论文检索

Seunghyun Park, Yuanyuan Lei

While LLMs demonstrate impressive reasoning capabilities, they remain fragile in multi-step logic deduction, where a single transition error can propagate through the entire reasoning chain, leading to unstable performance. In this work, we identify logical connectives as primary points of this structural fragility. Through empirical analysis, we show that logical connective tokens function as high entropy forking points, at which models frequently struggle to determine the correct logical direction. Motivated by this observation, we hypothesize that intervening in logical connective selection can guide LLMs towards the correct logical direction, thereby improving the overall reasoning chain. To validate this hypothesis, we propose a multi-layered framework that intervenes specifically at these logic-critical junctions in the reasoning process. Specifically, we introduce (1) Gradient-based Logical Steering to guide LLMs internal representations towards valid reasoning subspaces, (2) Localized Branching to resolve ambiguity via targeted look-ahead search, and (3) Targeted Transition Preference Optimization, a surgical reinforcement learning objective that selectively optimizes single-token preferences at logical pivots. Crucially, by concentrating intervention solely on logic-critical transitions, our framework achieves a favorable accuracy–efficiency trade-off compared to global inference time scaling methods like beam search and self-consistency.

Meixiu Long, Duolin Sun, Dan Yang, Yihan Jiao, Lei Liu, Jiahai Wang, Binbin Hu, Yue Shen, Jie Feng, Zhehao Tan 等

Large Language Models (LLMs) have emerged as powerful tools for passage reranking in information retrieval, leveraging their superior reasoning capabilities to address the limitations of conventional models on complex queries. However, current LLM-based reranking paradigms are fundamentally constrained by an efficiency-accuracy trade-off: (1) pointwise methods are efficient but ignore inter-document comparison, yielding suboptimal accuracy; (2) listwise methods capture global context but suffer from context-window constraints and prohibitive inference latency. To address these issues, we propose GroupRank, a novel paradigm that balances flexibility and context awareness. To unlock the full potential of groupwise reranking, we propose an answer-free data synthesis pipeline that fuses local pointwise signals with global listwise rankings. These samples facilitate supervised fine-tuning and reinforcement learning, with the latter guided by a specialized group-ranking reward comprising ranking-utility and group-alignment. These complementary components synergistically optimize document ordering and score calibration to reflect intrinsic query-document relevance.Experimental results show GroupRank achieves a state-of-the-art 65.2 NDCG@10 on BRIGHT and surpasses baselines by 2.1 points on R2MED, while delivering a 6.4\times inference speedup. The code is available at https://github.com/AQ-MedAI/Diver/tree/main/Reranker/GroupRank.

Yurui Pan, Ke Xu, Bo Peng

Alignment of large language models (LLMs) typically relies on supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF), or more recently direct preference optimization (DPO). However, existing objectives largely ignore the global geometry and topology of the representation space: they operate on local token-level likelihoods or scalar preference scores, and do not explicitly constrain how hidden states move from a user prompt to an answer.We view generation as tracing a semantic trajectory in hidden space, and propose a topology-enhanced alignment framework that regularizes these trajectories using 0-dimensional persistent homology. First, at the SFT stage, we introduce a Trajectory Topology Loss (TTL). For each batch, we treat mean-pooled embeddings of prompts and gold answers as a mixed point cloud, run a Union-Find-based 0D persistent homology algorithm, and extract ”prompt–answer bridge” edges that connect previously disconnected components. TTL encourages the model’s actual update direction from prompt to answer to align with these topologically derived bridges, rather than with arbitrary or per-example directions.Second, at the RLHF/DPO stage, we propose Topological Preference Optimization (TPO). TPO constructs topic-specific semantic preference vectors from an offline pipeline and aligns the semantic improvement direction between rejected and chosen responses with these vectors in an intermediate hidden layer. We further introduce an exponential-moving-average-based dynamic weighting scheme to balance DPO and TPO losses, and also explore a fully topological variant that applies persistent homology on the chosen/rejected embedding cloud.We instantiate our methods on Qwen2.5-7B-Instruct and evaluate on UltraChat and Anthropic HH-RLHF. Across both SFT and DPO training, topology-enhanced objectives consistently outperform strong non-topological baselines (including per-example, nearest-neighbor, and random direction regularizers) on automatic preference metrics and LLM-judge evaluations, while maintaining or slightly improving toxicity. These results suggest that incorporating persistent homology and trajectory geometry is a promising and practical direction for more controllable LLM alignment.

