# 持续记忆智能体：遗忘与恢复机制严重缺失

- 来源：Rohan Paul (@rohanpaul_ai)
- 发布时间：2026-07-27 15:03
- AIHOT 分数：28
- AIHOT 链接：https://aihot.virxact.com/items/cms2wimni05bfro3f5exldpmo
- 原文链接：https://x.com/rohanpaul_ai/status/2081636454485102753

## AI 摘要

一项对435篇论文的综述将始终在线的AI智能体重构为持久状态系统，其状态包括事实、偏好、权限、凭证、任务账本、触发器、来源和外部承诺。关键问题在于这些状态是否仍保持授权、可追溯、可变、可恢复且安全。在435篇论文中，269篇涉及检索、200篇涉及写入，但仅27篇提供回滚机制，72篇涉及权限管理——领域已构建了写入、组织、检索和行动的前向弧，却严重忽视了遗忘、审计和恢复的返回弧。

## 正文

The hardest memory failure may be an agent that remembers enough to act but not enough to explain or undo the action.

This survey reframes always-on agents as persistent-state systems.

Their durable state includes facts and preferences， but also permissions， credentials， task ledgers， triggers， provenance， and external commitments.

The key question is whether that state remains authorized， scoped， traceable， mutable， recoverable， and safe to act on.

Across a scoped corpus of 435 works， retrieval appears in 269 and writing in 200， while only 27 expose any rollback mechanism； authority appears in 72.

The field has built the forward arc of writing， organizing， retrieving， and acting far more thoroughly than the return arc of forgetting， auditing， and recovery.

- arxiv. org/abs/2606.30306

Title： "Always-OnAgents：A Survey of Persistent Memory， State， and Governance in LLMAgents"