论文检索

Zhaoyang Han, Yihe Liu, Kai Zhang, Ping Li

The rapid discovery of jailbreak prompts has revealed the alarming fragility of safety alignment in frontier large language models (LLMs). While jailbreak techniques play a critical role in red-teaming and safety evaluation, existing methods exhibit three key limitations: (i) poor transferability across model families, requiring model-specific manual tuning; (ii) heavy reliance on large-scale prompt enumeration or exhaustive search, causing prohibitive query costs and poor scalability; and (iii) high sensitivity to input preprocessing or refusal-oriented fine-tuning, leading to attack failures once the underlying model is updated. To address these, we propose Experience-driven Multi-agent Jailbreak Optimization (EMJO), which couples three collaborating agents (Attacker, Analyzer, and Judge) into a closed-loop “probe–evaluate–revise” process, together with a dynamic experience bank accumulating high-quality successful prompts and reusable strategy patterns across iterations and tasks. This design enables query-efficient and transferable jailbreak optimization under black-box access. Extensive experiments on diverse LLMs demonstrate that EMJO consistently outperforms existing black-box jailbreak baselines, achieving up to 11% absolute improvement in attack success rate while reducing the average query cost by up to 7.9\times across two benchmark datasets. These results indicate that EMJO offers an effective and scalable paradigm for systematic jailbreak discovery.

Sifan Wu, Huan Zhang, Yizhan Li, Farshid Effaty, Hongyuan Mei, Amirreza Ataei, Bang Liu

The emergence of Multimodal Large Language Models (MLLMs) that integrate vision and language modalities has unlocked new potentials for scientific reasoning, outperforming prior benchmarks in both natural language and coding domains. Current materials science evaluation datasets such as MaScQA and SciQA remain largely text-based and fail to capture the visual and research-level analytic complexity required in materials discovery and design. We introduce MatVQA, a scalable benchmark specifically designed to address this gap. Generated via an automated pipeline, MArxivAgent, from recent materials literature, MatVQA features 1672 questions across four critical structure-property-performance (SPP) reasoning tasks. Uniquely, MatVQA employs an iterative process to eliminate textual shortcuts, compelling MLLMs to perform fine-grained, low-level visual analysis of material imagery (e.g., microscopy, diffraction patterns) integrated with multi-step scientific reasoning. Benchmarking 19 open- and closed-source MLLMs on MatVQA reveals substantial gaps in current multimodal reasoning capabilities. The MatVQA benchmark is publicly available[<https://huggingface.co/datasets/trqcbf/matvqa_v2>] to facilitate further research on applying MLLMs to complex materials science problems.

Xiaoyong Mei, Tingting Zuo, Da Chen, Guangyu Hu, Xiangyu Wen, Chao Duan, Mingyan Zhang, Fudan Zheng

Multi-role dialogue summarization requires modeling complex interactions among multiple speakers while preserving role-specific information and factual consistency. However, most existing methods optimize for automatic metrics such as ROUGE and BERTScore, which favor surface-level imitation of references rather than genuine gains in faithfulness or alignment with human preferences. We propose a novel framework that couples explicit cognitive-style reasoning with reward-based optimization for multi-role dialogue summarization. Our method first distills structured reasoning traces (e.g., step-by-step inferences and intermediate reflections) from a large teacher model and uses them as auxiliary supervision to initialize a reasoning-aware summarizer via staged supervised fine-tuning. It then applies GRPO with a dual-principle reward that blends metric-based signals with human-aligned criteria targeting key information coverage, implicit inference, factual faithfulness, and conciseness. Experiments on multilingual multi-role dialogue benchmarks show that our method matches strong baselines on ROUGE and BERTScore. Specifically, results on CSDS confirm the framework’s stability in semantic consistency, while in-depth analysis on SAMSum demonstrates clear gains in factual faithfulness and model-based preference alignment. These findings underscore the value of reasoning-aware and preference-aware training for reliable dialogue summarization. Code will be made accessible upon acceptance, checkpoints and datasets are now available at https://huggingface.co/NebulaPixel.

