Long-context tasks require LLMs to identify and preserve answer-relevant information from large contexts. Chunk-wise memory agents address this issue by sequentially reading document chunks, updating a compact memory, and generating the final answer from the accumulated memory. However, existing RL-based chunk-wise agents either rely on sparse final-answer rewards or use lexical intermediate rewards for memory and retrieval actions. These signals supervise task success or local overlap, but do not directly evaluate whether the final memory supports the ground-truth answer. We propose InfoMem, a reward mechanism for training chunk-wise memory agents that evaluates final-memory utility using answer-conditioned information. InfoMem measures how much the final memory increases the model's per-token log-likelihood of the ground-truth answer. To stabilize RL optimization, InfoMem applies this signal only to successful trajectories and normalizes it before reward composition. Under the same GRPO framework and training budget, InfoMem improves long-context memory-agent performance over comparable memory-agent RL baselines. Analyses show that effective final-memory rewards should operate on successful trajectories, be normalized before reward composition, and be conditioned on the answer rather than the query. Our code is available at https://github.com/GenSouKa1/InfoMem.
From Parameters to Answers: How LLMs Retrieve and Use Their Internal Knowledge
University of Science and Technology of China(中国科学技术大学);Singapore Management University(新加坡管理大学);The University of Tokyo(东京大学);The Chinese University of Hong Kong(香港中文大学)
LEC: Linear Expectation Constraints for Selection-Conditioned Risk Control in Selective Prediction and Routing Systems
University of Electronic Science and Technology of China(电子科技大学);Munich Center for Machine Learning(慕尼黑机器学习中心);University of North Carolina at Chapel Hill(北卡罗来纳大学教堂山分校);Shandong University(山东大学);Tongji University(同济大学);City University of Hong Kong(香港城市大学)
Skill-Conditioned Gated Self-Distillation for LLM Reasoning
Tsinghua University(清华大学);Fudan University(复旦大学);City University of Hong Kong(香港城市大学);Huazhong University of Science and Technology(华中科技大学);University of Wisconsin-Madison(威斯康星大学麦迪逊分校)
ModaLens: Measuring Image Sensitivity in Report-Conditioned Medical VLMs
Massachusetts Institute of Technology(麻省理工学院);American International School Vienna(维也纳美国国际学校);Substrate Labs;University of North Florida(北佛罗里达大学);University of North Carolina at Chapel Hill(北卡罗来纳大学教堂山分校);Motork;Collingwood School(科林伍德学校);Beth Israel Deaconess Medical Center(贝斯以色列女执事医疗中心)
Latent Variable Models for Visual Question Answering
Imperial College London(帝国理工学院);University of Sheffield(谢菲尔德大学);Department of Computing, Imperial College London(帝国理工学院计算机系);Department of Computer Science, University of Sheffield(谢菲尔德大学计算机科学系)
HYPER: Hypothesis-Guided Planning for Evidence-Seeking Retrieval
Graduate School of AI, KAIST, Republic of Korea(韩国科学技术院人工智能研究生院);School of Electrical Engineering, KAIST, Republic of Korea(韩国科学技术院电气工程学院);Department of Industrial Engineering, Yonsei University, Republic of Korea(延世大学工业工程系)
MemChain: Learning Interpretable Memory Traces for Memory-Augmented LLM Agents
Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所);School of Advanced Interdisciplinary Sciences, University of Chinese Academy of Sciences(中国科学院大学交叉科学学院);Memorax AI(Memorax人工智能公司);School of Artificial Intelligence, University of Chinese Academy of Sciences(中国科学院大学人工智能学院)
EAGER: Asking and Answering Questions for Automatic Reward Shaping in Language-guided RL
Inria(法国国家信息与自动化研究所);Université de Bordeaux(波尔多大学);ISIR Sorbonne Université(索邦大学智能系统与机器人研究所);Univ Angers(昂热大学);Microsoft Research Montreal(微软研究院蒙特利尔分院)
Cross-modal Knowledge Reasoning for Knowledge-based Visual Question Answering
Institute of Information Engineering, Chinese Academy of Sciences(中国科学院信息工程研究所);School of Cyber Security, University of Chinese Academy of Sciences(中国科学院大学网络空间安全学院);Microsoft Research Asia(微软亚洲研究院);College of Mathematics, Physics and Information Engineering, Jiaxing University(嘉兴大学数理与信息工程学院)
Query-Conditioned Knowledge Alignment for Reliable Cross-System Medical Reasoning
Shenzhen Institute for Advanced Study, University of Electronic Science and Technology of China(深圳先进研究院,电子科技大学);Department of Electrical and Computer Engineering, University of Waterloo(滑铁卢大学电气与计算机工程系);School of Computer Science and Engineering, Kyungpook National University(庆北国立大学计算机科学与工程学院)