ResearchBenchmarksCode Reasoning on HumanEval+ (avg@10)Follow41.6Avg Score (k=10)AntiSD40.35240.6764141.324May 12, 2026Evaluation ResultsMethodMethodLinksAvg Score (k=10)AntiSDModel=Qwen3-8BModel=Qwen3-8B2026.0541.6GRPOModel=Qwen3-8BModel=Qwen3-8B2026.0540.4