Zhuoqun Li, Zhaopei Huang, Wenxuan Wang, Qin Jin

Understanding multimodal user personas in long-term dialogues is essential for building personalized and human-like dialogue systems. However, existing datasets suffer from limited persona diversity and static, overly simplified settings, making them insufficient for capturing the complexity of real-world interactions. To address these limitations, we introduce LongMP-Bench, a benchmark designed to evaluate the capabilities of models in understanding evolving user personas within long-term multimodal dialogues. We present a multi-step, scalable data construction pipeline that generates long-term interaction records centered around multimodal personas, followed by human refinement for quality assurance. The resulting dataset contains long conversations from 150 users, each exhibiting visual consistency and dynamic persona development over time. Built on this dataset, we define a suite of tasks to comprehensively assess models’ ability to track persona evolution, integrate visual and textual inputs, and apply persona understanding in realistic dialogue scenarios. Extensive experiments on LongMP-Bench highlight the substantial challenges in multimodal persona understanding, especially in tracking persona shifts and leveraging multimodal context effectively. We will release our benchmark and code to facilitate future research in multimodal and personalized dialogue systems.

Tian Xueyun, Wei Li, Bingbing Xu, Heng Dong, Yuanzhuo Wang, Huawei Shen

Recent Omni-multimodal Large Language Models show promise in unified audio, vision, and text modeling. However, streaming audio-video understanding remains challenging, as existing approaches suffer from disjointed capabilities: they typically exhibit incomplete modality support or lack autonomous proactive monitoring. To address this, we present ROMA, **a real-time omni-multimodal assistant for unified reactive and proactive interaction**. ROMA processes continuous inputs as synchronized multimodal units, aligning dense audio with discrete video frames to handle granularity mismatches. For online decision-making, we introduce a lightweight *speak head* that decouples response initiation from generation to ensure precise triggering without task conflict. We train ROMA with a curated streaming dataset and a two-stage curriculum that progressively optimizes for streaming format adaptation and proactive responsiveness. To standardize the fragmented evaluation landscape, we reorganize diverse benchmarks into a unified suite covering both proactive (alert, narration) and reactive (QA) settings. Extensive experiments across 12 benchmarks demonstrate ROMA achieves state-of-the-art performance on proactive tasks while competitive in reactive settings, validating its robustness in unified real-time omni-multimodal understanding. Code and benchmark are available [here](https://eureka-maggie.github.io/ROMA_show/).

Xiang Li, Runhai Jiao, Ruifan Li, Dongnan Wu, Ruojiao Qiao, Lei Liu

Reinforcement learning, with its interpretable path reasoning, has emerged as a promising paradigm for multi-hop question answering over knowledge graphs. However, existing approaches suffer from two inherent limitations: (1) lacking effective intermediate guidance, agents often fall into aimless exploration when confronted with complex multi-hop questions; and (2) policy networks focus on local neighborhood information, making it difficult to anticipate the long-term consequences of decisions. To address these challenges, we propose a Progressive Planning and Reinforced Reasoning (PPRR) framework. Specifically, we introduce large language models as multi-hop reasoning planners, converting decomposed sub-question sequences into stepwise decision guidance and thereby granting the agent human-like, step-by-step problem-solving capabilities. In addition, we design a structure-aware lookahead policy network, which explicitly models inter-node dependencies along the multi-hop reasoning process and performs lookahead value evaluations for candidate actions, thereby enhancing the agent’s global state awareness and decision foresight in complex environments. Finally, we conducted extensive experiments on four public multi-hop question answering benchmarks and one domain-specific dataset. The results demonstrate that our framework surpasses state-of-the-art methods while demonstrating strong generalization.

Finn Schmidt, Jan Philip Wahle, Terry Ruas, Bela Gipp

Automatic evaluation metrics are central to the development of machine translation systems, yet their robustness under domain shift remains unclear. Most metrics are developed on the Workshop on Machine Translation (WMT) benchmarks, raising concerns about their robustness to unseen domains. Prior studies that analyze unseen domains vary translation systems, annotators, or evaluation conditions, confounding domain effects with human annotation noise.To address these biases, we introduce a systematic multi-annotator **C**ross-**D**omain **E**rror-**S**pan-**A**nnotation dataset (CD-ESA), comprising 18.8k human error span annotations across three language pairs, where we fix annotators within each language pair and evaluate translations of the same six translation systems across one seen news domain and two unseen technical domains. Using this dataset, we first find that automatic metrics appear surprisingly robust to domain-shifts at the segment level (up to 0.69 agreement), but this robustness largely disappears once we account for human label variation. Averaging annotations increases inter-annotator agreement by up to +0.11. Metrics struggle on the unseen chemical domain compared to humans (inter-annotator agreement of 0.78–0.83 vs. 0.96).We recommend comparing metric–human agreement against inter-annotator agreement, rather than comparing raw metric–human agreement alone, when evaluating across different domains.

