceeds the pretraining context window, primarily due to the out-of-distribution (OOD) behavior of Rotary Position Embedding (RoPE). Recent studies mitigate this problem by remapping OOD positions into the in-distribution range with fixed mapping strategies, ignoring the dynamic relationship between input length and the model’s effective context window. To this end, we propose Length-aware Multi-grained Positional Encoding (LaMPE), a training-free method that fully utilizes the model’s effective context window for adaptive long-context scaling in LLMs. Motivated by the left-skewed frequency distribution of relative positions, LaMPE establishes a dynamic relationship between mapping length and input length through a parametric scaled sigmoid function to adaptively allocate positional capacity across varying input lengths. Meanwhile, LaMPE devises a novel multi-grained attention mechanism that strategically allocates positional resolution across different sequence regions to capture both fine-grained locality and long-range dependencies. Our method can be seamlessly applied to a wide range of RoPE-based LLMs without training. Extensive experiments on three representative LLMs across five mainstream long-context benchmarks demonstrate that LaMPE achieves significant performance improvements compared to existing length extrapolation methods.

Wenqing Wu, Yi Zhao, Yuzhuo Wang, Siyou Li, Juexi Shao, Yunfei Long, Chengzhi Zhang

Novelty is a core requirement in academic publishing and a central focus of peer review, yet the growing volume of submissions has placed increasing pressure on human reviewers. While large language models (LLMs), including those fine-tuned on peer review data, have shown promising results in generating review comments, the absence of dedicated benchmarks has limited systematic evaluation of their ability to assess research novelty. To address this gap, we introduce NovBench, the first large-scale benchmark designed to evaluate LLMs’ capability to generate novelty evaluations in support of human peer review. NovBench comprises 1,684 paper–review pairs from a leading NLP conference, including novelty descriptions extracted from paper introductions and corresponding expert-written novelty evaluations. We focus on both sources because the introduction provides a standardized and explicit articulation of novelty claims, while expert-written novelty evaluations constitute one of the current gold standards of human judgment. Furthermore, we propose a four-dimensional evaluation framework (including Relevance, Correctness, Coverage, and Clarity) to assess the quality of LLM-generated novelty evaluations. Extensive experiments on both general and specialized LLMs under different prompting strategies reveal that current models exhibit limited understanding of scientific novelty, and that fine-tuned models often suffer from instruction-following deficiencies. Our findings underscore the need for targeted fine-tuning strategies that jointly improve novelty comprehension and instruction adherence.

Xuexiang Wen, Hang Yu, Linchao Zhu, Gaoang Wang

While Reinforcement Learning with Verifiable Rewards (RLVR) has recently emerged as a promising post-training paradigm for Large Language Models (LLMs), its dependency on the gold label or domain-specific verifiers limits its scalability to new tasks and domains. In this work, we propose **Verifier-free Intrinsic Gradient-Norm Reward (VIGOR)**, a simple reward that uses only the policy model itself. Given a prompt, VIGOR samples a group of completions and assigns higher within-group rewards to outputs that induce smaller \ell_2 norms of the teacher-forced negative log-likelihood gradients under the current parameters. Intuitively, lower gradient norms suggest the completion aligns better with the current policy, serving as an intrinsic preference signal for policy optimization. To make this intrinsic signal practical for RL, we correct the systematic length bias of averaged token-level gradients with a \sqrt{T} scaling, and apply group-wise rank shaping to stabilize reward scales across prompts. Across mathematical reasoning benchmarks, VIGOR outperforms the state-of-the-art Reinforcement Learning from Internal Feedback (RLIF) baseline INTUITOR, and it also exhibits cross-domain transfer to code benchmarks when trained only on math data. For instance, on Qwen2.5-7B-Base post-trained on MATH, VIGOR improves the average math accuracy by +3.31% and the average code accuracy by +1.91% over INTUITOR, while exhibiting more stable training dynamics.

