arXiv cs.AI· Lihuan Li, Ruiyi Yang, Jiru Wang, Yifan Duan, Xiachong Lin, Hao Xue, Yang Song, Flora Salim·· 4 小时前AI 评分49
CarryOn:为长程 LLM 智能体学习上下文保留策略
CarryOn: Learning What to Keep in Context for Long-Horizon LLM Agents
AI 导读
arXiv 论文提出 CarryOn,一个轻量级上下文管理器,将紧凑的执行状态与细粒度、带来源链接的证据配对,通过学习型选择器和自适应保留策略管理长程 LLM 智能体的上下文,无需额外的 LLM 调用。
正文 · 原文
Abstract:Long-horizon LLM agents can coordinate tools and applications to complete complex, multi-step tasks. As their interaction histories grow, effective context management becomes essential for reliable and efficient execution. Existing methods reduce context through summarization, pruning, or history selection. However, shorter contexts do not necessarily support better decisions: summaries can omit exact details needed for later actions, while retaining entire interactions can carry substantial irrelevant information. We propose CarryOn, a lightweight context manager that pairs compact execution state with fine-grained, source-linked evidence. A learned selector and adaptive retention policy keep relevant evidence available without additional LLM calls for context management. We introduce two complementary evaluation metrics: $S_\alpha$ prioritizes task success while accounting for resource use, and normalized cost per success (NCS) measures the resources spent per successful task. Across AppWorld, OfficeBench, and Tau2 Telecom with three LLM Actors, CarryOn achieves 75.7% macro-averaged task success, exceeding the strongest baseline by 6.7 percentage points. It also achieves the best average scores on both metrics.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.17808 [cs.CL] |
| (or arXiv:2611.17808v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.17808 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Lihuan Li [view email]
[v1]
Fri, 13 Nov 2026 06:43:19 UTC (2,184 KB)
来源:arXiv cs.AI · arxiv.org