TaDA: Training-free recipe for Decoding with Adaptive KV Cache Compression and Mean-centering
机构 * AMD
Comments ACL-2025 industry-track accepted
期刊&会议
ACL
Annual Meeting of the Association for Computational Linguistics · 会议 · Natural Language Processing
A MISMATCHED Benchmark for Scientific Natural Language Inference
机构 * Computer Science（计算机科学） ； University of Illinois Chicago（伊利诺伊大学芝加哥分校） ； Kansas State University（堪萨斯州立大学）
Comments Accepted to Findings of ACL 2025
Hierarchical-Task-Aware Multi-modal Mixture of Incremental LoRA Experts for Embodied Continual Learning
机构 * Ping An Technology (Shenzhen) Co., Ltd.（平安科技（深圳）有限公司） ； Tsinghua Shenzhen International Graduate School（清华大学深圳国际研究生院） ； Tsinghua University（清华大学） ； Huazhong University of Science and Technology（华中科技大学）
Comments Accepted by the 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)
Safe: Enhancing Mathematical Reasoning in Large Language Models via Retrospective Step-aware Formal Verification
机构 * School of Computer Science, National Key Laboratory for Multimedia Information Processing, PKU-Anker LLM Lab, Peking University（计算机学院、多媒体信息处理国家重点实验室、PKU-Anker LLM实验室、北京大学） ； Huawei Noah’s Ark Lab（华为诺亚实验室） ； The Hong Kong University of Science and Technology（香港科技大学） ； The Hong Kong Polytechnic University（香港理工大学）
Comments Accepted in ACL 2025
Selecting Demonstrations for Many-Shot In-Context Learning via Gradient Matching
机构 * School of Computer Science and Engineering, Beihang University, China（北京航空航天大学计算机科学与工程学院） ； Meituan, Inc.（美团公司） ； Beijing Institute for General Artificial Intelligence, China（北京通用人工智能研究院） ； Ping An Technology, China（平安科技） ； Department of Computer Science, University of Manchester, UK（曼彻斯特大学计算机科学系）
Comments accepted to the ACL2025 Findings
Is It JUST Semantics? A Case Study of Discourse Particle Understanding in LLMs
机构 * Linguistics, The University of Texas at Austin（德克萨斯大学奥斯汀分校语言学系） ； Toyota Technological Institute at Chicago（芝加哥丰田技术研究所） ； Allen Institute for AI（人工智能研究院） ； University of Washington（华盛顿大学）
Comments To be published in Findings of The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025). The main paper is 5 pages and contains 3 figures and 1 table. In total, the paper is 12 pages and contains 8 figures and 5 tables (References + Appendix)
Understanding and Meeting Practitioner Needs When Measuring Representational Harms Caused by LLM-Based Systems
机构 * Cornell University（康奈尔大学） ； Microsoft Research（微软研究院）
Comments Findings of the Association for Computational Linguistics: ACL 2025
Aligning Large Language Models with Implicit Preferences from User-Generated Content
机构 * University of Notre Dame（诺特大学） ； Amazon.com Inc（亚马逊公司）
Comments Accepted to ACL 2025 Main Conference
Behavioural vs. Representational Systematicity in End-to-End Models: An Opinionated Survey
机构 * School of Informatics, University of Edinburgh（信息学院，爱丁堡大学） ； Chemnitz University of Technology（技术大学，克明茨） ； Institute of Logic, Language, and Computation, University of Amsterdam（逻辑、语言与计算研究所，阿姆斯特丹大学） ； School of Philosophy, Psychology & Language Sciences, University of Edinburgh（哲学、心理学与语言科学学院，爱丁堡大学）
Comments To appear at ACL 2025 Main Conference
Empaths at SemEval-2025 Task 11: Retrieval-Augmented Approach to Perceived Emotions Prediction
机构 * HSE University（俄罗斯人民友谊大学）
Comments Accepted to SemEval-2025, an ACL 2025 workshop
An Efficient Task-Oriented Dialogue Policy: Evolutionary Reinforcement Learning Injected by Elite Individuals
机构 * School of Computer Science and Technology, Changsha University of Science and Technology, China（长沙理工大学计算机科学与技术学院） ； School of Computer Science and Engineering, Central South University, China（中南大学计算机科学与工程学院） ； School of Information and Computing Sciences, Utrecht University, Netherlands（乌特勒支大学信息与计算科学学院）
Comments Accepted to ACL 2025 (Main Track)
