Files
myWiki/concepts/reliable-state-long-running-agents.md
2026-06-01 10:46:01 +08:00

45 lines
1.4 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

---
title: "Reliable State in Long-Running Agents长期运行中的可靠状态"
created: 2026-05-23
updated: 2026-05-23
type: concept
tags: [agent, state, context, memory, long-running]
sources: [raw/papers/agent-harness-engineering-survey-2026.md]
confidence: medium
---
# Maintaining Reliable State in Long-Running Agents
> 开放问题 2/5最深的上下文问题不是装更多 token而是保持 Agent 工作状态与真实任务状态在长时间跨度上的**对齐**。
## 状态退化机制
长期运行的编程、研究和运维 Agent 反复执行:
- 摘要 → 可能删除约束
- 检索 → 可能扭曲优先级
- 压缩 → 可能保留过时假设
- 外化信息 → 可能丢失上下文
## 重新框架:状态估计
Zhang et al. (2025) 和 Du (2026) 将 Agent 记忆形式化为**写入-管理-读取循环**。
开放问题:
- 能否量化每次压缩/检索/遗忘步骤中丢失多少任务相关信息?
- 能否界定 Agent 内部状态与真实任务状态之间的**发散程度**
## 未来系统需求
- 不确定性感知摘要
- 记忆事实的溯源provenance
- 矛盾处理
- 显式陈旧标记
- 从持久化产物重建状态(而非信任自身压缩历史)
- 记忆策略不仅按召回准确率评判,还要看是否防止了下游操作错误
## 相关概念
- [[context-management]]
- [[context-state-estimation]]
- [[agent-harness-engineering-survey]]