SWE-Pruner: Self-Adaptive Context Pruning for Coding Agents

ACL ARR 2026 January Submission3385 Authors

Keywords: coding agent, code generation, long context compression

Abstract: LLM agents have demonstrated remarkable capabilities in software development, but their performance is hampered by long interaction contexts, which incur high API costs and latency. While various context compression approaches have emerged to tackle this challenge, they typically rely on fixed metrics such as perplexity (PPL), ignoring the task-specific nature of code understanding. As a result, they frequently disrupt syntactic and logical structures and fail to retain critical implementation details. In this paper, we propose SWE-Pruner, a self-adaptive context pruning framework tailored for coding agents. Drawing inspiration from how human programmers "selectively skim" source code during development and debugging, SWE-Pruner performs task-aware adaptive pruning for long contexts. Given the current task, the agent formulates an explicit goal (e.g., "focus on error handling") as a hint to guide the pruning targets. A lightweight neural skimmer (0.6B parameters) is trained to dynamically select relevant lines from the surrounding context given the goal. Evaluations across four benchmarks and multiple models validate SWE-Pruner's effectiveness in various scenarios, achieving 23-54% token reduction on agent tasks like SWE-Bench Verified and up to 14.84x compression on single-turn tasks like LongCodeQA with minimal performance impact.

Paper Type: Long

Research Area: Code Models

Research Area Keywords: LLM/AI agents, code models, code generation and understanding

Contribution Types: NLP engineering experiment, Approaches low compute settings-efficiency, Publicly available software and/or pre-trained models

Languages Studied: Python

Submission Number: 3385