Superpose Task-specific Features for Model Merging
机构 * Tsinghua University（清华大学） ； Huawei（华为） ； State Key laboratory of Space Network and Communications（空间网络与通信国家重点实验室）
Comments EMNLP 2025
期刊&会议
EMNLP
Conference on Empirical Methods in Natural Language Processing · 会议 · Natural Language Processing
RAcQUEt: Unveiling the Dangers of Overlooked Referential Ambiguity in Visual LLMs
机构 * Amsterdam UMC, Department of Medical Informatics（阿姆斯特丹大学医学中心，医学信息学系） ； Center for Information and Language Processing, LMU Munich（信息与语言处理中心，慕尼黑大学） ； Munich Center for Machine Learning (MCML), Munich（慕尼黑机器学习中心（MCML）） ； Institute for Logic, Language and Computation (ILLC), University of Amsterdam（逻辑、语言与计算研究所（ILLC），阿姆斯特丹大学）
Comments Accepted to EMNLP 2025 (Main Conference)
Unlocking Legal Knowledge: A Multilingual Dataset for Judicial Summarization in Switzerland
机构 * University of Bern（伯尔尼大学） ； Bern University of Applied Sciences（伯尔尼应用科学大学） ； Stanford University（斯坦福大学） ； University of Fribourg（弗里堡大学）
Comments Accepted to EMNLP 2025 Findings
3DS: Medical Domain Adaptation of LLMs via Decomposed Difficulty-based Data Selection
机构 * School of Computer Science, Peking University, Beijing, China（北京大学计算机学院） ； Key Laboratory of High Confidence Software Technologies, Ministry of Education（教育部高可信软件技术重点实验室） ； National Engineering Research Center for Software Engineering, Peking University, China（软件工程国家级工程研究中心） ； Big Data Technology Research Center, Nanhu Laboratory, Jiaxing, China（南湖实验室大数据技术研究中心） ； Peking University Information Technology Institute, Tianjin Binhai, China（北京大学信息科学技术研究院）
Comments Accepted to EMNLP 2025 (Main Conference)
FG-PRM: Fine-grained Hallucination Detection and Mitigation in Language Model Mathematical Reasoning
机构 * Department of Computer Science, University of Texas at Dallas（德克萨斯大学达拉斯分校计算机科学系） ； Department of Electrical Engineering Computer Science, University of Michigen（密歇根大学电子工程与计算机科学系）
Comments Findings of EMNLP 2025
Read to Hear: A Zero-Shot Pronunciation Assessment Using Textual Descriptions and LLMs
Comments EMNLP 2025 MainConference
Calibrating LLMs for Text-to-SQL Parsing by Leveraging Sub-clause Frequencies
机构 * Carnegie Mellon University（卡内基梅隆大学） ； Bloomberg（彭博）
Comments EMNLP 2025 main conference
Enhancing Multi-Agent Debate System Performance via Confidence Expression
机构 * National University of Singapore（新加坡国立大学）
Comments EMNLP'25 Findings
You Are What You Train: Effects of Data Composition on Training Context-aware Machine Translation Models
机构 * Department of Advanced Computing Sciences（先进计算科学系）
Comments EMNLP 2025 main conference
Slim-SC: Thought Pruning for Efficient Scaling with Self-Consistency
机构 * NTU Singapore（新加坡国立大学） ； KTH Royal Institute of Technology（皇家理工学院） ； Microsoft（微软公司）
Comments Accepted by EMNLP 2025 (Oral), 9 pages
Large Language Models Discriminate Against Speakers of German Dialects
机构 * Johannes Gutenberg University Mainz（明斯特-贾格施泰因大学） ； Data Science Group, University of Hamburg（汉堡大学数据科学组） ； Allen Institute for AI（人工智能研究所） ； University of Washington（华盛顿大学） ； University of Colorado Boulder（科罗拉多大学博尔德分校）
Comments Accepted to EMNLP 2025 Main
Improving Context Fidelity via Native Retrieval-Augmented Reasoning
Comments Accepted as a main conference paper at EMNLP 2025
Sparse Neurons Carry Strong Signals of Question Ambiguity in LLMs
机构 * Brown University（布朗大学） ； Drexel University（德雷塞尔大学）
Comments To be appeared in EMNLP 2025 (main)
Humor in Pixels: Benchmarking Large Multimodal Models Understanding of Online Comics
机构 * Singapore University of Technology and Design（新加坡科技设计大学）
Comments 27 pages, 8 figures, EMNLP 2025 Findings
Context Copying Modulation: The Role of Entropy Neurons in Managing Parametric and Contextual Knowledge Conflicts
