学习该记住什么：一种基于认知的多因素记忆价值模型
Learning What to Remember: A Cognitively Grounded Multi-Factor Value Model for Agentic Memory
AI总结：
针对长期LLM代理的记忆管理问题，提出一种基于认知心理学的多因素记忆价值函数，通过无梯度优化学习权重，统一控制编码深度、遗忘风险和检索排名，在LongMemEval上显著优于单一因素和近因策略。
摘要：
长期运行的LLM代理积累的交互历史远超任何上下文窗口，迫使面临一个持续决策：在固定记忆预算下，哪些内容应深度编码、哪些应遗忘、哪些应检索。生产系统采用语义相似性或近因性——两者对于遗忘决策都是错误指定的，因为遗忘决策是在未来查询未知的整合时刻做出的。我们提出一个多因素记忆价值函数 V(m)=∑_i w_i f_i(m)，涵盖七个可解释因素（情感强度、目标相关性、价值对齐、自我/用户相关性、任务效用、可靠性和使用历史），这些因素来自认知心理学，其权重通过无梯度优化器从下游目标中学习，并且该单一标量统一控制编码深度、遗忘风险和检索排名。我们提出一个方法论观点：在LongMemEval上，针对保留的评估问题对目标相关性进行评分，使得黄金证据保留率达到≈0.98——这衡量的是检索，而非遗忘。在现实盲态模式下，学习到的多因素价值在479个可用案例中保留了0.770±0.011的黄金证据，而均匀权重为0.657，最佳单一因素为0.518，近因性为0.368；每对差距的95%自助法置信区间均高于零，且基于相同因素的神经网络与线性模型持平。学习到的权重是可解释的——可靠性、情感强度和自我/用户相关性占主导，而查询时的目标相似性在遗忘决策中被正确降权。一个带有植入混淆的受控合成任务证实，学习器恢复了分离性权重（保留率1.00），而均匀权重失败（0.62）。该基础架构是开源的；所有实验在单CPU上运行，无需API调用。
英文摘要：
Long-running LLM agents accumulate interaction histories far larger than any context window, forcing a standing decision: what to encode deeply, what to forget, and what to retrieve under a fixed memory budget. Production systems answer with semantic similarity or recency -- both mis-specified for the forgetting decision, which is made at consolidation time before the future query is known. We propose a multi-factor memory value function V(m)=\sum_i w_i f_i(m) over seven interpretable factors (emotional intensity, goal relevance, value alignment, self/user relevance, task utility, reliability, and usage history) drawn from cognitive psychology, whose weights are learned from a downstream objective by a gradient-free optimiser, and whose single scalar uniformly controls encoding depth, forget risk, and retrieval rank. We make a methodological point: on LongMemEval, scoring goal relevance against the held-out evaluation question saturates gold-evidence retention at \approx 0.98 -- this measures retrieval, not forgetting. In the realistic blind regime, a learned multi-factor value retains 0.770 \pm 0.011 of gold evidence across 479 usable cases, versus 0.657 for uniform weights, 0.518 for the best single factor, and 0.368 for recency; every paired gap's 95% bootstrap CI is above zero, and a neural network over the same factors ties the linear model. The learned weights are interpretable -- reliability, emotional intensity, and self/user relevance dominate, while query-time goal similarity is correctly down-weighted for the forgetting decision. A controlled synthetic task with planted confounds confirms the learner recovers a separating weighting (1.00 retention) where uniform weighting fails (0.62). The substrate is open-source; all experiments run on a single CPU with no API calls.
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。
