Efficient Context Selection for Long-Context QA: No Tuning, No Iteration, Just Adaptive-$k$
机构 * University of Notre Dame（内布拉斯加大学达灵顿分校） ； Megagon Labs（梅加贡实验室）
Comments 26 pages, 16 tables, 5 figures. Accepted at EMNLP 2025 (Main)
期刊&会议
EMNLP
Conference on Empirical Methods in Natural Language Processing · 会议 · Natural Language Processing
Static Word Embeddings for Sentence Semantic Representation
机构 * ZOZO Research（ZOZO研究院）
Comments 17 pages; accepted to the Main Conference of EMNLP 2025
How Do Large Vision-Language Models See Text in Image? Unveiling the Distinctive Role of OCR Heads
Comments EMNLP 2025 Oral
DEBATE, TRAIN, EVOLVE: Self Evolution of Language Model Reasoning
机构 * Department of Computer Science, Virginia Tech, USA（弗吉尼亚理工大学计算机科学系）
Comments Accepted to EMNLP 2025 Main Conference
ThinkEdit: Interpretable Weight Editing to Mitigate Overly Short Thinking in Reasoning Models
机构 * UCSD CSE（加州大学圣地亚哥分校计算机科学系） ； UCSD HDSI（加州大学圣地亚哥分校人类发展与应用科学学院）
Comments Accepted to EMNLP 2025
Value Profiles for Encoding Human Variation
机构 * Department of Computer Science, University of Washington（华盛顿大学计算机科学系） ； Google DeepMind（谷歌DeepMind）
Comments EMNLP 2025
Towards Reasoning Ability of Small Language Models
机构 * Department of Computer Science, Virginia Tech, USA（弗吉尼亚理工大学计算机科学系） ； Department of Physics, Clarendon Laboratory, University of Oxford, UK（牛津大学物理系，克伦德尔实验室） ； NVIDIA Corporation（英伟达公司）
Comments Accepted to EMNLP 2025 Main Conference
SAFE-SQL: Self-Augmented In-Context Learning with Fine-grained Example Selection for Text-to-SQL
机构 * Department of Artificial Intelligence, Chung-Ang University（人工智能系， Chung-Ang 大学） ； New York University（纽约大学）
Comments EMNLP 2025 (Main Conference)
Where Fact Ends and Fairness Begins: Redefining AI Bias Evaluation through Cognitive Biases
机构 * The Chinese University of Hong Kong（香港中文大学） ； University of California, Los Angeles（加州大学洛杉矶分校）
Comments Accepted to EMNLP 2025 (Findings)
Mind the Value-Action Gap: Do LLMs Act in Alignment with Their Values?
机构 * University of Washington（华盛顿大学） ； NYU Shanghai（纽约大学上海校区） ； New York University（纽约大学）
Comments EMNLP 2025 Main Paper
Dagger Behind Smile: Fool LLMs with a Happy Ending Story
机构 * S-Lab, Nanyang Technological University, Singapore（南洋理工大学新加坡分校S实验室） ； College of Computing and Data Science, Nanyang Technological University, Singapore（南洋理工大学计算机与数据科学学院）
Comments EMNLP 2025 Findings
Turning Logic Against Itself : Probing Model Defenses Through Contrastive Questions
机构 * Ubiquitous Knowledge Processing Lab (UKP Lab), Department of Computer Science and Hessian Center for AI (hessian.AI), Technical University of Darmstadt（德累斯顿技术大学计算机科学系、海斯堡人工智能中心（hessian.AI）、通用知识处理实验室（UKP Lab））
Comments Accepted at EMNLP 2025 (Main)
Cut the Deadwood Out: Backdoor Purification via Guided Module Substitution
机构 * National University of Singapore（新加坡国立大学） ； Macquarie University（麦考瑞大学） ； University College London（伦敦大学学院） ； Monash University（墨尔本大学）
Comments Accepted to Findings of the Association for Computational Linguistics: EMNLP 2025
Watermark under Fire: A Robustness Evaluation of LLM Watermarking
Comments 25 pages. Accepted by The 2025 Conference on Empirical Methods in Natural Language Processing (EMNLP 2025)
Preemptive Detection and Correction of Misaligned Actions in LLM Agents
机构 * Ubiquitous Knowledge Processing Lab, Department of Computer Science, TU Darmstadt（图宾根大学计算机科学系） ； National Research Center for Applied Cybersecurity ATHENE, Germany（德国应用网络安全国家研究中心ATHENE） ； Department of Electrical and Computer Engineering & Ingenuity Labs Research Institute, Queen’s University, Canada（加拿大女王大学电气与计算机工程系及创新实验室研究院）
Comments Accepted by EMNLP 2025
Hardware-Aware Parallel Prompt Decoding for Memory-Efficient Acceleration of LLM Inference
