[论文] Grounding Agent Memory: Environment-Probing Curation for Enterprise Ag...

论文概要

研究领域: cs.AI, cs.SE
作者: Susheel Suresh, Hazel Mak, Sahil Bhatnagar, Chhaya Methani, Alejandro Gutierrez Munoz
发布时间: 2026-09-13
arXiv: 2609.11060

中文摘要

持久记忆正在进入生产导向的智能体平台，帮助长期智能体跨会话积累经验。然而，仅限于已完成轨迹的后任务策展智能体可能保留错误、过度概括部分证据或保留陈旧知识。我们引入环境探测策展，一种部署兼容的扩展，为现有异步策展智能体提供最低权限的只读世界工具来检查、确定范围和刷新候选记忆。它不需要模型重新训练，也不改变任务智能体、检索器、记忆表示和生产写入权限。在生产类 GitHub Copilot（GHCP）工具中，我们比较无状态执行、完整上下文学习、GHCP + Mem 和 GHCP + Mem（含环境探测）在 CLBench 数据库探索和 90 个适应性 APEX 管理咨询任务上的表现。在 CLBench 上，探测将通过率从 39% 提高到 73%，折扣奖励从 8.60 提高到 22.60，同时将每问题查询从 8.8 减少到 4.7，任务智能体成本从 $3.38 降低到 $1.68。

原文摘要

Persistent memory is entering production-oriented agent platforms to help long-horizon agents accumulate experience across sessions. Yet a post-task curator agent restricted to completed trajectories can preserve errors, overgeneralize partial evidence, or retain stale knowledge. We introduce environment-probing curation, a deployment-compatible extension that gives an existing asynchronous curator agent least-privilege, read-only world tools to check, scope, and refresh candidate memories. It requires no model retraining and leaves the task agent, retriever, memory representation, and production write authority unchanged. In a production-like GitHub Copilot (GHCP) harness built on its SDK, we compare stateless execution, full in-context learning, GHCP + Mem, and GHCP + Mem (w/ Env Probing) on CLBench database exploration and 90 adapted APEX management-consulting tasks. On CLBench, probing raises pass rate from 39% to 73% and pass-discounted reward from 8.60 to 22.60 while reducing queries from 8.8 to 4.7 per question and task-agent cost from $3.38 to $1.68. Across six APEX worlds, all 18 memory-versus-baseline mean reward comparisons are positive and task-agent tool calls fall by 16--75%; probing gives the best task-agent reward gain per dollar in five worlds. Probing also attains higher mean reward than GHCP + Mem on both Sonnet 4.6 and Opus 4.7 without schema drift. Environment probing therefore turns existing agent-memory curation into an environment-informed, auditable process while preserving a compact task-time interface.

自动采集于 2026-09-13

#论文 #arXiv #AI #小凯