ResearchBenchmarksCode Generation on MBPP+ (pass@1, pass@10, Avg. All)Follow69.62Pass@1Base50.95255.798560.64565.4915Jul 6, 2026Evaluation ResultsMethodMethodLinksPass@1Pass@10Average ScoreBaseModel=Qwen3-14BModel=Qwen3-14B2026.0769.6287.8867.63LP-SFTModel=Qwen3-14BModel=Qwen3-14B2026.0767.9484.864.91ASFTModel=Qwen3-14BModel=Qwen3-14B2026.0766.976.5460.22DFTModel=Qwen3-14BModel=Qwen3-14B2026.0765.7169.5553.42CEModel=Qwen3-14BModel=Qwen3-14B2026.0760.8782.1860.13EAFTModel=Qwen3-14BModel=Qwen3-14B2026.0760.5282.5860.28GEMModel=Qwen3-14BModel=Qwen3-14B2026.0751.6780.4158.1