arXiv：cs.AI· Arman Behnam, Binghui Wang·· 4 小时前AI 评分43
因果记忆策略 CMP：通过对检索做干预让记忆效用可识别
Causal Memory Policy: Making Memory Utility Identifiable by Intervening on Retrieval
AI 导读
针对记忆增强大语言模型仅依赖已检索记忆、导致未被检索的记忆效用无法识别的问题，研究者提出因果框架 Causal Memory Policy（CMP），通过对检索本身做干预、预留固定上下文槽位并按已知倾向采样，用自归一化逆倾向加权估计记忆效用。
正文
Abstract:Memory-augmented large language models must decide which memories to retain, and recent systems do so by estimating each memory's effect on task performance. However, these estimates rely entirely on retrieved memories. When a memory is never retrieved, store-level interventions produce identical outcomes, leaving its utility unidentified. This is a retrieval-level positivity violation, invisible to diagnostics that examine only memory operations. We introduce Causal Memory Policy (CMP), a causal framework that restores identification by intervening on retrieval itself, reserving a fixed number of context slots for memories sampled with known propensities. CMP estimates memory utility by self-normalized inverse propensity weighting under a balanced assignment design. We prove the causal factorization of memory utility through retrieval, the unbiasedness and exact variance of the estimator, and the optimal decision rule under irreversible operations. Empirically, identification fails for 54% of required memories on LongMemEval and 67% on LoCoMo, and the failure persists in a deployed memory system. CMP improves discrimination between required and non-required memories from 0.54 to 0.66 AUC. Finally, we show that identified memory utility alone is insufficient for retention decisions: per-query utility reaches 0.78 AUC on the query for which it is estimated, yet no aggregation available to a retention policy predicts a memory's value on unseen queries. Code is available at: this https URL.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.02070 [cs.AI] |
| (or arXiv:2610.02070v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.02070 arXiv-issued DOI via DataCite |
Submission history
From: Arman Behnam [view email]
[v1]
Thu, 1 Oct 2026 17:11:25 UTC (131 KB)
[v2]
Fri, 2 Oct 2026 15:55:32 UTC (131 KB)
来源：arXiv：cs.AI · arxiv.org