发现大模型记忆会引发认知陷阱，导致推理错误。
MemTrapBench: Benchmarking Cognitive Traps in LLM Memory Use
- 设计新基准MemTrapBench，测试记忆如何扭曲推理和信念。
- 所有记忆策略在该基准上表现均差于无记忆，最差下降超10%。
- 提出AdaptiveMem方法，在不牺牲标准性能下规避认知陷阱。
记忆已成为大型语言模型的关键组件，使其能够保留信息并从长期交互中学习。然而，现有记忆基准主要评估信息是否被正确提取、存储和检索，而忽略了被检索的记忆如何重塑模型推理并影响当前任务表现。我们识别出记忆诱导的认知陷阱：即使记忆被准确记录且语义相关，仍可能扭曲模型推理或信念，降低当前任务性能。为系统评估这些失效模式，我们引入MemTrapBench，涵盖两种认知陷阱形式：推理固着与信念扭曲。在两大模型族和五个代表性记忆框架上的实验表明，MemTrapBench极具挑战性：所有评估的记忆策略均劣于无记忆设置，最强方法性能下降超过10%。为缓解此类陷阱，我们提出AdaptiveMem，一种简单有效的推理时方法，指导大模型避免记忆陷阱。AdaptiveMem在保持或提升标准记忆基准性能的同时，有效缓解了认知陷阱。
原文摘要 · Abstract (English)
Memory has become a key component of large language models, enabling them to retain information and learn from long-term interactions. However, existing memory benchmarks mainly evaluate whether information is correctly extracted, stored, and retrieved, while largely overlooking how retrieved memories reshape model reasoning and affect performance on the current task. We identify memory-induced cognitive traps: even faithfully recorded and semantically relevant memories can distort model reasoning or beliefs and degrade current task performance. To systematically evaluate these failure modes, we introduce MemTrapBench, which covers two forms of cognitive traps: Reasoning Fixation and Belief Distortion. Experiments across two model families and five representative memory frameworks show that MemTrapBench is challenging: all evaluated memory strategies underperform the no-memory setting, with even the strongest methods suffering drops of more than 10%. To mitigate these cognitive traps, we propose AdaptiveMem, a simple yet effective inference-time method that instructs LLMs to avoid memory traps. AdaptiveMem mitigates cognitive traps on MemTrapBench while preserving or improving performance on standard memory benchmarks across diverse memory frameworks.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。