Xiaolong Wei, Zerun Zhu, Simin Niu, Xingyu Zhang, Peiying Yu, Changxuan Xiao, Yuchen Li, Jicheng Yang, Zhejun Zhao, Chong Meng 等

A fundamental challenge in creative writing lies in reconciling the inherent tension between maintaining global coherence in long-form narratives and preserving local expressiveness in short-form texts. While long-context generation necessitates explicit macroscopic planning, short-form creativity often demands spontaneous, constraint-free expression. Existing alignment paradigms, however, typically employ static reward signals and rely heavily on high-quality supervised data, which is costly and difficult to scale. To address this, we propose UniCreative, a unified reference-free reinforcement learning framework. We first introduce AC-GenRM, an adaptive constraint-aware reward model that dynamically synthesizes query-specific criteria to provide fine-grained preference judgments. Leveraging these signals, we propose ACPO, a policy optimization algorithm that aligns models with human preferences across both content quality and structural paradigms without supervised fine-tuning and ground-truth references. Empirical results demonstrate that AC-GenRM aligns closely with expert evaluations, while ACPO significantly enhances performance across diverse writing tasks. Crucially, our analysis reveals an emergent meta-cognitive ability: the model learns to autonomously differentiate between tasks requiring rigorous planning and those favoring direct generation, validating the effectiveness of our direct alignment approach.

Siyuan Deng, Yerong Li, Fanhua Shang, Hongying Liu

Despite advances in safety alignment, Large Language Models (LLMs) remain vulnerable to jailbreaking attacks. However, prevailing methods suffer from a dichotomy of limitations: they either rely on prohibitive iterative optimization in the input space (leading to high computational costs) or fail to penetrate the model’s internal decision-making processes. In this work, we identify a critical structural vulnerability: the ”Attention Sink” mechanism—originally designed to maintain generation stability by anchoring to initial tokens—unintentionally serves as a computational anchor for safety alignment. We hypothesize and empirically verify that safety guardrails are not globally distributed but are predominantly ”front-loaded” in specific attention heads of shallow layers. To audit this structural fragility, we propose FLASH (Focused Layer Attention Sink Hijacking), a novel diagnostic auditing framework that executes a surgical intervention. By precisely scaling attention scores in these vulnerable layers, we dismantle the model’s internal safety anchor. To ensure the attack’s robustness and coherence against the resulting internal noise, we synergize this intervention with multi-variant query rewriting and an adaptive dynamic decoding strategy. Extensive experiments on Llama-3, Qwen-3, and others demonstrate that FLASH achieves a state-of-the-art Attack Success Rate of over 77% with an unprecedented efficiency of 1.53 queries on average. This work marks a paradigm shift from brute-force optimization to mechanism-driven diagnostic auditing, exposing a fundamental trade-off between architectural stability and safety security.

Yu Pan, Xiongfei Wu, Yang Yuguang, Jixun Yao, Maxime Cordy, Lei Ma, Jianjun Zhao

Despite recent advances in speech-to-speech translation (S2ST), it remains difficult to achieve both high translation accuracy and practical flexibility. In this paper, we present S2ST-Omni, a compositional S2ST framework that integrates a high-accuracy speech-to-text translation (S2TT) frontend with a modular, plug-and-play text-to-speech (TTS) backend, enabling independent optimization of translation and synthesis. On the S2TT side, we introduce a hybrid adapter that follows a "local-then-global" strategy to bridge the pretrained Whisper encoder and Qwen3 LLM, yielding a hierarchical acoustic-to-semantic abstraction. Building on this bridge, we further propose a hierarchical language-aware architecture that injects source-language information at two complementary levels. At the acoustic level, Language-Aware Dual-CTC operates on intermediate adapter features and employs FiLM-style feature modulation with a learnable gate, encouraging the model to learn language-specific but content-faithful acoustic representations. At the linguistic level, Language-Aware Prompting dynamically constructs source-language-conditioned prompts that activate language-specific translation knowledge in the LLM. To enable efficient optimization, we design a task-specific progressive fine-tuning strategy that first stabilizes speech-text alignment and then improves translation via LoRA on top of this converged foundation. The TTS backend remains fully modular and can be instantiated with any state-of-the-art synthesizer without retraining the S2TT frontend. Experiments on CVSS-C show that S2ST-Omni consistently achieves the best BLEU and ASR-BLEU across French, German, and Spanish to English directions, outperforming strong recent S2ST baselines.