机构 * Imperial College London, UK（伦敦帝国学院） ； Hong Kong Polytechnic University, Hong Kong（香港理工大学） ； Samsung AI Center, Cambridge, UK（三星人工智能中心）
Comments Accepted at EMNLP 2025. The code for this implementation is available at https://github.com/hmarkc/parallel-prompt-decoding
Multi-level Diagnosis and Evaluation for Robust Tabular Feature Engineering with Large Language Models
机构 * Computer Science and Engineering Korea University（计算机科学与工程韩国大学）
Comments Accepted to Findings of EMNLP 2025
Paired by the Teacher: Turning Unpaired Data into High-Fidelity Pairs for Low-Resource Text Generation
机构 * Center for Language and Speech Processing, Johns Hopkins University（语言与语音处理中心，约翰霍普金斯大学）
Comments Accepted at EMNLP 2025 (Main Conference)
AIPOM: Agent-aware Interactive Planning for Multi-Agent Systems
Comments EMNLP 2025 Demo
Evaluating Spatiotemporal Consistency in Automatically Generated Sewing Instructions
Comments 18 pages, 14 figures; to be published in EMNLP 2025 proceedings
SeaPO: Strategic Error Amplification for Robust Preference Optimization of Large Language Models
机构 * Institute of Computing and Intelligence, Harbin Institute of Technology, Shenzhen（计算与智能学院，哈尔滨工业大学，深圳） ； Huawei Cloud Computing Technologies Co., Ltd.（华为云计算技术有限公司） ； School of Intelligence Science and Engineering, Harbin Institute of Technology, Shenzhen（智能科学与工程学院，哈尔滨工业大学，深圳）
Comments EMNLP 2025 Findings
Bias Mitigation or Cultural Commonsense? Evaluating LLMs with a Japanese Dataset
机构 * The University of Tokyo（东京大学） ； Riken（理化学研究所） ； University of Liverpool（利物浦大学）
Comments Accepted to EMNLP 2025 main
Beyond Repetition: Text Simplification and Curriculum Learning for Data-Constrained Pretraining
机构 * Samsung R&D Institute Philippines（三星菲律宾研发中心）
Comments To be published in BabyLM Workshop at EMNLP 2025
AlignX: Advancing Multilingual Large Language Models with Multilingual Representation Alignment
机构 * Key Laboratory of Intelligent Information Processing, Institute of Computing Technology, Chinese Academy of Sciences (ICT/CAS)（智能信息处理重点实验室，计算技术研究所，中国科学院） ； Key Laboratory of AI Safety, Chinese Academy of Sciences（人工智能安全重点实验室，中国科学院） ； University of Chinese Academy of Sciences, Beijing, China（中国科学院大学，北京，中国） ； Baidu Inc.（百度公司）
Comments Accepted to EMNLP 2025 Main Conference. The code will be available at https://github.com/ictnlp/AlignX
MoVa: Towards Generalizable Classification of Human Morals and Values
机构 * The Australian National University（澳大利亚国立大学） ； University of Chicago（芝加哥大学） ； University of Pennsylvania（宾夕法尼亚大学） ； Microsoft Research Asia（微软亚洲研究院）
Comments 9 pages, 10 figures and tables, EMNLP 2025 main conference
Retrieval-augmented GUI Agents with Generative Guidelines
机构 * Emory University（埃默里大学） ； Tencent AI Lab（腾讯AI实验室）
Comments Accepted to EMNLP 2025 (Main Conference)
EduVidQA: Generating and Evaluating Long-form Answers to Student Questions based on Lecture Videos
机构 * Indian Institute of Technology, Kharagpur（印度理工学院，克哈格普尔） ； Panjab University, Chandigarh（旁遮普大学，昌迪加尔）
Comments EMNLP 2025 (Main)
ResFormer: All-Time Reservoir Memory for Long Sequence Classification
机构 * Gateway Academic Center, Stevens Institute of Technology（Gateway学术中心，史蒂文斯理工学院）
Comments Accepted at EMNLP 2025. To appear in the proceedings
ByteSized32Refactored: Towards an Extensible Interactive Text Games Corpus for LLM World Modeling and Evaluation
机构 * Johns Hopkins University（约翰霍普金斯大学） ； Liaoning Technical University（辽宁技术大学） ； Microsoft Research Montréal（微软研究院蒙特利尔分校） ； Central University of Finance and Economics（中央财经大学）
Comments 14 pages,15 figures, Accepted to the 5th Wordplay: When Language Meets Games Workshop, EMNLP 2025
Everyday Physics in Korean Contexts: A Culturally Grounded Physical Reasoning Benchmark
Comments Accepted to MRL@EMNLP 2025
