论文检索

Yongfeng Huang, Ruiying Chen, James Cheng

Retrieval-Augmented Generation (RAG) is widely employed to mitigate risks such as hallucinations and knowledge obsolescence in medical question answering, yet its predominantly single-round, static retrieval paradigm misaligns with the multi-stage process of clinical reasoning. This compressed workflow induces two structural deficiencies: question-to-query translation often lacks clinically grounded semantic interpretation, and retrieval lacks iterative sufficiency feedback, making it difficult to form reliable evidence chains. We argue that both issues stem from a deeper cause—overloading a single reasoning chain with heterogeneous tasks of interpretation, exploration, and adjudication—and that the remedy is to reconstruct the workflow via task decoupling and dynamic multi-round exploration. To this end, we propose the Self-Evolving Multi-Agent framework **SEMA-RAG**, which assigns these roles to three specialist agents: **Interpreter Agent** for clinical schema interpretation, **Explorer Agent** for sufficiency-driven self-evolving retrieval, and **Arbiter Agent** for evidence adjudication and answer selection. Across five benchmarks and five LLM backbones, SEMA-RAG improves the strongest baseline by **+6.46** accuracy points on average, measured per backbone.

Xinyi Zeng, Xue Yang, Jingyuan Zhang, Huanqian Yan, Xiang Chen, Kaiwen Wei, Hankun Kang, Yu Tian

Multimodal large language models (MLLMs) are gaining increasing attention. Due to the heterogeneity of their input features, they face significant challenges in terms of jailbreak defenses. Current defense methods rely on costly fine-tuning or inefficient post-hoc interventions, limiting their ability to address novel attacks and involving performance trade-offs. To address the above issues, we explore the endogenous safety capabilities within MLLMs and quantify their intrinsic ability to discern harmfulness at both encoding and decoding stages. We observe that 1) MLLMs can distinguish the harmful and harmless inputs during decoding process, 2) Image-based attacks are more stealthy. Based on these insights, we introduce SafeSteer, a decoding-level defense mechanism for MLLMs. Specifically, it employs a lightweight discriminator, based on the MLLM’s own discriminative ability, to iteratively steer the decoding process toward safety. A safety alignment vector is also integrated to handle complex multimodal threats. Experiments on multiple MLLMs demonstrate that our proposed method can improve safety performance by up to 33.40% without fine-tuning.

Yiheng Tao, Kaiwen Cheng, Zhiwei Nie, Chang Liu, Jie Chen

Generative commonsense reasoning (GCR) requires models to synthesize coherent narratives that simultaneously satisfy lexical constraints and commonsense logic. Although ensemble-based LLM strategies are widely adopted to alleviate the fragility of single-chain reasoning, we uncover a counterintuitive homogeneity trap in GCR. Specifically, we observe that increasing the number of reasoning chains can degrade performance, as the generated chains tend to collapse into a narrow semantic region, thereby reinforcing shared biases rather than providing complementary evidence. We posit that escaping this trap requires fundamentally broadening semantic coverage via heterogeneous sources. Our investigation into the nature of diversity reveals that deep semantic diversity, rather than surface-level lexical variation, is the decisive prerequisite for effective integration. Motivated by this insight, we propose an Explore-then-Integrate framework, in which high–semantic-entropy explorers capture diverse concept bindings, and a powerful integrator performs compositional synthesis to merge valid fragments into coherent narratives. Crucially, to ensure that the observed performance gains arise from accurate logical composition rather than trivial best-candidate selection, we introduce a provenance-aware evaluation suite that explicitly quantifies the heterogeneous origins of synthesized outputs. Extensive experiments on multiple benchmarks demonstrate the consistent superiority of our approach across a range of metrics. Notably, our method achieves over 10% improvement in overall accuracy on NoRa and in SPICE score on CommonGen-Lite.

Han Jang, Junhyeok Lee, Heeseong Eum, Kyu Sung Choi

Medical Vision-Language Models (Med-VLMs) have achieved expert-level proficiency in interpreting diagnostic imaging. However, current models are predominantly trained on professional literature, limiting their ability to communicate findings in the lay register required for patient-centered care. While text-centric research has actively developed resources for simplifying medical jargon, there is a critical absence of large-scale multimodal benchmarks designed to facilitate lay-accessible medical image understanding. To bridge this resource gap, we introduce MedLayBench-V, the first large-scale multimodal benchmark dedicated to expert-lay semantic alignment. Unlike naive simplification approaches that risk hallucination, our dataset is constructed via a Structured Concept-Grounded Refinement (SCGR) pipeline. This method enforces strict semantic equivalence by integrating Unified Medical Language System (UMLS) Concept Unique Identifiers (CUIs) with micro-level entity constraints. MedLayBench-V provides a verified foundation for training and evaluating next-generation Med-VLMs capable of bridging the communication divide between clinical experts and patients.

