Coding Agents are Effective Long-Context Processors

TL;DR

This paper demonstrates that coding agents, which organize text in file systems and manipulate it with native tools, significantly outperform traditional LLMs in processing long contexts across various tasks.

Contribution

The work introduces a novel approach where coding agents externalize long-context processing, leveraging file system organization and native tools to improve performance.

Findings

Coding agents outperform state-of-the-art by 17.3% on average.

Native tool proficiency enhances long-context reasoning.

File system familiarity enables efficient navigation of large text corpora.

Abstract

Large Language Models (LLMs) have demonstrated remarkable progress in scaling to access massive contexts. However, the access is via the latent and uninterpretable attention mechanisms, and LLMs fail to effective process long context, exhibiting significant performance degradation as context length increases. In this work, we study whether long-context processing can be externalized from latent attention into explicit, executable interactions, by allowing coding agents to organize text in file systems and manipulate it using its native tools. We evaluate off-the-shelf frontier coding agents as the general interface for tasks that require processing long contexts, including long-context reasoning, retrieval-augmented generation, and open-domain question answering with large-scale corpus contains up to three trillion tokens. Across multiple benchmarks, these agents outperform published state-of-the-art by 17.3% on average. We attribute this efficacy to two key factors: native tool proficiency, which enables agents to leverage executable code and terminal commands rather than passive semantic queries, and file system familiarity, which allows them to navigate massive text corpora as directory structures. These findings suggest that delegating long-context processing to coding agents offers an effective alternative to semantic search or context window scaling, opening new directions for long-context processing in LLMs.