ResearchBenchmarksCoding on HumanEval (Accuracy and Avg Length)Follow75AccuracyBase40.118449.174258.2367.2858Apr 7, 2026Evaluation ResultsMethodMethodLinksAccuracyAverage LengthBaseModel=DeepSeek-R1-Dist...Model=DeepSeek-R1-Distill-7B2026.04753,975.97ETRModel=DeepSeek-R1-Dist...Model=DeepSeek-R1-Distill-7B2026.04751,057.99ETRModel=Qwen3-4BModel=Qwen3-4B2026.0453.661,838.6BaseModel=Qwen3-4BModel=Qwen3-4B2026.0441.463,397.2