Yibo Zhang, Kaiwen Luo, Liang Lin, Shilinlu Yan, Jin Wang, Yaoqi Guo, Yitian Chen, Yalan Qin, Zhenhong Zhou, Kun Wang 等

While Audio Large Models (ALLMs) have achieved remarkable proficiency, their robustness remains brittle in real-world deployment. Existing evaluations largely rely on synthetic Gaussian noise or simplistic single-source interference, failing to capture the intricate, multi-layered acoustic dynamics—or "Acoustic Ecology"—that characterize authentic physical environments. To bridge this ecological gap, we introduce RSA-Bench, a comprehensive robustness benchmark designed to stress-test ALLMs through high-fidelity auditory scene simulations. Unlike traditional methods, we construct evaluation samples by naturally superimposing diverse environmental soundscapes—spanning Pasture, Extreme Weather, Classroom, and Outdoors—onto clean speech signals across a spectrum of interference intensities. By evaluating models on six core tasks ranging from fundamental perception to complex reasoning, our study unveils three macro-level insights: (I) The Perception-Cognition Gap: Models maintain relative resilience in low-level recognition but suffer a functional collapse in high-order reasoning tasks under stress; (II) Scenario Sensitivity: "Vocal-like" interference (e.g., children playing) proves significantly more destructive than mechanical noise, challenging the model’s auditory attention mechanisms; and (III) The Denoising Paradox: Standard speech enhancement often exacerbates performance degradation, as ALLMs prove highly sensitive to the semantic distortions introduced by denoising artifacts.

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.

Seunghee Kim, Ingyu Bang, Seokgyu Jang, Changhyeon Kim, Sanghwan Bae, Jihun Choi, Richeng Xuan, Taeuk Kim

Multimodal Large Language Models (MLLMs) have increasingly supported omni-modal processing across text, vision, and speech. However, existing evaluation frameworks for such models suffer from critical limitations, including modality shortcuts and biased reasoning paths. To address these challenges, we propose OMHBench, a novel benchmark designed to rigorously evaluate omni-modal multi-hop reasoning. It consists of 6,144 questions with balanced reasoning paths that are jointly grounded across all three modalities. Extensive evaluation of 13 state-of-the-art models reveals that (1) a large performance gap exists between proprietary and open-source MLLMs and (2) even proprietary models exhibit high sensitivity to reasoning path variations, resulting in asymmetric omni-modal grounding. Notably, models struggle when processing the speech modality, underscoring the need for balanced, multi-hop evaluation of omni-modal intelligence.

Langming Liu, Kangtao Lv, Haibin Chen, Weidong Zhang, Yejing Wang, Shilei Liu, Xin Tong, Yujin Yuan, Yongwei Wang, Wenbo Su 等

Large language models (LLMs), despite their powerful capabilities, suffer from factual hallucinations where they generate verifiable falsehoods. We identify a root of this issue: the imbalanced data distribution in the pretraining corpus, which leads to a state of "low-probability truth" and "high-probability falsehood". Recent approaches, such as teaching models to say "I don’t know" or post-hoc knowledge editing, either evade the problem or face catastrophic forgetting. To address this issue from its root, we propose PretrainRL, a novel framework that integrates reinforcement learning into the pretraining phase to consolidate factual knowledge. The core principle of PretrainRL is "debiasing then learning." It actively reshapes the model’s probability distribution by down-weighting high-probability falsehoods, thereby making "room" for low-probability truths to be learned effectively. To enable this, we design an efficient negative sampling strategy to discover these high-probability falsehoods and introduce novel metrics to evaluate the model’s probabilistic state concerning factual knowledge. Extensive experiments on three public benchmarks demonstrate that PretrainRL significantly alleviates factual hallucinations and outperforms state-of-the-art methods.

Kun Zhou, Jiakai He, Wenmian Yang, Zhensheng Wang, Yiquan Zhang, Weijia Jia

Presentation slides are a primary medium for data-driven reporting, yet keeping complex, analytics-style decks up to date remains labor-intensive. Existing automation methods mostly follow fixed template filling and cannot support dynamic updates for diverse, user-authored slide decks. We therefore define “Dynamic Slide Update via Natural Language Instructions on User-provided Templates” and introduce DynaSlide, a large-scale benchmark with 20,036 real-world instruction–execution triples (source slide, user instruction, target slide) grounded in a shared external database and built from business reporting slides under bring-your-own-template (BYO-template) conditions. To tackle this task, we propose SlideAgent, an agent-based framework that combines multimodal slide parsing, natural language instruction grounding, and tool-augmented reasoning for tables, charts, and textual conclusions. SlideAgent updates content while preserving layout and style, providing a strong reference baseline on DynaSlide. We further design end-to-end and component-level evaluation protocols that reveal key challenges and opportunities for future research. The dataset and code are available at https://anonymous.4open.science/r/604E/.

