EMNLP

Can Federated Learning Safeguard Private Data in LLM Training? Vulnerabilities, Attacks, and Defense Evaluation

Wenkai Guo, Xuefeng Liu, Haolin Wang, Jianwei Niu, Shaojie Tang, Jing Yuan

机构 * State Key Laboratory of Virtual Reality Technology and Systems, School of Computer Science and Engineering, Beihang University, Beijing, China（虚拟现实技术与系统国家重点实验室，计算机科学与工程学院，北京航空航天大学） ； Hangzhou Innovation Institute of Beihang University, Zhejiang Key Laboratory of Industrial Big Data and Robot Intelligent Systems, Hangzhou, China（北京航空航天大学杭州创新研究院，浙江省工业大数据与机器人智能系统重点实验室） ； Center for AI Business Innovation, Department of Management Science and Systems, University at Buffalo, Buffalo, New York, USA（人工智能商业创新中心，管理科学与系统系，布法罗大学） ； University of North Texas, Denton, Texas, USA（德克萨斯大学达文波特分校） ； Zhongguancun Laboratory, Beijing, China（中关村实验室） 纠错

Comments 28 pages, 32 figures, accepted to the Findings of EMNLP 2025

SAMULE: Self-Learning Agents Enhanced by Multi-level Reflection

Yubin Ge, Salvatore Romeo, Jason Cai, Monica Sunkara, Yi Zhang

机构 * AWS AI Labs（AWS人工智能实验室） 纠错

Comments Accepted at EMNLP 2025 Main Conference

ZERA: Zero-init Instruction Evolving Refinement Agent -- From Zero Instructions to Structured Prompts via Principle-based Optimization

Seungyoun Yi, Minsoo Khang, Sungrae Park

机构 * Upstage AI Research（Upstage人工智能研究院） 纠错

Comments 9 pages, 4 figures. To appear in EMNLP 2025 Main Conference (Oral Presentation)

TactfulToM: Do LLMs Have the Theory of Mind Ability to Understand White Lies?

Yiwei Liu, Emma Jane Pretty, Jiahao Huang, Saku Sugawara

机构 * EPFL（苏黎世联邦理工学院） ； Tampere University（塔尔库大学） ； University of Tokyo（东京大学） ； National Institute of Informatics（信息研究所） 纠错

Comments Accepted to EMNLP 2025

LM-Searcher: Cross-domain Neural Architecture Search with LLMs via Unified Numerical Encoding

Yuxuan Hu, Jihao Liu, Ke Wang, Jinliang Zhen, Weikang Shi, Manyuan Zhang, Qi Dou, Rui Liu, Aojun Zhou, Hongsheng Li

机构 * CUHK MMLab（香港中文大学MML实验室） ； CUHK CURI（香港中文大学CURI） ； Tsinghua University（清华大学） ； Shanghai AI Laboratory（上海人工智能实验室） ； CPII under InnoHK（创新香港下的CPII） 纠错

Comments EMNLP 2025 Main

MathBuddy: A Multimodal System for Affective Math Tutoring

Debanjana Kar, Leopold Böss, Dacia Braca, Sebastian Maximilian Dennerlein, Nina Christine Hubig, Philipp Wintersberger, Yufang Hou

机构 * IT:U Interdisciplinary Transformation University Austria（奥地利 interdisciplinary Transformation 大学） ； IBM Research India（印度 IBM 研究院） 纠错

Comments Accepted at EMNLP 2025 (Demo Track)

THCM-CAL: Temporal-Hierarchical Causal Modelling with Conformal Calibration for Clinical Risk Prediction

Xin Zhang, Qiyu Wei, Yingjie Zhu, Fanyi Wu, Sophia Ananiadou

机构 * The University of Manchester（曼彻斯特大学） ； Harbin Institute of Technology（哈尔滨工业大学） 纠错

Comments Accepted at EMNLP 2025

ConsistentChat: Building Skeleton-Guided Consistent Multi-Turn Dialogues for Large Language Models from Scratch

Jiawei Chen, Xinyan Guan, Qianhao Yuan, Guozhao Mo, Weixiang Zhou, Yaojie Lu, Hongyu Lin, Ben He, Le Sun, Xianpei Han