Mengna Gao, Dapeng Yin, Shuyue Zhu, Bingxuan Hou, Zhanpeng Ni, Junli Wang

Retrieval-Augmented Generation (RAG) systems rely on Optical Character Recognition (OCR) to ingest knowledge from unstructured documents. However, OCR engines often struggle with complex layouts, introducing Structural Noise, such as line insertion and paragraph interleaving, which disrupts the semantic flow of the text. Existing evaluations largely overlook this dimension, operating on the assumption of structurally perfect input. To bridge this gap, we introduce StruNRAG, a dedicated benchmark for evaluating RAG robustness against OCR-induced structural perturbations. We construct a bilingual dataset of 2,132 question-answer pairs derived from complex Chinese and English documents and systematically inject three categories of real-world structural noise: line insertion, paragraph interleaving, and line interleaving. Our evaluation of mainstream retrievers and Large Language Models (LLMs) reveals a nuanced interaction between noise and pipeline stages: while structural distortions consistently degrade retrieval performance, the generation stage exhibits unexpected robustness. Advanced LLMs demonstrate robustness against local noise (e.g., line insertion), but struggle to maintain reasoning capabilities under severe structural disruption that fragments global context. These findings indicate that while LLMs are capable of compensating for minor parsing errors, future RAG optimizations must take into account the effects of structural noise. Our code and datasets are available at [https://github.com/GaoMengnana/StruNRAG](https://github.com/GaoMengnana/StruNRAG).

Shijia Xu, Yu Wang, Xiaolong Jia, Zhou Wu, Kai Liu, April Xiaowen Dong

Despite the adoption of Large Language Models (LLMs) in legal AI, automated contract revision remains impeded because generic models often treat strict legal constraints as mere suggestions. To address this safety gap, we introduce the Risk-Constrained Bilevel Stackelberg Framework (RCBSF), modeling high-stakes revision as a rigorous strategic interaction rather than an open-ended conversation. RCBSF establishes a hierarchical Leader-Follower structure: a Global Prescriptive Agent (GPA) leader imposes definitive risk budgets, while a follower system—comprising a Constrained Revision Agent (CRA) and a Local Verification Agent (LVA)—iteratively optimizes the output within these strict boundaries. We theoretically prove this bilevel formulation converges to an equilibrium yielding strictly superior utility over unguided methods. Empirically, RCBSF achieves state-of-the-art performance, surpassing iterative baselines with an average Risk Resolution Rate (RRR) of 84.21% and enhanced token efficiency. Our code is available at https://github.com/xjiacs/RCBSF .

Lin Xv, Xian Gao, Ting Liu, Yuzhuo fu

Low-rank decomposition, particularly Singular Value Decomposition (SVD), is a pivotal technique for mitigating the storage and computational demands of Large Language Models (LLMs). However, prevalent SVD-based approaches overlook the critical phenomenon that decomposition errors exhibit significant disparity across different components of the parameter matrix, often leading to suboptimal approximation. Furthermore, existing methods lack a direct metric to evaluate the importance of individual weight matrices. To address these limitations, we propose **Duo-SVD** (**Du**al-level **O**ptimization **SVD**), a novel training-free framework that synergizes optimization at both the column and the module levels. First, Duo-SVD incorporates a Column-Preserving Strategy that explicitly retains columns exhibiting high decomposition errors, while applying low-rank approximation solely to those with lower errors. Second, at the module level, we employ a Module-Adaptive Allocation Strategy that formulates ratio allocation as a global constrained optimization problem based on perturbation-induced model deviation. Extensive experiments demonstrate that Duo-SVD consistently outperforms state-of-the-art SVD-based baselines and structured pruning methods, establishing it as a superior paradigm for efficient LLM compression.

