论文检索

Haolin Li, Shuyang Jiang, Ruipeng Zhang, Jiangchao Yao, Ya Zhang, Yanfeng Wang

While large language models hold promise for complex medical applications, their development is hindered by the scarcity of high-quality reasoning data. To address this issue, existing approaches typically distill chain-of-thought reasoning traces from large proprietary models via supervised fine-tuning, then conduct reinforcement learning (RL). These methods exhibit limited improvement on underrepresented domains like rare diseases while incurring substantial costs from generating complex reasoning chains. To efficiently enhance medical reasoning, we propose MedSSR, a Medical Knowledge-enhanced data Synthesis and Semi-supervised Reinforcement learning framework. Our framework first employs rare disease knowledge to synthesize distribution-controllable reasoning questions. We then utilize the policy model itself to generate high-quality pseudo-labels. This enables a two-stage, intrinsic-to-extrinsic training paradigm: self-supervised RL on the pseudo-labeled synthetic data, followed by supervised RL on the human-annotated real data. MedSSR scales model training efficiently without relying on costly trace distillation. Extensive experiments on Qwen and Llama demonstrate that our method outperforms existing methods across ten medical benchmarks, achieving up to **+5.93%** gain on rare-disease tasks. Our code is available at https://github.com/tdlhl/MedSSR.

Tianyu Dong, Yangyang Liu, Jiang Zhou, Xinwei Wu, Xiaohu Zhao, Hao Wang, Heng Liu, Linlong Xu, Longyue Wang, Weihua Luo 等

Sparse Mixture-of-Experts (MoE) architectures have emerged as an increasingly influential paradigm as they offer a strategic balance between parameter scalability and computational efficiency. However, low-resource language tokens are often routed to different experts than those predominantly activated by high-resource inputs, which limits cross-lingual expert sharing. This cross-lingual routing divergence consequently hinders their efficacy in multilingual contexts. To address this issue, we propose SARA (Semantically Anchored Routing Alignment), a framework designed to transfer specialized capabilities from high-resource languages as anchors to low-resource languages. SARA explicitly aligns the routing distribution of multilingual inputs with high-resource semantic anchors using a symmetric Jensen-Shannon (JS) divergence constraint. Unlike traditional distillation methods that operate on output logits, SARA directly aligns the internal routing distributions of MoE layers, encouraging mechanistic consistency in expert selection across languages. We conduct experiments on 2 LLMs across 5 low-resource languages and 3 benchmarks. Experiment results demonstrate that SARA outperforms standard instruction tuning (e.g., +0.8% on Qwen3-30B-A3B and +1.2% on Phi-3.5-MoE-instruct on Global-MMLU benchmark). Further analyses show that SARA effectively addresses performance bottlenecks in low-resource languages, providing a scalable pathway to enhance multilingual capabilities in sparse architectures.

Zhenyang Cai, Jiaming Zhang, Junjie Zhao, Ziyi Zeng, Yanchao Li, Liang Jingyi, Junying Chen, Yunjin Yang, Jiajun You, Shuzhi Deng 等

Reliable interpretation of multimodal dental data is essential for automated oral healthcare, yet current multimodal large language models (MLLMs) show limited understanding of dental images. Although complex reasoning improves performance, its gains in dentistry are substantially smaller than in other medical domains, suggesting that complex reasoning is not yet sufficiently incentivized for dental diagnosis, likely due to insufficient domain knowledge and limited reinforcement learning on dental questions. We present DentalGPT, a dentistry-specialized MLLM trained via staged multimodal alignment and reinforcement learning. By constructing the largest annotated multimodal dental dataset to date with over 120k images, multimodal alignment provides the necessary domain knowledge foundation to support and incentivize complex reasoning, which is further strengthened through reinforcement learning. Experiments on expert-annotated benchmarks and dental subsets of medical VQA benchmarks show that DentalGPT achieves superior performance on disease classification and dental VQA tasks, outperforming many state-of-the-art MLLMs despite its compact 7B parameter scale.

Qingyu Ren, Tianjun Pan, Xingzhou Chen, Xuhong Wang

Large language models have achieved remarkable progress in text generation but still struggle with generative writing tasks. In terms of evaluation, existing evaluation benchmarks include few requirement types and writing reward models are not evaluated. In terms of training, existing studies often enhance writing ability through reinforcement learning with verifiable rewards (RLVR). Howerver, existing reward model training remains coarse-grained. To address these issues, we introduce W²Bench, a comprehensive evaluation benchmark, and WRL, a fine-grained training framework. W²Bench covers five task categories and seven requirement types, enabling systematic evaluation of both writing and writing reward models by measuring the correlation between reward rankings and golden rankings. WRL constructs positive and negative samples by dropping instruction requirements to construct positive and negative examples, allowing more precise reward model training. Experiments show that our models achieve substantial improvements on various writing benchmarks and exhibit strong generalization. We will release our code and data to support future research.