机构 * Chinese Information Processing Laboratory, Institute of Software, Chinese Academy of Sciences, Beijing, China（中国科学院软件研究所中文信息处理实验室） ； University of Chinese Academy of Sciences, Beijing, China（中国科学院大学） 纠错

Comments EMNLP 2025 Main Conference (Oral Presentation)

BabyLM's First Constructions: Causal probing provides a signal of learning

Joshua Rozner, Leonie Weissweiler, Cory Shain

Comments Accepted to EMNLP 2025

VideoPASTA: 7K Preference Pairs That Matter for Video-LLM Alignment

Yogesh Kulkarni, Pooyan Fazli

机构 * Arizona State University（亚利桑那州立大学） 纠错

Comments EMNLP 2025 (Main)

On the Perception Bottleneck of VLMs for Chart Understanding

Junteng Liu, Weihao Zeng, Xiwen Zhang, Yijun Wang, Zifei Shan, Junxian He

机构 * The Hong Kong University of Science and Technology（香港理工大学） 纠错

Comments EMNLP 2025: Camera-ready version

Constructions are Revealed in Word Distributions

Joshua Rozner, Leonie Weissweiler, Kyle Mahowald, Cory Shain

机构 * Stanford University（斯坦福大学） ； Uppsala University（乌普萨拉大学） ； The University of Texas at Austin（德克萨斯大学奥斯汀分校） 纠错

Comments Accepted to EMNLP 2025

Scaling Rich Style-Prompted Text-to-Speech Datasets

Anuj Diwan, Zhisheng Zheng, David Harwath, Eunsol Choi

机构 * Department of Computer Science, The University of Texas at Austin（德克萨斯大学奥斯汀分校计算机科学系） ； Department of Computer Science and Data Science, New York University（纽约大学计算机科学与数据科学系） 纠错

Comments EMNLP 2025

The Validation Gap: A Mechanistic Analysis of How Language Models Compute Arithmetic but Fail to Validate It

Leonardo Bertolazzi, Philipp Mondorf, Barbara Plank, Raffaella Bernardi

机构 * DISI, University of Trento（特伦托大学DISI中心） ； MaiNLP, Center for Information and Language Processing, LMU Munich（慕尼黑大学信息与语言处理中心） ； Munich Center for Machine Learning (MCML), Munich, Germany（慕尼黑机器学习中心） ； Free University of Bozen-Bolzano, Italy（博兹纳自由大学） 纠错

Comments EMNLP 2025 Main, 38 pages, 33 figures

Context-Aware Reasoning On Parametric Knowledge for Inferring Causal Variables

Ivaxi Sheth, Sahar Abdelnabi, Mario Fritz

机构 * CISPA Helmholtz Center for Information Security（CISPA 欧洲信息安全中心） ； Microsoft（微软公司） 纠错

Comments EMNLP'25 Findings

Interdisciplinary Research in Conversation: A Case Study in Computational Morphology for Language Documentation

Enora Rice, Katharina von der Wense, Alexis Palmer

Comments Accepted to EMNLP 2025

Procedural Environment Generation for Tool-Use Agents

Michael Sullivan, Mareike Hartmann, Alexander Koller

机构 * Department of Language Science and Technology（语言科学与技术系） ； Saarland Informatics Campus（萨尔兰信息学校区） ； Saarland University（萨尔兰大学） 纠错

Comments 16 pages, 3 figures; accepted at EMNLP 2025

DISCO Balances the Scales: Adaptive Domain- and Difficulty-Aware Reinforcement Learning on Imbalanced Data

Yuhang Zhou, Jing Zhu, Shengyi Qian, Zhuokai Zhao, Xiyao Wang, Xiaoyu Liu, Ming Li, Paiheng Xu, Wei Ai, Furong Huang

Comments Accepted by EMNLP 2025 Findings

LLMs Reproduce Stereotypes of Sexual and Gender Minorities

Ruby Ostrow, Adam Lopez

机构 * University of Edinburgh（爱丁堡大学） 纠错

Comments 13 pages, 5 figures, 9 tables (including bibliography and appendix). Accepted to Findings of EMNLP 2025