Shengmin Piao, Sanghyun Park

Recent advances in large reasoning models have been driven by reinforcement learning and test-time scaling, accompanied by growing interest in latent rather than purely textual reasoning. However, existing latent reasoning methods lack mechanisms to ensure stable reasoning dynamics in latent space and a systematic way to interleave implicit and explicit reasoning. We introduce SpiralThinker, a unified framework that performs iterative updates over latent representations while enabling interleaved reasoning across latent and textual steps. At its core, SpiralThinker employs a progressive alignment objective and structured annotations to stabilize latent reasoning and maintain coherence with textual reasoning. Across mathematical, logical, and commonsense reasoning tasks, SpiralThinker achieves state-of-the-art performance among latent reasoning baselines. Detailed analyses reveal that both iteration and alignment are indispensable, the numbers of latent tokens and iterations exhibit dataset-specific optima, and appropriate alignment proves critical for an effective iterative process. Overall, SpiralThinker bridges iterative computation and latent reasoning, demonstrating that aligned iterative updates can reliably steer reasoning in the latent space.

Dongyang Zhen, Niping Duan, Huan Zhou, Qingbin Cui

GPS trajectories encode rich behavioral information about how people move, organize activities, and form daily routines. Recent advances in large language models (LLMs) raise a natural question: can such models infer and summarize travel behavior directly from mobility traces? This paper introduces TravelBehaviorQA, a large-scale benchmark dataset that reframes trajectory analysis as a language-based behavioral understanding task. The dataset links raw GPS trajectories with human-grounded question-answering (QA) pairs that capture travel intensity, temporal structure, activity patterns, mode usage, and behavioral routines. Unlike prior mobility datasets focused on prediction or classification, TravelBehaviorQA emphasizes semantic interpretation through a unified mix of deterministic and open-ended questions. In this benchmark, we construct over 143k QA instances spanning users and years, and evaluate a broad range of state-of-the-art LLMs under controlled settings. Our results reveal substantial gaps between factual extraction and genuine behavioral reasoning, showing that model scale alone is insufficient and that trajectory representation is a primary bottleneck. TravelBehaviorQA exposes critical limitations of current models and establishes a rigorous benchmark for advancing language-based understanding of human mobility behavior.

Ruiran Yan, Wen Xiong, Ze Liu, Chaozhuo Li, Hao Liao, Defu Lian, Zheng Liu

Large language models (LLMs) have demonstrated that explicitly performing step-by-step thinking before producing final outputs can substantially improve performance on complex tasks, as exemplified by recent reasoning-oriented models such as OpenAI O1 and DeepSeek R1. Inspired by these advancements, we propose the O1 Embedder, a novel approach aiming to endow retrieval models with similar capabilities to address challenges like multi-task retrieval, zero-shot retrieval, and tasks requiring intensive reasoning of complex relationships. The O1 Embedder generates preliminary thoughts for input queries before document retrieval. To realize this objective, we address two fundamental challenges in integrating thinking mechanisms into dense retrieval. First, retrieval tasks lack explicit supervision for intermediate thinking processes, making it difficult to define thoughts that are truly useful for retrieval. We address this challenge with a data synthesis framework following an “Exploration-Refinement” process, ensuring alignment with retrieval utility. Second, effectively integrating thought generation with representation learning requires a unified modeling framework that can jointly support generation and embedding within a single model. O1 Embedder addresses this challenge by jointly optimizing thought generation and dense retrieval in an end-to-end manner, enhancing retrieval accuracy while reducing complexity through a single deployable model. Extensive evaluations across diverse datasets demonstrate significant performance improvements, highlighting the effectiveness and generalization capability of O1 Embedder.

Xin Wu, Yuqi Bu, Mengchen Zhao, Qingbao Huang, Yi Cai

Recent advancements in large language models (LLMs) have demonstrated their remarkable capabilities in natural language understanding and generation. However, these models still struggle with formal logical reasoning, often producing coherent yet invalid conclusions due to limitations in representing boundaries and relational structures through text alone. Human cognition frequently relies on visual representations to clarify logical structures involving category membership, inclusion, and relational hierarchies. Inspired by this, we investigate whether incorporating visual logic diagrams into LLMs’ reasoning workflows can enhance their performance on formal logic tasks. We study this question in a controlled setting using syllogistic and conditional reasoning with programmatically generated Venn, Euler, and Linear diagrams. Across three Vision Language Models (VLMs) families, diagrams help in some settings but can also hurt performance, especially on logically invalid cases where models may over-rely on a single static visual instantiation. We therefore present this work as a reproducible evaluation framework and empirical analysis of when logic diagrams help or hinder language-conditioned reasoning.