Zheng Wu, Xingyu Lou, Xinbei Ma, Yansi Li, Weiwen Liu, Weinan Zhang, Jun Wang, Zhuosheng Zhang

Large Language Model (LLM)-based agents significantly extend the utility of LLMs by interacting with dynamic environments. However, enabling agents to continually learn new tasks without catastrophic forgetting remains a critical challenge, known as the stability–plasticity dilemma.In this work, we argue that this dilemma fundamentally arises from the failure to explicitly distinguish between common knowledge shared across tasks and conflicting knowledge introduced by task-specific interference. To address this, we propose Agent-Dice, a parameter fusion framework based on directional consensus evaluation.Concretely, Agent-Dice disentangles knowledge updates through a two-stage process: geometric consensus filtering to prune conflicting gradients, and curvature-based importance weighting to amplify shared semantics.We provide a rigorous theoretical analysis that establishes the validity of the proposed fusion scheme and offers insight into the origins of the stability–plasticity dilemma. Extensive experiments on GUI agents and tool-use agent domains demonstrate that Agent-Dice exhibits outstanding continual learning performance with minimal computational overhead and parameter updates.

Zhibo Man, Shaoyang Xu, Yujie Zhang, Yi Feng, Yuanmeng Chen, Yufeng Chen, Xu Jinan, Wenxuan Zhang

Large language models (LLMs)-based multi-agent systems have recently shown strong potential for machine translation (MT). However, their application to multi-domain translation (MDT) remains under-explored, particularly in addressing cross-domain word ambiguity. To investigate whether multi-agent approaches can help disambiguation in MDT, we propose a multi-agent collaborative disambiguation framework for MDT (MACD), which leverages the collaborative capabilities of LLMs for disambiguation. MACD consists of four cooperating agents responsible for domain allocation, general translation, domain disambiguation, and translation fusion. Experimental results show that MACD significantly improves translation performance across multiple domains and enhances disambiguation accuracy. Our approach reveals several findings on multi-agent collaboration in resolving word ambiguities.

Chenkai Xu, Xu Wang, Zhenyi Liao, Yishun Li, TianQi Hou, Zhijie Deng

Consistency models (CMs) have shown promise in the efficient generation of both image and text. This raises the natural question of whether we can learn a unified CM for efficient multimodal generation (e.g., text-to-image) and understanding (e.g., image-to-text). Intuitively, such a model could be acquired by applying the consistency distillation (CD) to existing unified multimodal models. However, the key challenge is establishing a unified denoising perspective for both image and text generation, which is essential for establishing the consistency mapping. To tackle this, at the representation level, we advocate for discrete tokens for both modalities to best preserve language modeling capabilities. Critically, instead of defining the text denoising trajectory via recent discrete diffusion language modeling principles, we specify it using the parallel decoding trace of an autoregressive language model, benefiting from the latter’s superior performance in general text generation tasks. The denoising trajectory of image tokens adheres to standard discrete diffusion. We train our unified consistency models (UniCMs) on these combined multimodal trajectories simultaneously with a unified objective. We introduce a trajectory segmentation strategy to improve the training convergence. Empirically, in text-to-image generation, UniCMs outperform SD3 on GenEval and Image Reward, while requiring only approximately {1}/{8} of the sampling time. Meanwhile, in image-to-text generation, UniCMs surpass Show-o on the MMMU benchmark while being 1.5 \times faster at long-sequence generating speed.

Zayd Muhammad Kawakibi Zuhri, Erland Hilman Fuadi, Alham Fikri Aji

We introduce softpick, a rectified, not sum-to-one, drop-in replacement for softmax in transformer attention mechanisms that eliminates attention sink and massive activations. Our experiments with 340M and 1.8B parameter models demonstrate that softpick achieves 0% sink rate consistently. The softpick transformers produce hidden states with significantly lower kurtosis and creates sparse attention maps. Quantized models using softpick outperform softmax on standard benchmarks, with a particularly pronounced advantage at lower bit precisions. Our analysis and discussion shows how softpick has the potential to open new possibilities for quantization, low-precision training, sparsity optimization, pruning, and interpretability. Our code: https://github.com/zaydzuhri/softpick-attention.

