SWE-AGILE: A Software Agent Framework for Efficiently Managing Dynamic Reasoning Context
Summary: arXiv:2604.11716v1 Announce Type: new
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.
Introduction
The introduction of advanced reasoning models has opened new avenues for improving autonomous Software Engineering (SWE). However, existing methods often struggle with the complexity of multi-turn tasks, leading to inefficiencies and potential errors. SWE-AGILE aims to rectify these issues by introducing a framework that balances depth of reasoning with operational efficiency.
Key Features of SWE-AGILE
- Dynamic Reasoning Context: This innovative approach allows agents to maintain a relevant context without overwhelming the system with excessive data.
- Sliding Window Concept: By focusing on a limited scope of reasoning, SWE-AGILE minimizes the risk of losing crucial information while also preventing redundant processing.
- Reasoning Digests: Historical reasoning is compressed into concise summaries, ensuring that agents can quickly access essential insights without sifting through voluminous data.
- Empirical Validation: SWE-AGILE has demonstrated superior performance on SWE-Bench-Verified, showcasing its effectiveness across various tasks.
Impact on Autonomous Software Engineering
The introduction of SWE-AGILE signifies a major advancement in the field of autonomous Software Engineering. By effectively managing reasoning context, this framework not only improves the efficiency of agents but also enhances their ability to navigate complex scenarios. The implications of this development are profound, potentially leading to more robust and reliable software solutions.
Conclusion
SWE-AGILE represents a significant step forward in the integration of advanced reasoning strategies within autonomous systems. With its unique features and empirical success, it sets a new benchmark for future research and applications in the realm of Software Engineering. Researchers and practitioners are encouraged to explore the framework further, contributing to the ongoing evolution of autonomous software agents.