Learning What to Share and What to Personalize: Hierarchical Strategy Co-Evolution for Agent Memory

arXiv:2608.25329v1 Announce Type: cross Abstract: Memory-augmented agents maintain compact user profiles throughout extended conversations, enabling personalized and consistent responses without the need to process the entire dialogue history. The quality of these user profiles relies on the underlying memory management strategy: at each step, the agent must determine what to retain, compress, or discard. However, existing methods typically employ a static, one-size-fits-all strategy established before training. In practice, the optimal memory decision is inherently user-specific and dynamically evolves alongside policy optimization. To address this, we propose extbf{HiPS} (extbf{Hi}erarchical extbf{P}ersonalized extbf{S}trategy), a framework that decouples memory management into a globally shared foundation and a user-specific adaptive tier. Specifically, HiPS employs extbf{Universal Strategy} to extract shared principles from cross-persona trajectories, alongside extbf{Persona Delta Distillation} to generate tailored rules for users whose behaviors diverge from general patterns. extbf{Cross-Level Rule Flow} dynamically calibrates their boundary by promoting broadly validated personal rules and demoting contradicted global ones. The architecture establishes a co-evolution loop where a mechanism guarantees that all strategy refinements are anchored to task outcomes. Extensive experiments demonstrate consistent improvements over memory-augmented baselines.

Related

• Learning How and What to Memorize: Cognition-Inspired Two-Stage Optimization for Evolving Memory
• MemReader: From Passive to Active Extraction for Long-Term Agent Memory
• When Memory Lies: An Empirical Study of Spatial Memory Staleness in VLM Agents
• Learning User-Aware Recall: Personalized Retrieval in Long-Term Conversational Memory

Source: arXiv cs.CL | 2026-08-27