Zihan Wang, Hao Wang, Shi Feng, Xiaocui Yang, Daling Wang, Yiqun Zhang, Jinghao Lin, Xiaozhong Ji, Haihua Yang

Medical reasoning models remain constrained by parametric knowledge and are thus susceptible to forgetting and hallucinations. DeepResearch (DR) models ground outputs in verifiable evidence from tools and perform strongly in general domains, but their direct transfer to medical field yields relatively limited gains. We attribute this to two gaps: task characteristic and tool-use scaling. Medical questions require evidence interpretation in a knowledge-intensive clinical context; while general DR models can retrieve information, they often lack clinical-context reasoning and thus “find it but fail to use it,” leaving performance limited by medical abilities. Moreover, in medical scenarios, blindly scaling tool-call can inject noisy context, derailing sensitive medical reasoning and prompting repetitive evidence-seeking along incorrect paths. Therefore, we propose DeepMed. For data, we deploy a multi-hop med-search QA synthesis method supporting the model to apply the DR paradigm in medical contexts. For training, we introduce a difficulty-aware turn-penalty to suppress excessive tool-call growth. For inference, we bring a monitor to help validate hypotheses within a controlled number of steps and avoid context rot. Overall, on seven medical benchmarks, DeepMed improves its base model by 9.79% on average and outperforms larger medical reasoning and DR models.

Guixian Xu, Yide Liang, Zeli Su, Xuexian Song, Ziyin Zhang, Yushuang Dong, Ting Zhang, Xu Han

Vision–language models (VLMs) have progressed rapidly, but Tibetan remains largely underserved due to the lack of infrastructure for reproducible training and evaluation. To help address this gap, we introduce FTibSuite, a resource-centric foundation for Tibetan VLM research that provides an end-to-end training-and-evaluation workflow and includes human-verified multimodal annotations, partially filling a long-standing shortage of Tibetan multimodal resources. FTibSuite comprises FTibData, FTibBench, and a reproducible baseline model, FTibVLM, built on Qwen3-VL-8B-Instruct. FTibVLM adopts a three-stage adaptation pipeline consisting of Tibetan continual pretraining, image–text alignment, and multimodal instruction tuning. For systematic evaluation, FTibBench adapts five established multimodal benchmarks to Tibetan and offers a reproducible evaluation protocol to support consistent comparisons across models. Specifically, FTibBench includes Tibetan versions of MMBench, MME, POPE, BinaryVQA, and COREVQA. Experiments on FTibBench demonstrate that FTibVLM consistently improves Tibetan multimodal performance. For instance, FTibVLM attains 76.01 accuracy on BinaryVQA, indicating that Tibetan performance can be competitive with high-resource settings on this diagnostic task. We also observe substantial gains on other benchmarks, including an improvement on MMBench (dev) from 42.97 to 67.78 and an increase in POPE-random accuracy from 47.53 to 80.56, underscoring the practical value of the proposed workflow and resources.

Tzu-Chi Liu, Hui-Ying Yang, Shiow-Ching Shun, Yu-Chi Chen, Lu-Yen Anny Chen, Yong-Sheng Chen

Clinical dialogue systems are increasingly vital for patient education and follow-up care; however, effective responses often depend on the clinical context that patients often fail to provide in detail. Responding directly to vague messages can therefore lead to generic or clinically misaligned advice, a challenge that is particularly pronounced in post-op oral cancer (OC) care due to speech impairment and functional limitations. Moreover, post-op OC patients often experience psychological distress, making safety-aware language more likely to arise in dialogue. Dialogue systems in this setting must therefore address both clarifying missing clinical context and ensuring psychological safety. We propose a safety-aware dialogue system that applies information-gain guided clarification before RAG-based response generation and screens user utterances for emotional distress and suicidal ideation. Expert evaluations show that the proposed system improves the quality and clinical appropriateness of generated responses relative to strong baselines, while the safety module closely aligns with expert judgments on clinically concerning utterances. Furthermore, we release a clinically curated Chinese post-op OC QA dataset with expert-validated annotations, which we use throughout our experiments.

