Preference-Aware Memory Update for Long-Term LLM Agents
This addresses a specific bottleneck in long-term LLM agents for conversational AI applications, but is incremental as it builds on existing memory storage and retrieval approaches.
The paper tackles the problem of dynamic memory updating for long-term LLM agents, which lack mechanisms for refining preference memory as user behaviors evolve, and proposes a Preference-Aware Memory Update Mechanism (PAMU) that integrates sliding window averages with exponential moving averages to capture short-term and long-term user tendencies, showing significant improvements in output quality on five task scenarios of the LoCoMo dataset.
One of the key factors influencing the reasoning capabilities of LLM-based agents is their ability to leverage long-term memory. Integrating long-term memory mechanisms allows agents to make informed decisions grounded in historical interactions. While recent advances have significantly improved the storage and retrieval components, by encoding memory into dense vectors for similarity search or organizing memory as structured knowledge graphs most existing approaches fall short in memory updating. In particular, they lack mechanisms for dynamically refining preference memory representations in response to evolving user behaviors and contexts. To address this gap, we propose a Preference-Aware Memory Update Mechanism (PAMU) that enables dynamic and personalized memory refinement. By integrating sliding window averages (SW) with exponential moving averages (EMA), PAMU constructs a fused preference-aware representation that captures both short-term fluctuations and long-term user tendencies. We conduct experiments on five task scenarios of the LoCoMo dataset, and the results show that our mechanism can significantly improve the output quality of LLM in five baselines, validating its effectiveness in long-term conversations.