CoRe-MMRAG: Cross-Source Knowledge Reconciliation for Multimodal RAG
Comments Accepted to ACL 2025 Main
Spectral Insights into Data-Oblivious Critical Layers in Large Language Models
机构 * Dartmouth College（达特茅斯学院）
Comments Accepted by Findings of ACL2025
NorEval: A Norwegian Language Understanding and Generation Evaluation Benchmark
Comments Accepted for Findings of the Association for Computational Linguistics: ACL 2025
MMS-LLaMA: Efficient LLM-based Audio-Visual Speech Recognition with Minimal Multimodal Speech Tokens
机构 * Integrated Vision and Language Lab, KAIST, South Korea（集成视觉与语言实验室，韩国科学技术院）
Comments Accepted at Findings of ACL 2025. The code and models are available https://github.com/JeongHun0716/MMS-LLaMA
Implicit Cross-Lingual Rewarding for Efficient Multilingual Preference Alignment
Comments Camera ready version for ACL 2025 Findings
When Claims Evolve: Evaluating and Enhancing the Robustness of Embedding Models Against Misinformation Edits
Comments Accepted to ACL 2025 Findings
ATLaS: Agent Tuning via Learning Critical Steps
Comments ACL2025, Camera-ready
Towards Robust ESG Analysis Against Greenwashing Risks: Aspect-Action Analysis with Cross-Category Generalization
机构 * National University of Singapore（新加坡国立大学） ； Nanyang Technological University（南洋理工大学）
Comments Proceedings of the Association for Computational Linguistics Main Conference (ACL 2025)
Should I Trust You? Detecting Deception in Negotiations using Counterfactual RL
机构 * Department of Computer Science, University of Maryland（大学计算机科学系） ； Northwestern University（西北大学） ； School of Computer Science, University of Sydney（悉尼大学计算机科学学院） ； Information Sciences Institute, University of Southern California（南加州大学信息科学研究所）
Comments ACL Findings 2025
Lost in Transcription, Found in Distribution Shift: Demystifying Hallucination in Speech Foundation Models
机构 * Carnegie Mellon University（卡内基梅隆大学） ； MBZUAI（穆扎布伊人工智能研究所）
Comments ACL2025 camera-ready
On the Robust Approximation of ASR Metrics
机构 * Carnegie Mellon University（卡内基梅隆大学） ； MBZUAI（穆斯林人工智能研究所）
Comments ACL 2025 camera-ready
GRAF: Graph Retrieval Augmented by Facts for Romanian Legal Multi-Choice Question Answering
机构 * National University of Science and Technology POLITEHNICA Bucharest, Faculty of Automatic Control and Computers（波兰技术大学布加勒斯特分校国家科学与技术大学，自动控制与计算机学院） ； Technical University of Munich（慕尼黑技术大学） ； Paris 1 Panthéon-Sorbonne University（巴黎第一大学） ； University of Bucharest（布加勒斯特大学）
Comments Accepted to ACL 2025 Findings
FastDraft: How to Train Your Draft
机构 * Intel Labs（英特尔实验室）
Comments Accepted at ACL 2025
What Happened in LLMs Layers when Trained for Fast vs. Slow Thinking: A Gradient Perspective
机构 * University of Maryland（马里兰大学） ； University of Chicago（芝加哥大学）
Comments ACL2025 main, Camera-ready
Self-Correction is More than Refinement: A Learning Framework for Visual and Language Reasoning Tasks
机构 * University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校） ； Georgia Institute of Technology（佐治亚理工学院） ； The Hong Kong University of Science and Technology（香港科技大学）
Comments Accepted by ACL 2025 Findings
Inducing lexicons of in-group language with socio-temporal context
机构 * University of Melbourne（墨尔本大学）
Comments Accepted to ACL 2025
Leveraging LLMs for Bangla Grammar Error Correction:Error Categorization, Synthetic Data, and Model Evaluation
Comments Accepted at ACL Findings, 2025
The Impossibility of Fair LLMs
机构 * University of Chicago（芝加哥大学） ； Stanford University（斯坦福大学） ； Drexel University（德雷塞尔大学） ； University of California, Berkeley（加州大学伯克利分校）
Comments Published in ACL 2025
Mosaic-IT: Cost-Free Compositional Data Synthesis for Instruction Tuning
机构 * University of Maryland（马里兰大学） ； Texas A&M University（德克萨斯A&M大学） ； Stony Brook University（石溪大学）
Comments ACL2025, Camera-ready
