The authors developed SWE-Critix, a post-training pipeline for a Qwen3-30B mixture-of-experts model.
Post-training makes coding-agent trajectory verifiers substantially more accurate
SWE-Critix raises successful-versus-failed trajectory classification accuracy from 57% to 78% while generating written explanations.
Chinese Tech
Jiajun Hu · Chun Yong Chong
Leibniz Research Center Huawei · Monash University Malaysia
Research Digest··2 min read
Hu and Chong train a general-purpose verifier to assess an entire coding agent’s execution history, rather than judging only its final patch or returning a scalar score.
Why this paper
From Leibniz Research Center Huawei and Monash University Malaysia
In one line
Post-training a 30B coding-trajectory verifier with supervised fine-tuning and reinforcement learning raises completion-classification accuracy from 57% to 78%.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ✓Reports numbers on named benchmarks
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§