LLM代理中的‘记忆诅咒’
TL;DR · AI 摘要
研究发现长历史记录会降低LLM代理的协作能力，称为‘记忆诅咒’。
核心要点
- 7个LLM在4种社会困境游戏中，扩展历史记录导致18/28组合合作下降。
- 长历史使模型关注过去而非未来收益，Forward-looking intent减弱。
- LoRA适配器和记忆净化可有效缓解‘记忆诅咒’。
结构提纲
按章节快速跳转。
思维导图
用一张图看清主题之间的关系。
查看大纲文本（无障碍 / 无 JS 友好）
- Memory Curse in LLM Agents
- 现象
- 长历史降低协作能力
- 实验
- 7个LLM，4种社会困境游戏，500轮测试
- 机制
- Forward-looking intent减弱，关注过去而非未来
- 解决方案
- LoRA适配器
- 仅基于forward-looking traces训练
- 记忆净化
- 保持prompt长度固定，替换合成合作记录
金句 / Highlights
值得收藏与分享的关键句。
长历史显然会降低代理的能力，使其更加遵循历史并规避风险。
扩展可访问的历史记录导致28种模型-游戏组合中有18种合作下降。
仅基于前瞻性轨迹训练的LoRA适配器可以缓解衰减，并零样本迁移至新游戏。
(bookmark it)
Long histories apparently degrades agents as they become increasingly history-following and risk-minimizing.
Across 7 LLMs and 4 social dilemma games over 500 rounds, expanding accessible history degraded cooperation in 18 of https://t.co/f2gVtfkZ2r" / X
// The Memory Curse in LLM Agents // (bookmark it) Long histories apparently degrades agents as they become increasingly history-following and risk-minimizing. Across 7 LLMs and 4 social dilemma games over 500 rounds, expanding accessible history degraded cooperation in 18 of 28 model–game combinations. They call it the memory curse. Lexical analysis of 378,000 reasoning traces shows the mechanism: it's not that agents become paranoid, it's that forward-looking intent erodes. Long histories pull the model into reasoning about past slights instead of future payoffs. A LoRA adapter trained only on forward-looking traces mitigates the decay and transfers zero-shot to new games. Memory sanitization, keeping prompt length fixed but swapping in synthetic cooperative records, restores cooperation, proving the trigger is content, not length. And ablating explicit Chain-of-Thought often reduces the collapse, meaning deliberation actively amplifies the curse. Paper: arxiv.org/abs/2605.08060 Learn to build effective AI agents in our academy: academy.dair.ai