Yunzhi Shen, Hao Zhou, Xin Huang, Xue Han, Junlan Feng, Shujian Huang

Reinforcement learning (RL) has shown strong promise for LLM-based machine translation, with recent methods such as GRPO demonstrating notable gains; nevertheless, translation-oriented RL remains challenged by high-variance policy gradients induced by Monte Carlo baselines, as well as a large trajectory space that favors global exploration over fine-grained local optimization. We introduce PEGRL, a two-stage RL framework that uses post-editing as an auxiliary task to stabilize training and guide overall optimization. At each step, translation outputs are sampled to construct post-editing inputs, enabling lower-variance gradients from the post-editing task to propagate through the entire framework while jointly supporting both global exploration and fine-grained local optimization. A task-specific weighting scheme further emphasizes the post-editing gradient, producing a biased yet more sample-efficient estimator. Experiments on English\toFinnish, English\toTurkish, and English\leftrightarrowChinese show consistent gains over RL baselines, and for English\toTurkish, performance on COMETKiwi is comparable to advanced LLM-based systems (DeepSeek-V3.2). Our code and a set of representative pretrained models are publicly available at https://github.com/NJUNLP/peg-rl and https://huggingface.co/collections/DGME/pegrl.

Yishu Lei, Shuwei He, Hu Jing, Dan Zhang, Xianlong Luo, Danxiang Zhu, Shikun Feng, Rui Liu, Jingzhou HE, Yu Sun 等

Extending the input modality of Large Language Models (LLMs) to the audio domain is essential for achieving comprehensive multimodal perception. However, it is well-known that acoustic information is intrinsically heterogeneous, entangling attributes such as speech, music, and environmental context. Existing research is limited to a dense, parameter-shared adapter to model these diverse patterns, which induces gradient conflict during optimization, as parameter updates required for distinct attributes contradict each other. To address this limitation, we introduce the MoE-Adapter, a sparse Mixture-of-Experts (MoE) architecture designed to decouple acoustic information. Specifically, it employs a dynamic gating mechanism that routes audio tokens to specialized experts capturing complementary feature subspaces while retaining shared experts for global context, thereby mitigating gradient conflicts and enabling fine-grained feature learning. Comprehensive experiments show that the MoE-Adapter achieves superior performance on both audio semantic and paralinguistic tasks, consistently outperforming dense linear baselines with comparable computational costs. To facilitate future research, our code are publicly available at https://github.com/Alittleegg/Eureka-Audio.

Xudong Wang, Chaoning Zhang, Chenghao Li, Shuxu Chen, Qigan Sun, Jiaquan Zhang, Fachrina Dewi Puspitasari, Tae-Ho Kim, Jiwei Wei, Malu Zhang 等

Large Language Models (LLMs) have demonstrated strong capabilities in complex reasoning tasks, while recent prompting strategies such as Chain-of-Thought (CoT) have further elevated their performance in handling complex logical problems. Despite these advances, high-quality reasoning remains heavily reliant on manual static prompts and is sensitive to decoding configurations and task distributions, leading to performance fluctuations and limited transferability. Existing automatic prompt optimization methods typically adopt single-agent local search, failing to simultaneously optimize prompts and decoding hyperparameters within a unified framework to achieve stable global improvements. To address this limitation, we propose Agent-GWO, a dynamic prompt optimization framework for complex reasoning. Specifically, we unify prompt templates and decoding hyperparameters as inheritable agent configurations. By leveraging the leader-follower mechanism of the Grey Wolf Optimizer (GWO), we automatically select three leader agents (\alpha, \beta, and \delta) to guide the collaborative updates of the remaining agents, enabling iterative convergence toward robust optimal reasoning configurations that can be seamlessly integrated for inference. Extensive experiments on multiple mathematical and hybrid reasoning benchmarks across diverse LLM backbones show that Agent-GWO consistently improves accuracy and stability over existing prompt optimization methods.

