记忆诅咒：扩展回忆如何在LLM代理中侵蚀合作意图
The Memory Curse: How Expanded Recall Erodes Cooperative Intent in LLM Agents
AI总结：
研究发现扩展回忆会系统性地削弱多代理社会困境中的合作意图，通过三种分析揭示记忆内容对合作的影响，证明记忆是影响多代理行为的主动因素。
摘要：
扩展上下文窗口常被视为LLM能力的简单升级，但我们发现其在多代理社会困境中系统性失效。在7个LLM和4个游戏超过500轮的测试中，扩展可访问历史会破坏18种模型-游戏设置中的合作，我们将其称为记忆诅咒。通过三种分析揭示其机制：首先，对378,000条推理轨迹的词频分析表明，这种失效与侵蚀前瞻性意图有关，而非上升的偏执。我们通过针对性微调作为认知探针验证：仅在前瞻性轨迹上训练的LoRA适配器可缓解衰减并转移至不同游戏。其次，记忆净化固定提示长度，同时用合成合作记录替换可见历史，显著恢复合作，证明触发因素是记忆内容而非长度本身。最后，消去显式推理链往往减少崩溃，显示推理反而放大了记忆诅咒。共同，这些结果将记忆重新定义为多代理行为的主动决定因素：更长的回忆可根据其引发的推理模式稳定或支持合作。
英文摘要：
Context window expansion is often treated as a straightforward capability upgrade for LLMs, but we find it systematically fails in multi-agent social dilemmas. Across 7 LLMs and 4 games over 500 rounds, expanding accessible history degrades cooperation in 18 of 28 model--game settings, a pattern we term the memory curse. We isolate the underlying mechanism through three analyses. First, lexical analysis of 378,000 reasoning traces associates this breakdown with eroding forward-looking intent rather than rising paranoia. We validate this using targeted fine-tuning as a cognitive probe: a LoRA adapter trained exclusively on forward-looking traces mitigates the decay and transfers zero-shot to distinct games. Second, memory sanitization holds prompt length fixed while replacing visible history with synthetic cooperative records, which restores cooperation substantially, proving the trigger is memory content, not length alone. Finally, ablating explicit Chain-of-Thought reasoning often reduces the collapse, showing that deliberation paradoxically amplifies the memory curse. Together, these results recast memory as an active determinant of multi-agent behavior: longer recall can either destabilize or support cooperation depending on the reasoning patterns it elicits.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
