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

Hugging Face Daily Papers Papers

Summary

MemReread introduces a method for long-context reasoning that avoids intermediate retrieval by decomposing questions and rereading text to recover discarded information, achieving linear time complexity. It outperforms baseline frameworks on long-context reasoning tasks.

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.
Original Article
View Cached Full Text

Cached at: 05/14/26, 08:17 AM

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

Source: https://huggingface.co/papers/2605.10268

Abstract

MemReread addresses long-context reasoning challenges by avoiding intermediate retrieval and employing question decomposition with rereading to recover discarded information, maintaining linear time complexity.

To tacklelong-context reasoningtasks without the quadratic complexity of standardattention mechanisms, approaches based onagent memoryhave 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 integratedretrieval modulesthat allow agents to recall information previously discarded duringmemory 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 uponstreaming reading, MemReread circumvents intermediate retrieval. It triggersquestion decompositionandrereadingwhen 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 areinforcement learningframework that enhanceslength extrapolationcapability while dynamically determining the number ofrereadingpasses based on task complexity, thereby flexibly controlling computational overhead. Extensive experiments demonstrate that MemReread consistently outperforms baseline frameworks onlong-context reasoningtasks, while maintaining linear time complexity with respect to context length.

View arXiv pageView PDFGitHub1Add to collection

Get this paper in your agent:

hf papers read 2605\.10268

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2605.10268 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2605.10268 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2605.10268 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

ReM-MoA: Reasoning Memory Sustains Mixture-of-Agents Scaling

arXiv cs.AI

ReM-MoA introduces a memory-augmented Mixture-of-Agents framework that sustains scaling through ranked reasoning memory and curated diversified memory routing, outperforming prior MoA variants across five reasoning benchmarks.

ConvMem: Convolutional Memory for Long-Context Reasoning

arXiv cs.AI

ConvMem is a training-free, parallelizable framework that reformulates long-context reasoning in large language models as hierarchical convolution to improve efficiency, avoid overfitting, and outperform baseline methods.