Wenhan Liu, Xinyu Ma, Yutao Zhu, Yuchen Li, Daiting Shi, Dawei Yin, Zhicheng Dou

Agentic search has recently emerged as a powerful paradigm, where an agent interleaves multi-step reasoning with on-demand retrieval to solve complex questions. Despite its success, how to design a retriever for agentic search remains largely underexplored. Existing search agents typically rely on similarity-based retrievers, while similar passages are not always useful for final answer generation. In this paper, we propose a novel retriever training framework tailored for agentic search. Unlike retrievers designed for single-turn retrieval-augmented generation (RAG) that only rely on local passage utility, we propose to use both local query-passage relevance and global answer correctness to measure passage utility in a multi-turn agentic search. We further introduce an iterative training strategy, where the search agent and the retriever are optimized bidirectionally and iteratively. Different from RAG retrievers that are only trained once with fixed questions, our retriever is continuously improved using evolving and higher-quality queries from the agent. Extensive experiments on seven single-hop and multi-hop QA benchmarks demonstrate that our retriever, termed Agentic-R, consistently outperforms strong baselines across different search agents.

Ho Hung Lim, Yi Yang

Visual RAG has offered an alternative to traditional RAG. It treats documents as images and uses vision encoders to obtain vision patch tokens. However, hundreds of patch tokens per document create retrieval and storage challenges in a vector database. Practical deployment requires aggregating them into a single vector. This raises a critical question: does single-vector aggregation lose key information in financial documents? We develop a diagnostic benchmark using financial documents where changes in single digits can lead to significant semantic shifts. Our experiments show that single-vector aggregation collapses different documents with almost identical vectors. Metrics show that the patch level detects semantic changes, and confirm that aggregation obscures these details. We identify global texture dominance as the root cause. Our findings are consistent across model scales, retrieval-optimized embeddings, and multiple mitigation strategies, highlighting significant risks for single-vector visual document retrieval in financial applications.

Yuxuan Hu, Jianchao Tan, Jiaqi Zhang, Wen Zan, Pingwei Sun, Yifan Lu, Xunliang Cai, Jing Zhang

In this work, we conduct a systematic analysis of Native Sparse Attention (NSA) and propose targeted improvements that enhance long-context modeling. A key insight is that alternating between local (sliding-window) and global (compression/selective) attention across layers, rather than using fixed patterns, enables more effective propagation of long-range dependencies and substantially boosts performance on long-sequence tasks. Meanwhile, we further refine NSA’s branches with Latent Attention that the sliding-window branch is enhanced with Multi-head Latent Attention (MLA) while compression and selective branches adopt Group-head Latent Attention (GLA). These changes reduce KV-cache memory by 50% versus NSA while improving the model’s common-sense reasoning and long-text understanding capabilities. Experiments on models from 340M to 1.3B parameters (trained on 15B and 100B tokens) show our method matches or exceeds full attention and native sparse attention in both common-sense reasoning and long-context understanding tasks.

Zijian Zheng, Yonghe Lu

Short text clustering has gained significant prominence due to its ubiquity in real-world applications. Despite the recent success of contrastive clustering, existing paradigms still suffer from two critical bottlenecks: (1) conventional data augmentation provides limited semantic granularity and may introduce unintended noise; and (2) the absence of global optimization for cluster assignments often precipitates the accumulation of pseudo-label noise, thereby compromising semantic consistency. To bridge these gaps, we propose MAST, a Multi-view Alignment Strategy with Transport-based clustering. MAST constructs complementary structural views to capture multi-granularity semantic features and introduces a multi-view contrastive objective that jointly aligns original, augmented, and structure-enhanced embeddings. To mitigate representation over-smoothing, we incorporate structure-aware negative reweighting and intermediate-layer negative sampling. Furthermore, MAST employs high-confidence guided refinement and an optimal transport-based pseudo-label alignment mechanism to enforce global semantic consistency across multiple views. Extensive experiments on several benchmark datasets demonstrate that MAST consistently outperforms state-of-the-art methods, establishing a new competitive baseline for short text clustering.

