Coding agents fix more issues when shown whole files
Original title: Skip the Compressor: Coding Agents Fix More Issues When They Pick Whole Files
On 300 SWE-rebench issues, verbatim file selection beats compressors by 6.3 points with either model.
The paper argues that issue patches need the files that must change, shown exactly as written, not similarity-ranked chunks or pruned tokens. At an equal per-issue token budget on 300 recent SWE-rebench issues, selector CQ2-CARE resolves more issues than three compressors and three retrievers with GPT-5.6 and Claude Opus 5. It also beats SweRank and SWE-Pruner, two methods trained for issue resolution. Retrievers often miss files to edit, and compressors keep them but alter text so fewer patches apply.
- CQ2-CARE resolves 43.7% with GPT-5.6 and 46.7% with Claude Opus 5.
- It leads the best retriever by 9.3 and 12.7 points on the two models.
- It beats SweRank by 0.7 and 1.7 points and SWE-Pruner by 2.0 and 7.3.
- Patches are graded with the benchmark's tests at a matched token budget.
- Products
- CQ2-CARE
- AI models
- GPT-5.6Claude Opus 5
From the original article
1 min read at huggingface.co
A coding agent works on repositories that can be far larger than its context window can hold. Prompt compressors and retrievers offer solutions to this problem, but patching an issue needs two things at once: the files that must change, and their code exactly as written, since the model must reproduce the lines it edits.
By Xiaocong Yang, Rishit Chatterjee, Himanshu Udupi · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.