Qian Chen, Mengqiang Hu, Xin Guo

Dialogue summarization (DS) plays a vital role in improving customer service efficiency by automatically generating concise summaries from lengthy multi-turn dialogues. However, existing studies largely overlook the fine-grained sentiment dynamics expressed by customers, and most DS datasets lack detailed sentiment annotations. These limitations hinder both accurate service quality assessment and the development of sentiment-aware summarization models. To address these challenges, we propose a three-stage approach to building an aspect-aware sentiment dataset, comprising: (1) aspect-anchored dialogue rewriting, (2) dialogue-anchored explainable label generation, and (3) label-dialogue integrated summarization. Building upon this scheme, we construct FOCUS, a \textbf{F}ine-grained customer-\textbf{O}riented \textbf{C}hinese dialog\textbf{U}e \textbf{S}ummarization dataset. FOCUS is the first Chinese dataset with 12,948 dialogues annotated for multi-level aspects, sentiment polarity, opinion content, emotions, as well as customer-oriented formatted and free-style sentiment summaries. To demonstrate the challenges and utility of FOCUS, we benchmark a range of summarization models on FOCUS and observe that current methods often exhibit misalignment between aspects and sentiments. Meanwhile, we find that a Chain-of-Thought approach can enhance faithfulness and interpretability, highlighting promising directions for future research on this dataset. FOCUS serves as a valuable resource to advance research in sentiment-aware DS and related tasks.

Yule Xie, Jiaxin Ding, Cheng Deng, Shiqing Gao, Junran Zhang, Sibo Zhang, Zeyuan Wang, Ke Wu, Xin Ding, Luoyi Fu 等

Reinforcement learning (RL) has recently shown remarkable ability to enhance reasoning in large language models (LLMs), yet its potential in scientific domains beyond mathematics remains largely unexplored. Geoscience questions couple broad factual knowledge with multi-step inference and often rely on visual evidence such as maps, cross-sections, and diagrams, making them a challenging but verifiable testbed for RL-based reasoning. To enable this study, we introduce GeoMC-10K, a dataset of 10,000 geoscience multiple-choice questions spanning physical to human geography and high-school to professional levels; over 30% of the questions are image dependent. To support text-only RL on these multimodal questions, we design GeoM2T, a multi-agent framework that converts multimodal questions into descriptive text while preserving answerability and difficulty. Fine-tuning LLaMA-3.1-8B and Qwen-3-8B with Group Relative Policy Optimization (GRPO), incorporating a factual reward mechanism, yields GR1, which achieves absolute accuracy improvements of 5.9% and 13.3%, respectively, and it generalizes to out-of-distribution geoscience benchmarks. Together, GeoMC-10K, GeoM2T, and GR1 establish a scalable benchmark and baseline for RL-enhanced geoscience reasoning.

Daixin Shu, Jian Yang, Zhenhe Wu, Xianjie Wu, Xianfu Cheng, Guan Xiangyuan, Yanghai Wang, Pengfei Wu, Tingyang Yang, Hualei Zhu 等

Tabular data is a fundamental component of real-world information systems. However, existing multilingual table benchmarks suffer from geolinguistic imbalance - overrepresenting certain languages and lacking sufficient scale for rigorous cross-lingual analysis. To address these limitations, we introduce M3TQA, which is a comprehensive framework for massively multilingual multitask table question answering, including subsequent datasets M3TQA-BENCH and M3TQA-INSTRUCT, featuring tables expanded to 97 languages from Chinese and English sources. M3TQA-BENCH includes 6,606 professionally annotated question-answering pairs across four tasks designed to evaluate nuanced table reasoning capabilities. Additionally, we synthesized the training set M3TQA-INSTRUCT in 97 languages using Large Language Model (LLM). Experiments on state-of-the-art LLMs reveal critical insights into cross-lingual generalization, demonstrating that synthetically generated, unannotated training data can significantly boost performance, particularly for low-resource languages. M3TQA establishes a new standard for multilingual table understanding, providing both a challenging evaluation platform and a scalable methodology for future research.

