Memory Beyond Recall: A Dual-Process Cognitive Memory System for Self-Evolving LLM Agents
2026-06-08 • Computation and Language
Computation and LanguageArtificial Intelligence
AI summaryⓘ
The authors explain that long-term memory for language models is not just about finding the right piece of information but also about understanding changes and connections over time. They introduce DCPM, a new system that organizes memory into layers, from basic facts to complex ideas and intentions, inspired by how human thinking works in two parts: quick daytime recording and slower nighttime processing. Their tests show the nighttime system improves understanding of user changes across sessions more than just recalling exact information. This approach helps the model better guess what a user means based on their past behavior.
Long-term memoryLanguage model (LLM)Belief revisionDual-process theoryCognitive hierarchyDiachronic belief trajectoriesSchema inductionCross-domain inferenceMemory recallPersonaMem benchmarks
Authors
Tianxiang Fei, Mingyang Song, Mao Zheng, Xiang Yu
Abstract
Long-term memory for an LLM agent is more than retrieving the right passage at the right time. Current memory systems collapse belief revision, causal coupling, and cross-domain abstraction into a single retrieval surface tuned for surface recall, and consequently struggle on implicit personalisation that requires reasoning over how a user has evolved. We propose DCPM, which reorganises agent memory along a cognitive capability hierarchy ascending from raw inputs and atomic facts, through diachronic belief trajectories and identity, to domain schemas, latent intentions and cross-domain patterns. The hierarchy is driven by two processes inheriting the architectural split of dual-process theory: a synchronous daytime writer (System1) that records belief revisions as doubly linked supersedes chains, and an asynchronous nighttime engine (System2) that induces schemas and intentions and sweeps for cross-domain collisions abstracted into higher-level core schemas. On LongMemEval, PersonaMem and PersonaMem-v2, enabling System2 contributes most where the benchmark rewards implicit cross-session inference (up to +5.20 on PersonaMem-v2) and least on span recall, matching the architectural prediction.