SWE-AGILE: A Software Agent Framework for Efficiently Managing Dynamic Reasoning Context

TL;DR

SWE-AGILE is a novel software agent framework that balances reasoning depth and efficiency in dynamic contexts by maintaining a sliding window of detailed reasoning and compressing historical data.

Contribution

It introduces a Dynamic Reasoning Context strategy that effectively manages reasoning history, improving multi-turn SWE task performance.

Findings

Sets new performance standards for 7B-8B models on SWE-Bench-Verified.

Uses only 2.2k trajectories and 896 tasks for empirical evaluation.

Achieves better reasoning efficiency without losing context.

Abstract

Prior representative ReAct-style approaches in autonomous Software Engineering (SWE) typically lack the explicit System-2 reasoning required for deep analysis and handling complex edge cases. While recent reasoning models demonstrate the potential of extended Chain-of-Thought (CoT), applying them to the multi-turn SWE task creates a fundamental dilemma: retaining full reasoning history leads to context explosion and ``Lost-in-the-Middle'' degradation, while discarding it would force the agent to redundantly re-reason at every step. To address these challenges, we propose SWE-AGILE, a novel software agent framework designed to bridge the gap between reasoning depth, efficiency, and context constraints. SWE-AGILE introduces a Dynamic Reasoning Context strategy, maintaining a ``sliding window'' of detailed reasoning for immediate continuity to prevent redundant re-analyzing, while compressing historical reasoning content into concise Reasoning Digests. Empirically, SWE-AGILE sets a new standard for 7B-8B models on SWE-Bench-Verified using only 2.2k trajectories and 896 tasks. Code is available at https://github.com/KDEGroup/SWE-AGILE.