Tianhe Zhang, Yonghong Deng, Ping Jian, Zhen Yang, Boyang Wang, Xinyue Zhang

While LLMs demonstrate impressive reasoning capabilities, their internal decision dynamics remain opaque. To render these process interpretable and intervenable, we propose Dynamic Entropy Tracing, a mechanism-aware framework that interprets the evolving "choice state" of attention heads during CoT generation through stepwise head-wise option-logit and entropy tracing. Our analysis reveals distinct functional behaviors at attention heads: Steadfast Heads, characterized by consistently low entropy and producing a sharp, option-selective logit pattern with a stable top choice, and Wavering Heads, characterized by consistently high entropy and producing flat or oscillatory option logits without a persistent winner. Leveraging these traces, we identify a set of intervention targets and perform Selective Head Fine-Tuning, updating solely these selected heads against a frozen backbone. Experiments across the LLaMA and Qwen families reveal a striking plasticity hierarchy: fine-tuning just 30 Wavering Heads recovers over 98% of the performance achieved by full-parameter tuning, and in some settings modestly exceeds it. In contrast, intervening on Steadfast Heads yields much less gains. Our findings translate process-level mechanistic observables into a principled criterion for selective fine-tuning, offering a fundamental insight: the most effective tuning knobs are not the components that signal the final decision, but those that retain uncertainty, and thus plasticity, during its formation.

Hsiu-Yuan Huang, Chenming Tang, Weijie Liu, Clive Bai, Saiyong Yang, Yunfang Wu

Existing Reinforcement Learning from Verifiable Rewards (RLVR) methods, such as Group Relative Policy Optimization (GRPO), have achieved remarkable progress in improving the reasoning capabilities of Large Reasoning Models (LRMs). However, they exhibit limited exploration due to reliance on on-policy rollouts which are confined to the current policy’s distribution, resulting in narrow trajectory diversity. Recent approaches attempt to expand policy coverage by incorporating trajectories generated from stronger expert models, yet this reliance increases computational cost and such advanced models are often inaccessible. To address these issues, we propose In-Context Steered Policy Optimization (ICPO), a unified framework that leverages the inherent in-context learning capability of LRMs to provide expert guidance using existing datasets. ICPO introduces mixed-policy GRPO with implicit expert forcing, which expands exploration beyond the current policy distribution without requiring advanced LRM trajectories. To further stabilize optimization, ICPO integrates expert region reject sampling to filter unreliable off-policy trajectories and annealed expert-bonus reward shaping to balance early expert guidance with later autonomous improvement. Results demonstrate that ICPO consistently enhances RLVR performance and training stability on mathematical reasoning benchmarks, revealing a scalable and effective RLVR paradigm for LRMs. Our code is available at https://github.com/Celine-hxy/ICPO.

Sun Hui, Ding Yanfeng, Huidong Ma, Chang Xu, Keyan Jin, Lizheng Zu, Cheng Zhong, Xiaoguang Liu, Gang Wang, Wentong Cai

Lossless compression has made significant advancements in Genomics Data (GD) storage, sharing and management. Current learning-based methods are non-evolvable with problems of low-level compression modeling, limited adaptability, and user-unfriendly interface. To this end, we propose AgentGC, the first evolutionary Agent-based GD Compressor, consisting of 3 layers with multi-agent named Leader and Worker. Specifically, the 1) User layer provides a user-friendly interface via Leader combined with LLM; 2) Cognitive layer, driven by the Leader, integrates LLM to consider joint optimization of algorithm-dataset-system, addressing the issues of low-level modeling and limited adaptability; and 3) Compression layer, headed by Worker, performs compression decompression via a automated multi-knowledge learning-based compression framework. On top of AgentGC, we design 3 modes to support diverse scenarios: CP for compression-ratio priority, TP for throughput priority, and BM for balanced mode. Compared with 14 baselines on 9 datasets, the average compression ratios gains are 16.66%, 16.11%, and 16.33%, the throughput gains are 4.73x, 9.23x, and 9.15x, respectively.