Jaeeun Jang, Hansle Lee, Sangmin Kim

Reinforcement Learning with Verifiable Rewards (RLVR) and Reinforcement Learning from Internal Feedback (RLIF) often fail to benefit from test-time compute due to entropy collapse and the resulting loss of reasoning diversity. We show that this collapse is driven not by uniform entropy decay, but by premature overconfidence at a small number of structurally critical decision points. Based on a token-level analysis of GRPO-style policy optimization, we propose SCOPE (Structural Collapse-aware Optimization via Partial Entropy control), which assigns each generated token a redistribution score and applies selective KL regularization to only the top \sim 5% of tokens under this score. Across model scales and architectures on math reasoning benchmarks, SCOPE consistently improves performance under both RLVR and RLIF settings, demonstrating that targeted entropy control at a vanishingly small subset of tokens is sufficient to sustain reasoning diversity and effective test-time scaling.

Qingcheng Zeng, Yuheng Lu, Zeqi Zhou, Heli Qi, Puxuan Yu, Fuheng Zhao, Hitomi Yanaka, Weihao Xuan, Naoto Yokoya

Code-switching is a pervasive linguistic phenomenon in global communication, yet modern information retrieval systems remain predominantly designed for, and evaluated within, monolingual contexts. To bridge this critical disconnect, we present a holistic study dedicated to code-switching IR. We introduce CSR-L (Code-Switching Retrieval benchmark-Lite), constructing a dataset via human annotation to capture the authentic naturalness of mixed-language queries. Our evaluation across statistical, dense, and late-interaction paradigms reveals that code-switching acts as a fundamental performance bottleneck, degrading the effectiveness of even robust multilingual models. We demonstrate that this failure stems from substantial divergence in the embedding space between pure and code-switched text. Scaling this investigation, we propose CS-MTEB, a comprehensive benchmark covering 11 diverse tasks, where we observe performance declines of up to 27%. Finally, we show that standard multilingual techniques like vocabulary expansion are insufficient to resolve these deficits completely. These findings underscore the fragility of current systems and establish code-switching as a crucial frontier for future IR optimization.

Zhirui Chen, Peiyang Liu, Ling Shao

As Large Language Models (LLMs) scale to support context windows exceeding one million tokens, the linear growth of Key-Value (KV) cache imposes severe memory capacity and bandwidth bottlenecks, constraining the efficiency of long-context inference. Existing compression approaches typically prioritize tokens based on local saliency metrics to decouple prefill computation from decoding memory. However, these methods often rely on local saliency snapshots at a specific layer, thereby systematically discarding tokens that act as global information hubs across the network depth but appear temporarily dormant at the specific layer selected for pruning. To address this limitation, we propose StructKV, a structure-aware KV cache compression framework that introduces three core innovations: First, Global In-Degree Centrality aggregates attention patterns across the network depth to identify global information hubs. Second, Dynamic Pivot Detection utilizes information-theoretic metrics to adaptively locate the optimal layer for compression. Finally, Structural Propagation Decoupling separates the computational budget from the memory storage budget. Experimental results on the LongBench and RULER benchmarks demonstrate that StructKV effectively preserves long-range dependencies and retrieval robustness.

Tianxiang Ma, Weijie Feng, Xinyu Wang, Zhiyong Cheng

Emotion-Cause Pair Extraction in Conversations (ECPEC) aims to identify the set of causal relations between emotion utterances and their triggering causes within a dialogue. Most existing approaches formulate ECPEC as independent pairwise classification, overlooking the distinct semantics of emotion diffusion and cause explanation, and failing to capture globally consistent many-to-many conversational causality. To address these limitations, we revisit ECPEC from a semantic perspective and seek to disentangle emotion-oriented semantics from cause-oriented semantics, mapping them into two complementary representation spaces to better capture their distinct conversational roles. Building on this semantic decoupling, we naturally formulate ECPEC as a global alignment problem between the emotion-side and cause-side representations, and employ optimal transport to enable many-to-many and globally consistent emotion-cause matching. Based on this perspective, we propose a unified framework SCALE that instantiates the above semantic decoupling and alignment principle within a shared conversational structure. Extensive experiments on several benchmark datasets demonstrate that SCALE consistently achieves state-of-the-art performance.