PGMem: Tightly Coupled Persona-Memory Graph for Lifelong Personalized Agents

Long-term personalized dialogue agents must track user preferences as their personas evolve. Existing memory systems organize past events well, but store personas as flat profiles detached from the events that justify them. This loose coupling leads to the memory–persona validity gap and the persona-aware retrieval gap. We propose PGMem, a heterogeneous persona-memory graph that connects event and persona nodes through typed provenance and evidence edges, keeping each persona signal traceable to the events that support or revise it. At retrieval time, PGMem expands from query-relevant seeds and ranks signals by evidential validity. Across three benchmarks with small language model backbones, PG- Mem consistently outperforms summary-based, persona-aware, graph-structured, and agentic memory baselines, and improves performance as the context grows. The source code of PG- Mem is available at https://github.com/ wonjunchoi23/pgmem.

Introduction. Lifelong AI agents are increasingly expected to provide personalized interactions as large language models (LLMs) continue to advance (Zheng et al., 2026). Building such agents requires a memory module that organizes historical interactions and tracks the evolution of user personas (Zhang et al., 2025b). This dependence on explicit memory becomes even more pronounced for agents on small language models (SLMs) (Wang et al., 2025a), whose limited context windows and reasoning capabilities restrict their ability to recover such information directly from raw dialogue. One primary line of relevant work has advanced by improving memory construction and retrieval for long-term dialogue. Early methods stored past conversations as summaries or compressed memories (Zhong et al., 2024; Wang et al., 2025b; Chen et al., 2025). Subsequent work moved beyond flat compression by introducing hierarchical context management (Packer et al., 2023), segment-level memory (Pan et al., 2025), and agentic memory organization (Xu et al., 2025).

Discussion / Conclusion. In this work, we addressed a blind spot in longterm personalized dialogue: existing memory mechanisms organize past interactions well but leave the user’s evolving persona detached from the events that ground it. We traced this to two fundamental gaps—the memory–persona validity gap and the persona-aware retrieval gap. To close these gaps, we proposed PGMem, an evidence-guided heterogeneous persona-memory graph that connects event- and persona-level nodes through typed evidence edges. Rather than accumulating personas as a static profile, PGMem keeps every persona signal traceable to the evidence that supports or revises it, and at retrieval time expands from queryrelevant seeds along evidence edges to recover persona-critical evidence that direct query–memory similarity misses. Across three long-term personalized dialogue benchmarks, PGMem consistently outperforms full-history, summary-based, personaaware, graph-structured, and agentic memory baselines. We believe that organizing persona memory as an evidence graph offers a practical path toward reliable personalization, especially for resourceconstrained SLM backbones, where personalization must be carried by memory structure rather than backbone capacity.