Do All Autoregressive Transformers Remember Facts the Same Way? A Cross-Architecture Analysis of Recall Mechanisms
机构 * Chosun University（全州大学） ； Soongsil University（顺天大学） ； Kongju National University（康州国立大学）
Comments Accepted at EMNLP 2025
期刊&会议
EMNLP
Conference on Empirical Methods in Natural Language Processing · 会议 · Natural Language Processing
Simulating Identity, Propagating Bias: Abstraction and Stereotypes in LLM-Generated Text
机构 * Computational Linguistics and Text Mining Lab, Vrije Universiteit, Amsterdam（自由大学计算语言学与文本挖掘实验室） ； Università di Bologna（博洛尼亚大学） ； CLCG, University of Groningen（格罗宁根大学CLCG）
Comments Accepted to EMNLP Findings 2025
Acquiescence Bias in Large Language Models
机构 * Marburg University（马尔堡大学）
Comments Accepted to EMNLP 2025 Findings
Adversarial Attacks Against Automated Fact-Checking: A Survey
机构 * School of Computing, Macquarie University, Australia（麦考瑞大学计算机学院） ； CSIRO’s Data61, Australia（澳大利亚CSIRO数据61） ； UNSW Sydney, Australia（新南威尔士大学悉尼分校）
Comments Accepted to the Main Conference of EMNLP 2025. Resources are available at https://github.com/FanzhenLiu/Awesome-Automated-Fact-Checking-Attacks
CURE: Controlled Unlearning for Robust Embeddings -- Mitigating Conceptual Shortcuts in Pre-Trained Language Models
机构 * Technical University of Munich（慕尼黑技术大学）
Comments Accepted at the Conference on Empirical Methods in Natural Language Processing (EMNLP 2025)
Measuring Bias or Measuring the Task: Understanding the Brittle Nature of LLM Gender Biases
机构 * The University of Chicago（芝加哥大学） ； University of California, Los Angeles（加州大学洛杉矶分校）
Comments To be published at EMNLP 2025 (main conference)
A Survey on Training-free Alignment of Large Language Models
机构 * School of Computer Science, Wuhan University, Wuhan, China（武汉大学计算机学院） ； Faculty of Computer Science and Technology, Qilu University of Technology, Shandong, China（青岛科技大学计算机科学与技术学院） ； Zhongguancun Academy, Beijing, China（中关村学院）
Comments Accepted to EMNLP 2025 (findings), camera-ready version
Your Language Model Can Secretly Write Like Humans: Contrastive Paraphrase Attacks on LLM-Generated Text Detectors
机构 * Tsinghua Shenzhen International Graduate School, Tsinghua University（清华大学深圳国际研究生院，清华大学） ； Harbin Institute of Technology（哈尔滨工业大学） ； Pengcheng Laboratory（鹏城实验室）
Comments Accepted by EMNLP-2025
Prior Prompt Engineering for Reinforcement Fine-Tuning
Comments Accepted at EMNLP 2025, Main; 26 pages, 42 figures
To See a World in a Spark of Neuron: Disentangling Multi-task Interference for Training-free Model Merging
机构 * Xiamen University Malaysia（马来西亚厦门大学） ； The Hong Kong Polytechnic University（香港理工大学） ； Columbia University（哥伦比亚大学） ； Duke University（杜克大学） ； Harbin Institute of Technology (Shenzhen)（哈尔滨工业大学（深圳））
Comments Accepted to EMNLP 2025 Main Conference. This is the camera-ready version. Code: https://ZzzitaoFang.github.io/projects/NeuroMerging/
MPO: Boosting LLM Agents with Meta Plan Optimization
Comments EMNLP 2025 Findings
REGen: A Reliable Evaluation Framework for Generative Event Argument Extraction
机构 * Department of Computer Science, Dartmouth College（计算机科学系，达特茅斯学院）
Comments Accepted at EMNLP-2025
Beyond Seen Data: Improving KBQA Generalization Through Schema-Guided Logical Form Generation
机构 * The University of Melbourne（墨尔本大学）
Comments Accepted by EMNLP 2025
