Infini Memory: Maintainable Topic Documents for Long-Term LLM Agent Memory

2026-06-09

AI Summary

Current large language model agents lack a persistent memory mechanism that is maintainable, updatable, and capable of providing evidence across sessions. This work proposes a topic-document-based textual persistent memory architecture that organizes memories into structured thematic units, temporarily buffers new observations, and periodically integrates them to support factual revision while preserving metadata. Additionally, it introduces an LLM-driven iterative tool-calling mechanism for precise retrieval. The approach significantly enhances the coherence, updatability, and retrieval effectiveness of long-term memory, achieving a 64.7% overall score on MemoryAgentBench. Ablation studies confirm the complementary benefits of maintaining thematic structure and performing iterative evidence verification.

Abstract

Long-term LLM agents need persistent memory that can track changing facts and provide relevant evidence across sessions. Existing memory systems often store observations as isolated records, summaries, or indexed fragments, which makes evidence aggregation, fact revision, and memory maintenance difficult. We propose Infini Memory, a maintainable text-based persistent memory architecture that treats agent memory as topic-structured documents. Each topic document serves as a semantic unit for collecting related evidence, preserving metadata, and revising facts over time. New observations are first staged in a buffer and periodically consolidated into coherent textual contexts. At inference time, an agentic retrieval procedure lets the LLM read memory through iterative tool calls rather than a single retrieval step. On MemoryAgentBench, Infini Memory achieves 64.7% overall score. Ablations show that topic-structured maintenance and iterative evidence inspection improve complementary aspects of long-term memory use.

Problem

• long-term memory
• LLM agent
• memory maintenance
• fact revision
• evidence aggregation

Innovation

• topic-structured memory
• maintainable memory
• iterative retrieval
• long-term LLM agents
• evidence consolidation