Midan Shim, Seokju Hwang, KaeHyun Um, Kyong-Ho Lee

Large language models still struggle with faithfulness and hallucinations despite their remarkable reasoning abilities. In Knowledge Graph Question Answering (KGQA), semantic parsing-based approaches address the limitations by understanding constraints in a user’s question and converting them into a logical form to execute on a knowledge graph. However, existing KGQA benchmarks and methods are biased toward positive and calculation constraints. Negative constraints are neglected, although they frequently appear in real-world questions. In this paper, we introduce a new task, NEgative-conSTrained (NEST) KGQA, where each question contains at least one negative constraint, and a corresponding dataset, NestKGQA. We also design PyLF, a Python-formatted logical form, since existing logical forms are hardly suitable to express negation clearly while maintaining readability. Furthermore, NEST questions naturally contain multiple constraints. To mitigate their semantic complexity, we present a novel framework named CUCKOO, specialized to multiple-constrained questions and ensuring semantic executability. CUCKOO first generates a constraint-aware logical form draft and performs schema-guided semantic matching. It then selectively applies self-directed refinement only when executing improper logical forms yields an empty result, reducing cost while improving robustness. Experimental results demonstrate that CUCKOO consistently outperforms baselines on both conventional KGQA and NEST-KGQA benchmarks under few-shot settings.

Chunyu Ye, Yunhao Zhang, Jingyuan Sun, Chong Li, Yang Zhao, Shaonan Wang

Decoding language from the human brain remains a grand challenge for Brain-Computer Interfaces (BCIs). Current approaches typically rely on unimodal brain representations, neglecting the brain’s inherently multimodal processing. Inspired by the brain’s associative mechanisms, where viewing an image can evoke related sounds and linguistic representations, we propose a unified framework that leverages Multimodal Large Language Models (MLLMs) to align brain signals with a shared semantic space encompassing text, images, and audio. A router module dynamically selects and fuses modality-specific brain features according to the characteristics of each stimulus. Experiments on various fMRI datasets with textual, visual, and auditory stimuli demonstrate state-of-the-art performance, achieving an 8.48% average improvement on the most commonly used benchmark. We further extend our framework to EEG and MEG data, demonstrating flexibility and robustness across varying temporal and spatial resolutions. To our knowledge, this is the first unified BCI architecture capable of robustly decoding multimodal brain activity across diverse brain signals and stimulus types, offering a flexible solution for real-world applications.

Yu-Shiang Huang, Yun-Yu Lee, Tzu-Hsin Chou, Che Lin, Chuan-Ju Wang

Numerical precision is critical in financial NLP, yet embedding-based semantic similarity metrics exhibit numerical blindness—failing to distinguish contradictory values within similar contexts. We introduce NASH (Numerically Aware Scoring Hueristic), a model-agnostic metric that decouples numerical verification from textual semantic evaluation through a three-stage pipeline: (1) modal separation via numeric masking, (2) dual-channel similarity estimation through masked-text similarity and context-aware numeric alignment, and (3) IDF-weighted aggregation. NASH functions as a drop-in enhancement to existing embedding-based metrics. Validated on our proposed NumFinE financial numerical evaluation benchmark and established semantic similarity datasets (STS-B, Financial-STS), NASH achieves substantial improvements in numerical sensitivity (up to +159.6% on listwise ranking) while preserving general semantic performance, establishing a reliable standard for numeracy-aware evaluation.

