A Human-Inspired Reading Agent with Gist Memory of Very Long Contexts
Kuang-Huei Lee, Xinyun Chen, Hiroki Furuta, John F. Canny, Ian Fischer
摘要
Current Large Language Models (LLMs) are not only limited to some maximum context length, but also are not able to robustly consume long inputs. To address these limitations, we propose ReadAgent, an LLM agent system that increases effective context length up to 20× in our experiments. Inspired by how humans interactively read long documents, we implement ReadAgent as a simple prompting system that uses the advanced language capabilities of LLMs to (1) decide what content to store together in a memory episode, (2) compress those memory episodes into short episodic memories called gist memories, and (3) take actions to look up passages in the original text if ReadAgent needs to remind itself of relevant details to complete a task. We evaluate ReadAgent against baselines using retrieval methods, using the original long contexts, and using the gist memories. These evaluations are performed on three long-document reading comprehension tasks: QuALITY, NarrativeQA, and QMSum. ReadAgent outperforms the baselines on all three tasks while extending the effective context window by 3.5 -20×.
问 Lune
几分钟读透这篇 Paper,不用再花一下午。
Lune 能读完全文,也能读引用它的顶会 Paper,讲清它的结论、证据和局限,每一点都注明出自哪篇 Paper。不用硬啃,也不靠猜。
已有 37 篇顶会 Paper 引用了它。别等审稿人问起,先弄清它们发现了什么。
脉络
- Efficient Streaming Language Models with Attention Sinks 引用了 Big Bird: Transformers for Longer Sequences
- Big Bird: Transformers for Longer Sequences 引用了 Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
- Agentic Reasoning: A Streamlined Framework for Enhancing LLM Reasoning with Agentic Tools 引用了 Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
- Memory-R1: Enhancing Large Language Model Agents to Manage and Utilize Memories via Reinforcement Learning 引用了 A-Mem: Agentic Memory for LLM Agents
- MemoRAG: Boosting Long Context Processing with Global Memory-Enhanced Retrieval Augmentation 引用了 Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
- SimpleMem: Efficient Lifelong Memory for LLM Agents 引用了 Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
- SimpleMem: Efficient Lifelong Memory for LLM Agents 引用了 A-Mem: Agentic Memory for LLM Agents
- A-Mem: Agentic Memory for LLM Agents 引用了 Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
引用它的顶会 Paper37
- A-Mem: Agentic Memory for LLM AgentsWujiang Xu, Zujie Liang, Kai Mei, Hang Gao 等NeurIPS 2025 · 1,138 次引用
- Memory-R1: Enhancing Large Language Model Agents to Manage and Utilize Memories via Reinforcement LearningSikuan Yan, Xiufeng Yang, Zuchao Huang, Ercong Nie 等ACL 2026 · 140 次引用
- SimpleMem: Efficient Lifelong Memory for LLM AgentsJiaqi Liu, Yaofeng Su, Peng Xia, Siwei Han 等ICML 2026 · 112 次引用
- MemoRAG: Boosting Long Context Processing with Global Memory-Enhanced Retrieval AugmentationHongjin Qian, Zheng Liu, Peitian Zhang, Kelong Mao 等WWW 2025 · 92 次引用
- Agentic Reasoning: A Streamlined Framework for Enhancing LLM Reasoning with Agentic ToolsJunde Wu, Jiayuan Zhu, Yuyuan Liu, Min Xu 等ACL 2025 · 88 次引用
它引用的 Paper19
- Retrieval-Augmented Generation for Knowledge-Intensive NLP TasksPatrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni 等NeurIPS 2020 · 19,162 次引用
- Big Bird: Transformers for Longer SequencesManzil Zaheer, Guru Guruganesh, Kumar Avinava Dubey, Joshua Ainslie 等NeurIPS 2020 · 3,159 次引用
- Generative Agents: Interactive Simulacra of Human BehaviorJoon Sung Park, Joseph C. O'Brien, Carrie Jun Cai, Meredith Ringel Morris 等UIST 2023 · 1,882 次引用
- Efficient Streaming Language Models with Attention SinksGuangxuan Xiao, Yuandong Tian, Beidi Chen, Song Han 等ICLR 2024 · 1,714 次引用
- WebShop: Towards Scalable Real-World Web Interaction with Grounded Language AgentsShunyu Yao, Howard Chen, John Yang, Karthik NarasimhanNeurIPS 2022 · 1,477 次引用
相关 Paper
- DocAgent: An Agentic Framework for Multi-Modal Long-Context Document UnderstandingLi Sun, Liu He, Shuyue Jia, Yangfan He 等EMNLP 2025 · 1 次引用
- GraphReader: Building Graph-based Agent to Enhance Long-Context Abilities of Large Language ModelsShilong Li, Yancheng He, Hangyu Guo, Xingyuan Bu 等EMNLP Findings 2024 · 13 次引用
- Chain of Agents: Large Language Models Collaborating on Long-Context TasksYusen Zhang, Ruoxi Sun, Yanfei Chen, Tomas Pfister 等NeurIPS 2024 · 297 次引用
- EMA: An Episodic Memory Agent for Efficient and Selective MemoryHongyi Lan, Jiaqi Song, Zhengjia Zhong, Hui Li 等ACL Findings 2026
- CAM: A Constructivist View of Agentic Memory for LLM-Based Reading ComprehensionRui Li, Zeyu Zhang, Xiaohe Bo, Zihang Tian 等NeurIPS 2025 · 29 次引用