UNComp: Can Matrix Entropy Uncover Sparsity? -- A Compressor Design from an Uncertainty-Aware Perspective

Jing Xiong, Jianghan Shen, Fanghua Ye, Chaofan Tao, Zhongwei Wan, Jianqiao Lu, Xun Wu, Chuanyang Zheng, Zhijiang Guo, Min Yang, Lingpeng Kong, Ngai Wong

Comments Accepted at EMNLP 2025 (Main Conference)

Probing Gender Bias in Multilingual LLMs: A Case Study of Stereotypes in Persian

Ghazal Kalhor, Behnam Bahrak

Comments Accepted and forthcoming at the Widening Natural Language Processing Workshop (WiNLP 2025) at EMNLP 2025

Less is More: The Effectiveness of Compact Typological Language Representations

York Hay Ng, Phuong Hanh Hoang, En-Shiun Annie Lee

机构 * University of Toronto（多伦多大学） ； Ontario Tech University（安大略技术大学） 纠错

Comments Accepted to EMNLP 2025 Main Conference

Causal Understanding by LLMs: The Role of Uncertainty

Oscar Lithgow-Serrano, Vani Kanjirangat, Alessandro Antonucci

机构 * SUPSI, IDSIA（瑞士SUPSI和IDSIA） 纠错

Comments Accepted in second UncertaiNLP workshop at EMNLP 2025

From Input Perception to Predictive Insight: Modeling Model Blind Spots Before They Become Errors

Maggie Mi, Aline Villavicencio, Nafise Sadat Moosavi

机构 * University of Sheffield（谢菲尔德大学） ； University of Exeter（埃克塞特大学） ； The Alan Turing Institute（艾伦·图灵研究所） ； UFRN, Brazil（巴西UFRN） 纠错

Comments EMNLP 2025

Tokenization and Representation Biases in Multilingual Models on Dialectal NLP Tasks

Vani Kanjirangat, Tanja Samardžić, Ljiljana Dolamic, Fabio Rinaldi

机构 * SUPSI, IDSIA（SUPSI和IDSIA） 纠错

Comments Accepted in EMNLP-2025 Main conference

Do Before You Judge: Self-Reference as a Pathway to Better LLM Evaluation

Wei-Hsiang Lin, Sheng-Lun Wei, Hen-Hsen Huang, Hsin-Hsi Chen

机构 * Department of Computer Science and Information Engineering（计算机科学与信息工程系） ； National Taiwan University（国立台湾大学） ； Institute of Information Science, Academia Sinica（学术院信息研究所） ； AI Research Center (AINTU)（人工智能研究中心） 纠错

Comments Accepted as a long findings paper at EMNLP 2025

Mahānāma: A Unique Testbed for Literary Entity Discovery and Linking

Sujoy Sarkar, Gourav Sarkar, Manoj Balaji Jagadeeshan, Jivnesh Sandhan, Amrith Krishna, Pawan Goyal

机构 * Indian Institute of Technology, Kharagpur（印度理工学院，克拉格浦尔） ； Kyoto University（京都大学） ； BharatGen（巴哈特基因） 纠错

Comments Accepted to EMNLP 2025. This is the authors' version. The official version will appear in the ACL Anthology

EnAnchored-X2X: English-Anchored Optimization for Many-to-Many Translation

Sen Yang, Yu Bao, Yu Lu, Jiajun Chen, Shujian Huang, Shanbo Cheng

Comments Accepted to EMNLP 2025

CHURRO: Making History Readable with an Open-Weight Large Vision-Language Model for High-Accuracy, Low-Cost Historical Text Recognition

Sina J. Semnani, Han Zhang, Xinyan He, Merve Tekgürler, Monica S. Lam

机构 * Stanford University（斯坦福大学） 纠错

Comments EMNLP 2025

Personality Vector: Modulating Personality of Large Language Models by Model Merging

Seungjong Sun, Seo Yeon Baek, Jang Hyun Kim

机构 * Department of Human-Artificial Intelligence Interaction, Sungkyunkwan University（人类-人工智能交互系，成均馆大学） ； Department of Immersive Media Engineering, Sungkyunkwan University（沉浸式媒体工程系，成均馆大学） 纠错

Comments EMNLP 2025