MemSIF: From Structured Interactions to Dual-Track Fact Memory for LLM Agents

2026-08-03

AI Summary

This work addresses memory degradation in large language model agents during long-term interactions, caused by temporal-structural misalignment (TSM) and delayed utility manifestation (DUM). To mitigate these issues, the authors propose a structured interaction memory framework coupled with a dual-track factual memory mechanism. The approach organizes raw interactions into thematically coherent Topical Segments and temporally continuous Event Trajectories, while dynamically managing factual knowledge through CoreFact and ActiveFact tracks. Introducing a novel paradigm that transforms structured interactions into factual memory, the method employs multi-source verification and query-driven fact generation to substantially enhance the accuracy and utility of long-term memory. Evaluated on the LoCoMo and LongMemEval-S benchmarks, the proposed framework consistently achieves state-of-the-art performance across five prominent large language models, surpassing the strongest baseline by 2.29%–8.79% on LoCoMo and 2.87%–6.15% on LongMemEval-S in overall accuracy.

Abstract

Long-term memory is critical for LLM agents operating over long-horizon interactions. However, several persistent limitations of existing memory systems can be traced to two recurring misalignment patterns in long-term interaction settings: Temporal-Structural Misalignment (TSM) and Delayed Utility Manifestation (DUM). TSM arises when temporal proximity does not reliably align with topical or event-level relatedness, whereas DUM arises when write-time salience does not reliably predict future query utility. To mitigate these misalignment patterns, we propose MemSIF (Memory with Structured Interactions and Facts), a structured interaction-to-fact memory framework. Structured Interaction Memory organizes raw interactions into Topical Segments that preserve local topical coherence and Event Trajectories that maintain cross-time event continuity. Dual-Track Fact Memory uses two complementary tracks: CoreFact memory consolidates stable, schema-guided information at write time, whereas ActiveFact memory forms facts on demand and promotes those supported by multiple historical sources and recurring query demand for reuse. Experiments on LoCoMo and LongMemEval-S across five backbone LLMs show that MemSIF achieves the highest Total ACC in all settings, outperforming the strongest baseline by 2.29%-8.79% on LoCoMo and 2.87%-6.15% on LongMemEval-S. These results support the effectiveness of combining Structured Interaction Memory with Dual-Track Fact Memory to mitigate TSM and DUM. Code is available at https://github.com/luoyufeihaha/MemSIF.

Problem

• Temporal-Structural Misalignment
• Delayed Utility Manifestation
• long-term memory
• LLM agents
• memory systems

Innovation

• Structured Interaction Memory
• Dual-Track Fact Memory
• Temporal-Structural Misalignment
• Delayed Utility Manifestation
• Long-term Memory for LLM Agents