Qianjun Pan, Junyi Wang, Jie Zhou, Yutao Yang, Junsong Li, Kaiyin Xu, Yougen Zhou, Yihan Li, JingYuan Zhao, Qin Chen 等

To develop a reliable AI for psychological assessment, we introduce PsychEval, a multi-session, multi-therapy, and highly realistic benchmark designed to address three key challenges:**1) Can we train a highly realistic AI counselor?** Realistic counseling is a longitudinal task requiring sustained memory and dynamic goal tracking. We propose a multi-session benchmark (spanning 6-10 sessions across three distinct stages) that demands critical capabilities such as memory continuity, adaptive reasoning, and longitudinal planning. The dataset is annotated with extensive professional skills, comprising over 677 meta-skills and 4577 atomic skills. **2) How to train a multi-therapy AI counselor?** While existing models often focus on a single therapy, complex cases frequently require flexible strategies among various therapies. We construct a diverse dataset covering five therapeutic modalities alongside an integrative therapy with a unified three-stage clinical framework across six core psychological topics. **3) How to systematically evaluate an AI counselor?** We establish a holistic evaluation framework with 18 therapy-specific and therapy-shared metrics across Client-Level and Counselor-Level dimensions. To We also construct over 2,000 diverse client profiles. Extensive experimental analysis fully validates the superior quality and clinical fidelity of our dataset.Our datasets and evaluation framework are anonymously available at this repository.

Somraj Gautam, Anathapindika Dravichi, Gaurav Harit

We introduce INDOTABVQA, a benchmark for evaluating cross-lingual Table Visual Question Answering (VQA) on real-world document images in Bahasa Indonesia. The dataset comprises 1,593 document images across three visual styles (bordered, borderless, and colorful) with one or more tables, and 1,593 question-answer sets in four languages: Bahasa Indonesia, English, Hindi, and Arabic. This enables evaluation of Vision-Language Models (VLMs) in both monolingual (Bahasa documents with Bahasa questions) and cross-lingual settings (Bahasa documents with questions in other languages). We benchmark leading open-source VLMs (Qwen2.5-VL, Gemma- 3, LLaMA-3.2) and GPT-4o and reveal substantial performance gaps, particularly on structurally complex tables and in low-resource languages. Fine-tuning a compact 3B model and a LoRA- finetuned 7B model on our dataset yields 11.6% and 17.8% improvements in accuracy. Providing explicit table region coordinates as additional input further improves performance by 4-7%, demonstrating the value of Spatial priors for table-based reasoning. Our findings underscore the importance of language- diverse, domain-specific datasets and demonstrate that targeted fine-tuning can significantly enhance VLM performance on specialized document understanding tasks. INDOTABVQA provides a valuable resource for advancing research in cross-lingual, structure-aware document understanding, especially in underrepresented regions of the world. The dataset is publicly available via Hugging Face at: https://huggingface.co/datasets/NusaBharat/INDOTABVQA.

Yamen Ajjour, Carlotta Quensel, Nedim Lipka, Henning Wachsmuth

Argumentation skills are an essential toolkit for large language models (LLMs). These skills are crucial in various use cases, including self-reflection, debating collaboratively for diverse answers, and countering hate speech. In this paper, we create the first benchmark for a standardized evaluation of LLM-based approaches to computational argumentation, encompassing 33 datasets from previous work in unified form. Using the benchmark, we evaluate the generalizability of five LLM families across 46 computational argumentation tasks that cover mining arguments, assessing perspectives, assessing argument quality, reasoning about arguments, and generating arguments. On the benchmark, we conduct an extensive systematic analysis of the contribution of few-shot examples, reasoning steps, model size, and training skills to the performance of LLMs on the computational argumentation tasks in the benchmark.

Omri Asscher, Arif Ahmad, Ananya Agrawal, Monojit Choudhury

Translation is a fundamentally value-laden process that requires the translator to make decisions and judgments that have ethical implications. However, even though large language models (LLMs) are increasingly used for translation tasks, LLMs have not been systematically examined for their default ethical tendencies or their abilities to employ and prioritize specified ethical approaches in conflicted translation situations. To address this gap, we present ETHICA-MT, a framework for examining ethical reasoning and implementation in LLM-based machine translation. Drawing on diverse ethical approaches from the translation studies literature, we formalize a conceptual framework and construct a multilingual benchmark, ETHICA-MT BENCH, that covers six languages and highlights ethical conflicts arising from competing ethical approaches in a variety of translation scenarios. Our empirical study shows that current models predominantly default to an ethical stance favoring ‘faithful representation’ to the source text, and vary in their ability to implement specified ethics at the expense of others. Finally, we highlight the basic challenges of automatically and manually evaluating the models’ ethical stances.

