RePaCA, a Qwen2.5-Coder 3B model fine-tuned with GRPO, classifies Java patches as correct or overfitting with 83.1% accuracy and 84.8% F1 on a Defects4J-derived benchmark.
The patch also introduces a new variable `ydotonscale2` that is not used anywhere in the original code, which is not a good practice
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SE 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
RePaCA: Leveraging Reasoning Large Language Models for Static Automated Patch Correctness Assessment
RePaCA, a Qwen2.5-Coder 3B model fine-tuned with GRPO, classifies Java patches as correct or overfitting with 83.1% accuracy and 84.8% F1 on a Defects4J-derived benchmark.