20260601
This commit is contained in:
44
concepts/reliable-state-long-running-agents.md
Normal file
44
concepts/reliable-state-long-running-agents.md
Normal file
@@ -0,0 +1,44 @@
|
||||
---
|
||||
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]]
|
||||
Reference in New Issue
Block a user