On 300 GitHub issues from SWE-Bench Lite, a strong-first pipeline with weak-model refinement matches the strong model's resolution rate at about 60% of the cost.
Your response should contain specific implementation details that would help someone understand how to navigate, extend, and debug the codebase to solve issues
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
An Empirical Study on Strong-Weak Model Collaboration for Repo-level Code Generation
On 300 GitHub issues from SWE-Bench Lite, a strong-first pipeline with weak-model refinement matches the strong model's resolution rate at about 60% of the cost.