为大模型智能体设计可解释的记忆价值模型，精准决定该记住什么、忘掉什么。
Learning What to Remember: A Cognitively Grounded Multi-Factor Value Model for Agentic Memory
- 基于认知心理学七因子构建记忆价值函数，用无梯度优化学习权重。
- 在真实盲测中保留98%黄金证据，显著优于传统方法（最高达77%）。
- 权重可解释，适合需可信长期记忆的AI系统开发者使用。
长期运行的大语言模型智能体积累的交互历史远超上下文窗口容量，必须在固定记忆预算下决定哪些信息深度编码、哪些遗忘、哪些检索。现有系统依赖语义相似性或时间久近，但这些指标在巩固阶段无法预测未来查询，因而不适用于遗忘决策。本文提出一个基于七种可解释因素（情感强度、目标相关性、价值一致性、自我/用户相关性、任务效用、可靠性、使用历史）的多因子记忆价值函数 V(m)=∑w_i f_i(m)，权重由下游目标通过无梯度优化器学习，单一标量统一控制编码深度、遗忘风险与检索排序。在LongMemEval上，以目标相关性评分作为评估问题的参考时，黄金证据保留率可达约0.98——这衡量的是检索性能而非遗忘。在更真实的盲测场景中，学习到的多因子价值函数在479个有效案例中保留0.770±0.011的黄金证据，显著优于均匀权重（0.657）、最佳单因子（0.518）和时间久近（0.368）；所有成对差异的95%置信区间均高于零。神经网络模型表现与线性模型持平。学习到的权重具有可解释性：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被正确降权。在带有植入干扰项的受控合成任务中，该学习器实现了1.00的保留率，而均匀加权仅达0.62。实验全部在单个CPU上完成，代码开源。
原文摘要 · Abstract (English)
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.
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。