Bo Li, Shikun Zhang, Wei Ye

Instruction-tuned language models increasingly rely on large multi-turn dialogue corpora, but these datasets are often noisy and structurally inconsistent, with topic drift, repetitive chitchat, and mismatched answer formats across turns. We address this from a data selection perspective and propose MDS (Multi-turn Dialogue Selection), a dialogue-level framework that scores whole conversations rather than isolated turns. MDS combines a global coverage stage that performs bin-wise selection in the user-query trajectory space to retain representative yet non-redundant dialogues, with a local structural stage that evaluates within-dialogue reliability through entity-grounded topic grounding and information progress, together with query-answer form consistency for functional alignment. MDS outperforms strong single-turn selectors, dialogue-level LLM scorers, and heuristic baselines on three multi-turn benchmarks and an in-domain Banking test set, achieving the best overall rank across reference-free and reference-based metrics, and is more robust on long conversations under the same training budget. Code and resources are included in the supplementary materials.

Mohammed Ali, Abdelrahman Abdallah, Amit Agarwal, Hitesh Laxmichand Patel, Adam Jatowt

Existing benchmarks treat multi-turn conversation and reasoning-intensive retrieval separately, yet real-world information seeking requires both. To bridge this gap, we present a benchmark for reasoning-based conversational information retrieval comprising 707 conversations (2,971 turns) across eleven domains. To ensure quality, our Decomposition-and-Verification framework transforms complex queries into fact-grounded multi-turn dialogues through multi-level validation, where atomic facts are verified against sources and explicit retrieval reasoning is generated for each turn. Comprehensive evaluation reveals that combining conversation history with reasoning doubles retrieval performance (Baseline .236 \rightarrow History+Reasoning .479 nDCG@10), while reasoning-specialized models substantially outperform dense encoders. Despite these gains, further analysis highlights that implicit reasoning remains challenging, particularly when logical connections are not explicitly stated in the text. [<https://github.com/RECOR-Benchmark/RECOR>]

Dianqing Lin, Tian Lan, Jiali Zhu, Jiang Li, Wei Chen, Xu Liu, Aruukhan, Xiangdong Su, Hongxu Hou, Guanglai Gao

该来源暂未提供摘要。

Hongjin Qian, Zhao Cao, Zheng Liu

Complex reasoning in tool-augmented agent frameworks is inherently long-horizon, causing reasoning traces and transient tool artifacts to accumulate and strain the bounded working context of large language models. Without explicit memory mechanisms, such accumulation disrupts logical continuity and undermines task alignment. This positions memory not as an auxiliary efficiency concern, but as a core component for sustaining coherent, goal-directed reasoning over long horizons.We propose MemoBrain, an executive memory model for tool-augmented agents that constructs a dependency-aware memory over reasoning steps, capturing salient intermediate states and their logical relations. Operating as a co-pilot alongside the reasoning agent, MemoBrain organizes reasoning progress without blocking execution and actively manages the working context. Specifically, it prunes invalid steps, folds completed sub-trajectories, and preserves a compact, high-salience reasoning backbone under a fixed context budget. Together, these mechanisms enable explicit cognitive control over reasoning trajectories rather than passive context accumulation.We evaluate MemoBrain on challenging long-horizon benchmarks, including GAIA, WebWalker, and BrowseComp-Plus, demonstrating consistent improvements over strong baselines.

Jipeng Qiang, Jiahao Zhu, Yi Zhu, Chaowei Zhang

Live-stream E-commerce faces significant challenges from morphs, deliberate linguistic variants used to evade real-time voice filters and amplify product claims illegally. While critical for regulatory enforcement, Live Auditory Morph Resolution (LiveAMR) research is hindered by limited datasets: prior work relied on narrow, redundant health domain corpora, restricting model robustness. To bridge this gap, we introduce two datasets: (1) HealthAMR, a refined health-domain corpus via deduplication and re-annotation. (2) GeneralAMR, a general domain benchmark with 28K annotated sentences from 77 channels across 7 E-commerce categories. Further, we propose JointMRE, a multi-task framework that jointly resolves morphs and generates structured explanations, transferring grammatical insights from large language models to enhance generalization. Predictions are refined by our Conflict-aware Dual-output Refinement Framework (CDRF), which detects inconsistencies between corrections and explanations. Experiments show CDRF significantly improves morph resolution accuracy and interpretability. Our datasets and code are available [<https://anonymous.4open.science/r/Morph-Resolution-Datasets-and-Methods-611E>].