Minh-Vuong Nguyen, Fatemeh Shiri, Zhuang Li, Karin Verspoor

Large Language Models (LLMs) are increasingly being explored for clinical question answering and decision support, yet safe deployment critically requires reliable handling of patient measurements in heterogeneous clinical notes. Existing evaluations of LLMs for clinical numerical reasoning provide limited operation-level coverage, restricted primarily to arithmetic computation, and rarely assess the robustness of numerical understanding across clinical note formats. We introduce ClinicNumRobBench, a benchmark of 1,624 context-question instances with ground-truth answers that evaluates four main types of clinical numeracy: value retrieval, arithmetic computation, relational comparison, and aggregation. To stress-test robustness, ClinicNumRobBench presents longitudinal MIMIC-IV vital-sign records in three semantically equivalent representations, including a real-world note-style variant derived from the Open Patients dataset, and instantiates queries using 42 question templates. Experiments on 17 LLMs show that value retrieval is generally strong, with most models exceeding 85% accuracy, while relational comparison and aggregation remain challenging, with some models scoring below 15%. Fine-tuning on medical data can reduce numeracy relative to base models by over 30%, and performance drops under note-style variation indicate LLM sensitivity to format. ClinicNumRobBench offers a rigorous testbed for clinically reliable numerical reasoning. Code and data URL are available on https://github.com/MinhVuong2000/ClinicNumRobBench.

Bingguang Hao, Zengzhuang Xu, Maolin Wang, Yuntao Wen, Yicheng Chen, Cunyin Peng, Long Chen, Xiangyu Zhao, Jinjie Gu, Chenyi Zhuang 等

While Supervised Fine-Tuning (SFT) is the prevailing method for equipping Large Language Models (LLMs) with function calling capabilities, its effectiveness is often compromised by two critical challenges: 1) **Imbalanced Training Signals**, where lengthy Chain-of-Thought (CoT) reasoning tokens dominate the training signals over concise function calls in the learning objective, and 2) **Imbalanced Data Hardness**, characterized by a scarcity of hard training examples. To overcome these limitations, we propose Balanced Supervised Fine-tuning (**BalanceSFT**), a novel framework that incorporates two key components: a Self-adjusted Signal Balancing (SSB) loss that employs a learnable hyperparameter to dynamically adjust the token contributions of CoT reasoning and function calls, together with a Hard Data Re-sampling (HDR) strategy that establishes a feedback loop to selectively generate new, high-quality complex data guided by model errors. Extensive experiments demonstrate the effectiveness of our proposed BalanceSFT framework. With BalanceSFT, a 7B model achieves function calling performance that surpasses state-of-the-art models like GPT-5. Our code, models, and dataset are open-sourced.

Jiale Liu, Jiahao Zhang, Suhang Wang

Retrieval-Augmented Generation (RAG) is a powerful technique for enhancing Large Language Models (LLMs) with external, up-to-date knowledge. Graph RAG has emerged as an advanced paradigm that leverages graph-based knowledge structures to provide more coherent and contextually rich answers. However, the move from plain document retrieval to structured graph traversal introduces new, under-explored privacy risks. This paper investigates the data extraction vulnerabilities of the Graph RAG systems. We design and execute tailored data extraction attacks to probe their susceptibility to leaking both raw text and structured data, such as entities and their relationships. Our findings reveal a critical trade-off: while Graph RAG systems may reduce raw text leakage, they are significantly more vulnerable to the extraction of structured entity and relationship information. We also explore potential defense mechanisms to mitigate these novel attack surfaces. This work provides a foundational analysis of the unique privacy challenges in Graph RAG and offers insights for building more secure systems.

Chaoyang Wang, Kaituo Feng, Dongyang Chen, Zhongyu Wang, Zhixun Li, Sicheng Gao, Meng Meng, Xu Zhou, Manyuan Zhang, Yuzhang Shang 等

Recent advances have shown that multimodal large language models (MLLMs) benefit from multimodal interleaved chain-of-thought (CoT) with vision tool interactions. However, existing open-source models often exhibit blind tool-use reasoning patterns, invoking vision tools even when they are unnecessary, which significantly increases inference overhead and degrades model performance. To this end, we propose AdaTooler-V, an MLLM that performs adaptive tool-use by determining whether a visual problem truly requires tools. First, we introduce AT-GRPO, a reinforcement learning algorithm that adaptively adjusts reward scales based on the Tool Benefit Score of each sample, encouraging the model to invoke tools only when they provide genuine improvements. Moreover, we construct two datasets to support training: AdaTooler-V-CoT-100k for SFT cold start and AdaTooler-V-300k for RL with verifiable rewards across single-image, multi-image, and video data. Experiments across twelve benchmarks demonstrate the strong reasoning capability of AdaTooler-V, outperforming existing methods in diverse visual reasoning tasks. Notably, AdaTooler-V-7B achieves an accuracy of 89.8% on the high-resolution benchmark V*, surpassing the commercial proprietary model GPT-4o and Gemini 1.5 Pro.