List your company

Scientific Reasoning on Scientific Methods Reasoning Internal (test)

70.851-σ ScoreDirect-SFT
13.161228.138143.11558.0919Apr 6, 2026

Evaluation results

MethodLinks
Direct-SFTModel=Qwen32026.0470.8585.5684.6277.1
Reason-DAPOModel=Qwen32026.0469.437583.5272.73
Reason-SFT-DrGRPOModel=Qwen32026.0466.1976.4183.5271.35
Synthetic-Reason-SFT-DAPOModel=Qwen32026.0465.7972.5374.7268.93
Reason-SFT-DAPOModel=Qwen32026.0464.5779.2383.5271.35
Reasoning (high)Model=GPT-52026.0461.9461.2756.0461.1
Reasoning (med)Model=GPT-52026.0459.161.6256.0459.61
Reasoning (low)Model=GPT-52026.0458.758.4549.4557.65
Direct-SFTModel=Llama3.12026.0453.6458.167.0356.5
Base (Reasoning)Model=Gemini 2.5 Flash2026.0441.940.1436.2640.73
BaseModel=Llama3.12026.0437.3631.3327.9330.03
Reason-SFTModel=Qwen32026.0435.3238.945.0537.51
Base (Reasoning)Model=Llama3.12026.0426.5230.6321.9827.39
BaseModel=Qwen32026.0418.4226.0510.9920.14
Base (Reasoning)Model=Qwen32026.0415.3827.1126.1125.31