Junyoung Koh, Jaeyun Lee, Soo Yong Kim, Gyu Hyeong Choi, Jung In Koh, Jordan Phillips, Yeonjin Lee, Min Song

Recent work on music question answering (Music-QA) has primarily focused on single-track understanding, where models answer questions about an individual audio clip using its tags, captions, or metadata. However, listeners often describe music in comparative terms, and existing benchmarks do not systematically evaluate reasoning across multiple tracks. Building on the Jamendo-QA dataset, we introduce Jamendo-MT-QA, a dataset and benchmark for multi-track comparative question answering. From Creative Commons-licensed tracks on Jamendo, we construct 36,519 comparative QA items over 12,173 track pairs, with each pair yielding three question types: yes/no, short-answer, and sentence-level questions. We describe an LLM-assisted pipeline for generating and filtering comparative questions, and benchmark representative audio-language models using both automatic metrics and LLM-as-a-Judge evaluation.

Chia Hsiang Kao, Wenting Zhao, Cheryl Lam, Aarush Umap, Shreelekha Revankar, Samuel Speas, Snehal Bhagat, Rajeev Datta, Cheng Perng Phoo, Utkarsh Mall 等

Earth Observation (EO) provides critical planetary data for environmental monitoring, disaster management, climate science, and other scientific domains. In this work we ask: Are AI systems ready for reliable Earth Observation? To answer this, we introduce **UnivEARTH**, a coding benchmark of 408 yes/no questions from NASA Earth Observatory articles across 7 various topics and over 15 satellite instruments and sources. Using Google Earth Engine API as a tool in a zero-shot setup, LLM agents achieve an accuracy of 40.0% where the code fails to run over 44% of the time. To better understand LLM agent behavior, we also analyze the impact of using the JavaScript API versus Python and the effect of providing documentation. Furthermore, we find that using a reflexion framework significantly reduces errors: Claude-4.5-Sonnet, Gemini-2.5-Pro, and GPT-5 accuracies rise to around 60%. However, these results remain only marginally above random chance. Taken together, our findings identify significant challenges to be solved before AI agents can automate earth observation, and suggest paths forward.

Zhen Yang, Xinyue Zhang, Ping Jian, Chengzhi Li, Zhongbin Guo, Jiaping Feng, Wenpeng Lu

Despite the remarkable performance across numerous tasks, Large Language Models (LLMs) still exhibit notable deficiencies in temporal reasoning, even in simple event ordering tasks. For instance, a slight alteration in the temporal phrasing of the question (e.g., changing "Is event A before B?” to "Is event A after B?") can lead LLMs to hallucinate and produce inconsistent answers, reflecting a lack of robust temporal reasoning. Although many prior studies have focused on benchmarking and improving the temporal reasoning ability of LLMs, little is known about the intrinsic mechanisms within LLMs when performing temporal reasoning. In this work, we investigate the mechanistic interpretability of temporal ordering within event temporal reasoning through a structured "Identify-Interpret-Verify” pipeline. We first employ path patching to identify a sparse subset of attention heads that are causally responsible for reasoning outcomes. Detailed pattern analysis reveals that these key heads specialize in attending to either temporal keywords (semantic cues) or structural delimiters (syntactic cues). Furthermore, we rigorously validate the observed mechanism through comprehensive intervention-based experiments, ranging from head ablation to targeted attention modulation. We demonstrate that dynamically modulating the attention of these specific heads can robustly enhance model performance, which serves as strong empirical evidence that our identified mechanism faithfully captures the internal logic of temporal ordering in LLMs.

Hao An, Yibin Lou, Jiayi Guo, Yang Xu

Large language models (LLMs) often exhibit hallucinations due to their inability to accurately perceive their own knowledge boundaries. Existing abstention fine-tuning methods typically partition datasets directly based on response accuracy, causing models to suffer from severe label noise near the decision boundaries and consequently exhibit high rates of abstentions or hallucinations. This paper adopts a latent space representation perspective, revealing a "gray zone" near the decision hyperplane where internal belief ambiguity constitutes the core performance bottleneck. Based on this insight, we propose the **GeoDe** (**Geo**metric **De**noising) framework for abstention fine-tuning. This method constructs a truth hyperplane using linear probes and performs "geometric denoising" by employing geometric distance as a confidence signal for abstention decisions. This approach filters out ambiguous boundary samples while retaining high-fidelity signals for fine-tuning. Experiments across multiple models (Llama3, Qwen3) and benchmark datasets (TriviaQA, NQ, SciQ, SimpleQA) demonstrate that GeoDe significantly enhances model truthfulness and demonstrates strong generalization in out-of-distribution (OOD) scenarios. Code is available at https://github.com/Notbesidemoon/GeoDe.

