KEEPER：面向长程 LLM Agent 的控制充分文本记忆研究
KEEPER: Learning Control-Sufficient Textual Memory for Long-Horizon LLM Agents
AI 导读
arXiv 论文提出 KEEPER，在共享递归文本记忆接口下对比四种辅助目标在信息量与更新敏感性两个维度上的表现。结果显示，与奖励相关的预测目标在离线强化学习前后均取得最高任务性能，优于联合监督。针对基线轨迹的 claim 级分析记录了错误声明在收到纠正观察后仍持续存在，据此作者主张将错误持续性与有效信息保留纳入记忆可靠性评估。
正文 · 原文
Abstract:Recursive textual memory can carry earlier errors into later updates, yet task performance alone does not establish memory reliability. We define self-consistency bias as the persistence of incorrect claims already present in the incoming memory. We present KEEPER, a comparative analysis of four auxiliary objectives along information-content and update-sensitivity axes within a shared recursive interface, examining how supervision targets relate to agent performance. Among evaluated conditions, reward-related prediction achieve the highest task performance before and after offline reinforcement learning, outperforming joint supervision. A claim-level analysis of one baseline trajectory documents error persistence despite corrective observations. Together, these analyses motivate evaluating error persistence and valid-information retention alongside task outcomes.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.41482 [cs.CL] |
| (or arXiv:2610.41482v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.41482 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jinkyeong Choi [view email]
[v1]
Fri, 30 Oct 2026 09:36:06 UTC (410 KB)
来源：arXiv cs.AI · arxiv.org