Ruichao Yang, Yufan Bian, Wei Gao, Bo-Wen Zhang, Jing Ma, Hongzhan Lin, Ziyang Luo, Xiaobin Zhu, Xu-Cheng Yin

Current multimodal fake news detectors predominantly function as opaque classifiers, offering limited deductive transparency and little insight into how conflicting evidence is reconciled. To address this limitation, we propose Dialectical Structured Reasoning (DSR), a framework modeling fake news detection as an explicit dialectical process over multimodal social context. DSR instantiates two opposing agents: a Verifier, which constructs evidence paths supporting semantic consistency, and a Debunker, which actively explores exposing logical or factual contradictions. Then a differentiable Judge agent adjudicates between these competing perspectives by integrating local evidence with global parametric knowledge. Experiments on three benchmarks demonstrate that DSR achieves state-of-the-art performance while producing transparent, dialectically grounded explanations that closely mirror human reasoning process.

Yu Cui, Hang Fu, Haibin Zhang, Licheng Wang, Cong Zuo

Multi-agent debate (MAD) is an emerging approach to improving the reasoning capabilities of large language models (LLMs). Existing MAD methods rely on multiple rounds of interaction among agents to reach consensus, and the final output is decided by majority voting in the last round. However, this consensus-based design faces several limitations. First, multiple rounds of communication increases token overhead and limits scalability. Second, due to the inherent conformity of LLMs, agents that initially produce correct responses may be influenced by incorrect ones during the debate process, causing error propagation. Third, majority voting introduces randomness and unfairness in the decision-making phase, and can degrade the reasoning performance. To address these issues, we propose Free-MAD, an alternative and novel MAD framework that eliminates the need for consensus among agents. Free-MAD introduces a novel score-based decision mechanism that evaluates the entire debate trajectory rather than relying on the last round only. This mechanism tracks how each agent’s reasoning evolves, enabling more accurate and fair outcomes. In addition, Free-MAD reconstructs the debate phase by introducing anti-conformity, a mechanism that enables agents to mitigate excessive influence from the majority. Experiments on eight benchmark datasets demonstrate that Free-MAD significantly improves reasoning performance while requiring only a single-round debate and thus reducing token costs. We also show that compared to existing MAD approaches, Free-MAD exhibits improved robustness in real-world attack scenarios.

Md. Nazmul Islam Ananto, Shamit Fatin, Mohammed Eunus Ali, Md Rizwan Parvez

The popular path query - identifying the most frequented routes between locations from historical trajectory data - has important applications in urban planning, navigation optimization, and travel recommendations. While traditional algorithms and machine learning approaches have achieved success in this domain, they typically require model training, parameter tuning, and retraining when accommodating data updates. As Large Language Models (LLMs) demonstrate increasing capabilities in spatial and graph-based reasoning, there is growing interest in exploring how these models can be applied to geo-spatial problems.We introduce CompassLLM, a novel multi-agent framework that intelligently leverages the reasoning capabilities of LLMs into the geo-spatial domain to solve the popular path query. CompassLLM employs its agents in a two-stage pipeline: the SEARCH stage that identifies popular paths, and a GENERATE stage that synthesizes novel paths in the absence of an existing one in the historical trajectory data. Experiments on real and synthetic datasets show that CompassLLM demonstrates superior accuracy in SEARCH and competitive performance in GENERATE while being cost-effective.

ShangZhan Li, Xinyu Yin, Xuanyu Jin, Ye He, Yuxin Zhou, Yuxuan Li, Xu Han, Wanxiang Che, Qi Shi, Ting Liu 等

Vectorization via Single Instruction, Multiple Data (SIMD) architectures is a cornerstone of high-performance computing. To fully exploit hardware potential, developers often resort to explicit vectorization using intrinsics, as compiler-based auto-vectorization frequently yields suboptimal results due to conservative static analysis. While Large Language Models (LLMs) have demonstrated remarkable proficiency in general code generation, they struggle with explicit vectorization due to the scarcity of high-quality corpora and the strict semantic constraints of low-level hardware instructions. In this paper, we propose AutoVecCoder, a novel framework designed to empower LLMs with the capability of automated explicit vectorization. AutoVecCoder integrates two core components: VecPrompt, an automated data synthesis pipeline to inject domain-specific intrinsic knowledge; and VecRL, a reinforcement learning framework that aligns code generation with execution efficiency. AutoVecCoder-8B trained by this framework achieves state-of-the-art performance on the SSE and AVX subsets of SimdBench and, in some cases, generates implementations surpassing standard -O3 optimizations, effectively overcoming the inherent bottlenecks of traditional automated vectorization.