MemSearch-o1: Empowering Large Language Models with Reasoning-Aligned Memory Growth in Agentic Search

2026-04-19

AI Summary

This work addresses the challenges of memory inflation and information dilution in large language model (LLM)-based agent search, where iterative think-retrieve loops often obscure critical information due to inadequate modeling of fine-grained semantic relationships between queries and retrieved documents. To overcome this limitation, the authors propose MemSearch-o1, a novel framework that restructures memory management from sequential concatenation to a reasoning-aligned, structured mechanism. MemSearch-o1 dynamically generates fine-grained memory segments anchored to query seed tokens, optimizes their semantic contributions via a contribution function, and constructs globally connected memory paths to enable path-guided reasoning. Evaluated across eight benchmark datasets, the method substantially mitigates memory dilution and effectively enhances the reasoning capabilities of diverse LLMs, offering a new paradigm for building memory-aware agent systems.

Technology Category

• Search and Optimization: Learning to Search
• Machine Learning: Large Multimodal Models (LMMs)
• Multiagent Systems: Agent Communication

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Recent advances in large language models (LLMs) have scaled the potential for reasoning and agentic search, wherein models autonomously plan, retrieve, and reason over external knowledge to answer complex queries. However, the iterative think-search loop accumulates long system memories, leading to memory dilution problem. In addition, existing memory management methods struggle to capture fine-grained semantic relations between queries and documents and often lose substantial information. Therefore, we propose MemSearch-o1, an agentic search framework built on reasoning-aligned memory growth and retracing. MemSearch-o1 dynamically grows fine-grained memory fragments from memory seed tokens from the queries, then retraces and deeply refines the memory via a contribution function, and finally reorganizes a globally connected memory path. This shifts memory management from stream-like concatenation to structured, token-level growth with path-based reasoning. Experiments on eight benchmark datasets show that MemSearch-o1 substantially mitigates memory dilution, and more effectively activates the reasoning potential of diverse LLMs, establishing a solid foundation for memory-aware agentic intelligence.

Problem

• memory dilution
• agentic search
• memory management
• semantic relations
• large language models

Innovation

• reasoning-aligned memory
• memory growth
• agentic search
• memory dilution
• token-level memory