Hwanjun Song

Much of the alignment tuning literature is organized around optimization objectives, while the construction of alignment data is often treated implicitly. In this survey, we adopt a data centric perspective and reframe alignment tuning as a pipeline design problem. We decompose alignment data construction into three interacting stages, response synthesis, preference evaluation, and preference instantiation, and use this framework to organize existing alignment methods into a unified taxonomy. Through this lens, we identify recurring design trade-offs and failure modes observed across prior alignment methods, and distill a set of high level principles that clarify how pipeline design choices influence the resulting optimization signal. Finally, we outline open challenges for alignment data pipelines, including prompt-level alignment, agentic settings, and alignment under evolving objectives.

Hongjin Qian, Siqi Bao, Zhao Cao, Zheng Liu

Information seeking bridges the knowledge gap between a query and its answer. Although LLMs perform well broadly, their ability to close this gap is limited by pretraining and degrades on specialized or up-to-date queries. A common remedy augments LLMs with external knowledge, either by injecting retrieved evidence into context or interleaving retrieval with reasoning. The former limits exploration of layered dependencies, while the latter is bounded by context length, constraining efficiency and scalability. For complex tasks with intricate dependencies and large text volumes, both approaches become inadequate.To tackle this bottleneck, we present AWARE (Agentic Knowledge Warehouse), an agentic knowledge warehousing framework that transforms heterogeneous, unstructured data into minimal, task-conditioned knowledge representations consumable by LLMs. Rather than exposing raw text, AWARE constructs knowledge through intent planning, online multi-threaded exploration, and map-reduce evidence integration, producing compact, LLM-ready context under finite budgets. Specifically, it applies offline document structuring to generate document headers that support controlled access, performs exploration with targeted refinement to recover layered information dependencies, and integrates distributed evidence into task-aware representations for downstream answer generation. Experiments on GAIA, WebWalker, and BrowseComp-Plus show improvements over all baselines

Xiaojie Gu, Ziying Huang, Weicong Hong, Jian Xie, Renze Lou, Kai Zhang

Large Language Models (LLMs) internalize vast world knowledge as parametric memory, yet inevitably inherit the staleness and errors of their source corpora. Consequently, ensuring the reliability and malleability of these internal representations is imperative for trustworthy real-world deployment. Knowledge editing offers a pivotal paradigm for surgically modifying memory without retraining. However, while recent editors demonstrate high success rates on standard benchmarks, it remains questionable whether current evaluation frameworks that rely on assessing output under specific prompting conditions can reliably authenticate genuine memory modification. In this work, we introduce a rigorous diagnostic framework that subjects models to discriminative self-assessment under diverse contextual pressures, specifically designed to scrutinize the subtle behavioral nuances induced by memory modifications. This probing reveals a pervasive phenomenon of Surface Compliance, where editors achieve high benchmark scores by merely mimicking target outputs without structurally overwriting internal beliefs. Moreover, we find that recursive modifications accumulate representational residues, triggering cognitive instability and permanently diminishing the reversibility of the model’s memory state. These insights underscore the risks of current editing paradigms and highlight the pivotal role of robust memory modification in building trustworthy, long-term sustainable LLM systems.

Xuchen Li, Jing Chen, Xuzhao Li, Hao Liang, Xiaohuan Zhou, Taifeng Wang, Wentao Zhang

In mathematical reasoning tasks, the advancement of Large Language Models (LLMs) relies heavily on high-quality training data with clearly defined and well-graded difficulty levels. However, existing data synthesis methods often suffer from limited diversity and lack precise control over problem difficulty, making them insufficient for supporting efficient training paradigms such as curriculum learning. To address these challenges, we propose MathMixup, a novel data synthesis paradigm that systematically generates high-quality, difficulty-controllable mathematical reasoning problems through hybrid and decomposed strategies. Automated self-checking and manual screening are incorporated to ensure semantic clarity and a well-structured difficulty gradient in the synthesized data. Building on this, we construct the MathMixupQA dataset and design a curriculum learning strategy that leverages these graded problems, supporting flexible integration with other datasets. Experimental results show that MathMixup and its curriculum learning strategy significantly enhance the mathematical reasoning performance of LLMs. Fine-tuned Qwen2.5-7B achieves an average score of 52.6% across seven mathematical benchmarks, surpassing previous state-of-the-art methods. These results fully validate the effectiveness and broad applicability of MathMixup in improving the mathematical reasoning abilities of LLMs and advancing data-centric curriculum learning.