机构 * BNP Paribas（BNP巴黎银行） ； Sorbonne Université（索邦大学） ； Criteo AI Lab（Criteo人工智能实验室）
Comments Accepted at EMNLP 2025
Journal ref EMNLP 2025
IntrEx: A Dataset for Modeling Engagement in Educational Conversations
机构 * Department of Computer Science, University of Warwick, UK（沃里克大学计算机科学系） ； School of Computer Science, University of Sheffield, UK（谢菲尔德大学计算机科学学院） ； Department of Psychology, University of Warwick, UK（沃里克大学心理学系）
Comments EMNLP 2025 Findings camera-ready, 9+7 pages
Training Text-to-Molecule Models with Context-Aware Tokenization
机构 * Seoul National University（首尔国立大学） ； Moloco Inc.（Moloco公司） ； Korea Advanced Institute of Science and Technology (KAIST)（韩国科学技术院）
Comments EMNLP 2025 Findings
How Does Cognitive Bias Affect Large Language Models? A Case Study on the Anchoring Effect in Price Negotiation Simulations
机构 * Kyoto University（京都大学）
Comments 18 pages, 2 figures. Accepted to EMNLP 2025 findings
Language Models Identify Ambiguities and Exploit Loopholes
机构 * UNC Chapel Hill（北卡罗来纳大学教堂山分校） ； The University of Texas at Austin（德克萨斯大学奥斯汀分校）
Comments EMNLP 2025 camera-ready; Code: https://github.com/esteng/ambiguous-loophole-exploitation
Position Bias Mitigates Position Bias:Mitigate Position Bias Through Inter-Position Knowledge Distillation
机构 * MAIS, Institute of Automation, Chinese Academy of Sciences（信息与自动化研究所，中国科学院） ； School of Artificial Intelligence, University of Chinese Academy of Sciences（中国科学院大学人工智能学院） ； AMAP, Alibaba Group（AMAP，阿里巴巴集团）
Comments EMNLP 2025 Oral
Table-Text Alignment: Explaining Claim Verification Against Tables in Scientific Papers
机构 * National Institute of Informatics, Japan（日本信息机构国家研究所） ； The University of Tokyo, Japan（日本东京大学） ； National Taiwan University（台湾国立大学） ； JFLI, CNRS, Nantes Université, France（法国CNRS-纳antes大学联合研究所）
Comments EMNLP 2025 Findings; 9 pages; code and data are available at https://github.com/Alab-NII/SciTabAlign
MythTriage: Scalable Detection of Opioid Use Disorder Myths on a Video-Sharing Platform
机构 * Princeton University（普林斯顿大学） ； University of Washington（华盛顿大学） ； Georgia Institute of Technology（佐治亚理工学院）
Comments To appear at EMNLP 2025. Please cite EMNLP version when proceedings are available
Puzzled by Puzzles: When Vision-Language Models Can't Take a Hint
机构 * POSTECH ； University of California, Berkeley（加州大学伯克利分校）
Comments EMNLP 2025 Main Conference
From Automation to Autonomy: A Survey on Large Language Models in Scientific Discovery
机构 * Department of Computer Science and Engineering, HKUST, Hong Kong SAR, China（计算机科学与工程系，香港科技大学，香港特别行政区，中国）
Comments EMNLP 2025 Main
Do Large Language Models Truly Grasp Addition? A Rule-Focused Diagnostic Using Two-Integer Arithmetic
机构 * Zhejiang University（浙江大学） ； School of Engineering, Westlake University（西湖大学工程学院）
Comments Accepted by EMNLP'25 Main
What's Not Said Still Hurts: A Description-Based Evaluation Framework for Measuring Social Bias in LLMs
机构 * George Mason University（乔治·马歇尔大学）
Comments EMNLP Findings 2025
Mind the Style Gap: Meta-Evaluation of Style and Attribute Transfer Metrics
机构 * Department of Computer Science, Aarhus University, Denmark（计算机科学系，奥胡斯大学） ； Department of Computer Science, University of Copenhagen, Denmark（计算机科学系，哥本哈根大学）
Comments Accepted at EMNLP Findings 2025
LogiDynamics: Unraveling the Dynamics of Inductive, Abductive and Deductive Logical Inferences in LLM Reasoning
Comments EMNLP 2025 Main
Mirror-Consistency: Harnessing Inconsistency in Majority Voting
机构 * Shanghai Jiaotong University（上海交通大学） ； Tiansuan Lab, Ant Group Co., Ltd.（蚂蚁集团天算实验室） ； SJTU Paris Elite Institute of Technology（上海交通大学巴黎精英理工学院）
Comments EMNLP 2024 Findings
MAgICoRe: Multi-Agent, Iterative, Coarse-to-Fine Refinement for Reasoning
机构 * UNC Chapel Hill（北卡罗来纳大学教堂山分校）
Comments EMNLP 2025 (Camera-Ready)
