返回首页
arXiv AI··论文与技术

Agent-Controlled Forgetting for Tool-Using Agents: Reversible Context Curation in Practice

中文摘要

工具型智能体可通过“受控遗忘”将冗长工具输出替换为简短笔记,并将原始数据存入可恢复存档,实现高效可逆的上下文管理。

English Summary

Tool-using agents optimize context by replacing large tool outputs with concise notes while storing original data in a recoverable archive, enabling efficient and reversible context curation.

原文节选

arXiv:2610.10590v1 Announce Type: new Abstract: Tool-using agents repeatedly carry observations whose useful content can be much smaller than their original payload. We study agent-controlled forgetting: the acting model selects previously observed tool results, replaces each with a short note at its original position, and retains the exact original in a recoverable archive. A Python harness exposes batch archival and explicit recovery without task-specific model training, while protecting user instructions and assistant messages from these operations. In an exploratory OpenTelemetry debugging case followed by an unrelated implementation task, the method ended with 231,951 provider-reported prompt tokens versus 912,492 under retained history, used 50% fewer cumulative input tokens, and had an estimated API cost of USD 1.28-1.44 versus approximately USD 4.38. Both arms passed the two-case primary behavioral oracle; neither fully satisfied the follow-up evaluation. The method made more requests and took 17% longer. A contrasting application-development pair produced no context or cost saving, and an earlier continuation exhibited lower manually assessed quality despite reduced contex…