ResearchBenchmarksCode Reasoning on MBPP+Follow63.3Avg@10AntiSD60.90861.52962.1562.771May 12, 2026Evaluation ResultsMethodMethodLinksAvg@10AntiSDModel=Qwen3-8BModel=Qwen3-8B2026.0563.3GRPOModel=Qwen3-8BModel=Qwen3-8B2026.0561