From Atom to Community: Structured and Evolving Agent Memory for User Behavior Modeling

TL;DR

STEAM tackles the problem of memory for user behavior modeling by moving beyond a single unstructured memory to a structured, evolving store of atomic memories. Each memory captures a distinct interest with explicit links to observed behaviors and is organized into memory communities via prototype memories, enabling multi-order collaborative signals without heavy graph convolutions. The framework supports memory formation, consolidation, and formation, implemented in a local-to-global retrieval pipeline that scales through 2-hop BFS-based community exploration and batch updates. Empirical results on three real-world datasets show STEAM achieves state-of-the-art performance in recommendation, improves user-simulation fidelity, and enhances diversity, demonstrating the practical impact of structured memory for personalized, cross-user signaling in LLM-based agents.

Abstract

User behavior modeling lies at the heart of personalized applications like recommender systems. With LLM-based agents, user preference representation has evolved from latent embeddings to semantic memory. While existing memory mechanisms show promise in textual dialogues, modeling non-textual behaviors remains challenging, as preferences must be inferred from implicit signals like clicks without ground truth supervision. Current approaches rely on a single unstructured summary, updated through simple overwriting. However, this is suboptimal: users exhibit multi-faceted interests that get conflated, preferences evolve yet naive overwriting causes forgetting, and sparse individual interactions necessitate collaborative signals. We present STEAM (\textit{\textbf{ST}ructured and \textbf{E}volving \textbf{A}gent \textbf{M}emory}), a novel framework that reimagines how agent memory is organized and updated. STEAM decomposes preferences into atomic memory units, each capturing a distinct interest dimension with explicit links to observed behaviors. To exploit collaborative patterns, STEAM organizes similar memories across users into communities and generates prototype memories for signal propagation. The framework further incorporates adaptive evolution mechanisms, including consolidation for refining memories and formation for capturing emerging interests. Experiments on three real-world datasets demonstrate that STEAM substantially outperforms state-of-the-art baselines in recommendation accuracy, simulation fidelity, and diversity.

Figures (5)

• Figure 1: Comparison of unstructured single summary memory and our proposed STEAM for user behavior modeling in real recommendation scenario.
• Figure 2: Overview of STEAM. (a) and (b) together illustrate the memory construction process: (a) Memory Formation captures user preferences from behavior history. (b) Community Construction builds first-order collaborative relationships via relation extraction, explores communities through 2-hop BFS, and summarizes them as prototype memories. (c) Memory Retrieval first queries local agent memory, then indexes collaborative memories via corresponding prototypes. (d) Memory Evolution reflects on simulated behaviors to consolidate existing memory or form new ones.
• Figure 3: Comparison of recommendation diversity between AgentCF and STEAM.
• Figure 4: Case studies for collaborative signal exploration and how collaborative signals enhance recommendation.
• Figure 5: Performance comparison w.r.t. $k_1$ and $k_2$ of STEAM. The evaluate metric is NDCG@10.