Agent memory with entity-based indexing improves long-term interaction recall
EnSIMem: Entity-Structured Indexing for Long-Term Agent Memory
Artificial IntelligenceComputation and Language
Summary
When digital assistants talk to people over a long time, they need to remember facts, preferences, and past events correctly. Existing systems often jumble information into broad summaries or find bits of text without clear context. The authors created EnSIMem, a system that organizes conversations around specific entities and their properties, keeping track of when and where each fact came from. This helps the assistant answer questions more accurately by using exact remembered details instead of vague summaries.
What this means in practice
- •For virtual assistant developers: Develop assistants that recall user details and past events with precise entity-based memory over extended interactions.
- •For customer service platform engineers: Implement memory modules that reliably track customer preferences and interaction history to improve personalized support.
Authors
Xuanyu Meng, Xing Fan, Xinyi Fan, Chenlei Guo, Yixuan Xie, Jiawei Han
Abstract
An agent that interacts with users over long periods must recall facts, preferences, events, and changes from a continuously growing interaction history. Existing memory systems often compress interactions into generic summaries or retrieve anonymous text chunks, making it difficult for an agent to identify the correct entity, property, and supporting evidence. We present EnSIMem, an entity-structured long-term memory architecture for an agent. During offline construction, the system organizes interactions into theme-coherent episodes and builds dialogue-grounded index entries of the form [entity][entity type][property:value]. Each entry preserves its source turns, temporal information, and available multimodal fields. During online interaction, the agent's request is decomposed into evidence requirements whose properties are aligned with the memory index. Entity-property lookup and adaptive retrieval then collect the evidence needed for point, temporal, compositional, and aggregation reasoning. The agent generates its response from the preserved source evidence rather than from lossy memory summaries. On long-term agent-memory benchmarks, EnSIMem achieves high answer accuracy while maintaining compact contexts and favorable online efficiency. These results show that entity-structured indexing and episode-level provenance provide a reliable foundation for long-term memory in agents. The code of our model is available at https://github.com/RamonMeng/EnSIMem.