Zheng Li, Hongxin Ding, Chenyu Zhang, Weimin Xiong, Dawei Zhu, Sujian Li

Personality detection aims to label an individual’s traits via identifying linguistic cues from his or her written text. Previous approaches typically perform a direct mapping between text and trait labels or apply static reasoning to this task.In this paper, we argue that dynamic reasoning, underpinned by psychological theory, is essential for personality trait inference. To address this, we propose PsyPath, a novel framework that models personality detection as a process of psychologically-guided self-exploration. By enabling large language models (LLMs) to dynamically generate and answer psychologically meaningful questions, our method creates a dynamic reasoning path to explore the underlying dimensions of personality traits. This mechanism not only makes the reasoning process transparent, but also helps the model understand personality nuances in a way that mirrors expert psychological reasoning.For the "guided self-exploration", we propose a novel hybrid scoring mechanism to step-by-step evaluate the generated nodes in the reasoning paths that balances psychological coherence (black-box scoring) and model output dynamics (white-box scoring). This reasoning-based formulation inherently reflects how psychologists assess personality, as they rely on iterative, diagnostic reasoning. Experiments on two benchmark datasets demonstrate that PsyPath consistently outperforms strong baselines, yielding improvements in predictive accuracy and model interpretability.Moreover, the generated reasoning paths provide psychologically meaningful training data, significantly improving performance and psychologically grounded interpretability in downstream tasks.

Axel Delaval, Shujian Yang, Haicheng Wang, Han Qiu, Jialiang LU

Detecting toxic content using language models is crucial yet challenging. While substantial progress has been made in English, toxicity detection in French remains underdeveloped, primarily due to the lack of culturally relevant, human-annotated, large-scale datasets. In this work, we release TOXIFRENCH, a dataset of 53,622 French online comments, together with a 1,388-sample balanced benchmark split for systematic evaluation. The dataset is constructed via a semi-automated annotation pipeline that reduces manual labeling to only 10% through high-confidence LLM-based pre-annotation and human verification, while ensuring statistically near-perfect alignment with human-only annotation. We then benchmark a broad range of models and uncover a counterintuitive insight: Small Language Models (SLMs) often surpass larger models in robustness and generalization on this task. Motivated by this finding, we propose a novel Chain-of-Thought (CoT) fine-tuning strategy using a dynamic weighted loss that progressively emphasizes the model’s final decision, significantly improving faithfulness. Our fine-tuned 4B model (Qwen3-4B) achieves state-of-the-art performance on the benchmark, improving its balanced accuracy by 10% over its baseline and achieving better performance than GPT-4o and DeepSeek-R1 on our benchmark, while successfully retaining cross-lingual capabilities.

Yanyan Luo, Xue Han, Chunxu Zhao, Ruiqiao Bai, Yaxing Zhang, Qian Hu, Lijun Mei, Junlan Feng

While LLMs enable personalized chatbots, their effectiveness in child-centered personalization remains unclear, as systematic evaluation of child-specific preferences is still lacking. To address this gap, we introduce ChildEval, a benchmark for evaluating LLMs’ ability to infer and follow child-centered preferences in long-context conversations. ChildEval contains 29K synthesized persona profiles of children aged 3–6, providing relatively static background information. Each persona is associated with a child preference—which may align with, conflict with, or be independent of the persona—expressed either explicitly in a single sentence or implicitly through 6–10 turn dialogues. Explicit and implicit preferences are designed to reflect the same underlying preference but differ in expression, capturing dynamic aspects of preference expression rather than changes in the static persona. The benchmark spans five top-level and fourteen sub-level categories covering children’s daily lives and development. We further propose fine-grained, child-centric evaluation protocols to systematically assess open-source LLMs. Experimental results demonstrate how different personalized representations affect LLM responses and suggest that finetuning on ChildEval can enhance child-centered performance. Our code and dataset are available at https://github.com/ziyanluo/ChildEval.