MemReread: Enhancing Agentic Long-Context Reasoning via Memory-Guided Rereading

2026-05-11

AI Summary

This work addresses critical limitations in existing long-context reasoning methods, which suffer from evidence loss due to linear reading and interference from ineffective queries in retrieval-based recall. To overcome these issues, the authors propose MemReread, a novel framework that eschews conventional intermediate retrieval modules and instead integrates question decomposition with a memory-guided rereading mechanism. When the accumulated memory proves insufficient for answering a query, the framework dynamically triggers nonlinear rereading to recover indirectly relevant facts prematurely discarded during initial passes. By leveraging reinforcement learning to adaptively control the number of rereading iterations, MemReread achieves logically coherent multi-hop reasoning while preserving linear time complexity. Experimental results demonstrate that MemReread significantly outperforms current baselines across multiple long-context reasoning benchmarks.

Abstract

To tackle long-context reasoning tasks without the quadratic complexity of standard attention mechanisms, approaches based on agent memory have emerged, which typically maintain a dynamically updated memory when linearly processing document chunks. To mitigate the potential loss of latent evidence in this memorize-while-reading paradigm, recent works have integrated retrieval modules that allow agents to recall information previously discarded during memory overwriting. However, retrieval-based recall suffers from both evidence loss during memory formation and interference induced by invalid queries. To overcome these limitations, we propose MemReread. Built upon streaming reading, MemReread circumvents intermediate retrieval. It triggers question decomposition and rereading when the final memory is insufficient, enabling the recovery of indirect facts that were prematurely discarded. This design supports non-linear reasoning while preserving the inherent logical flow of document comprehension. To further enhance practicality, we introduce a reinforcement learning framework that enhances length extrapolation capability while dynamically determining the number of rereading passes based on task complexity, thereby flexibly controlling computational overhead. Extensive experiments demonstrate that MemReread consistently outperforms baseline frameworks on long-context reasoning tasks, while maintaining linear time complexity with respect to context length.

Problem

• long-context reasoning
• agent memory
• evidence loss
• retrieval interference
• memory overwriting

Innovation

• memory-guided rereading
• long-context reasoning
• agent memory
• reinforcement learning
• linear time complexity