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

TL;DR

MemReread is a novel memory-guided rereading approach that improves agentic long-context reasoning by recovering discarded evidence through question decomposition and dynamic rereading, maintaining linear complexity.

Contribution

It introduces MemReread, which enhances long-context reasoning by avoiding intermediate retrieval, enabling indirect fact recovery, and dynamically controlling rereading passes with reinforcement learning.

Findings

Outperforms baseline frameworks on long-context reasoning tasks.

Maintains linear time complexity with respect to context length.

Supports non-linear reasoning while preserving logical flow.

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.