List your company

Science Reasoning on GPQA (r* Accuracy and r_self)

30.9r* AccuracyPhi
9.16414.80720.4526.093May 6, 2026

Evaluation results

MethodLinks
PhiVariant=Base, Sampling=Power, Evaluation Protocol=All completions2026.0530.90.215
PhiVariant=Base, Sampling=Power, Evaluation Protocol=Self-reward Best-of-N2026.0530.90.214
PhiVariant=Distilled, Sampling=Temperature, Evaluation Protocol=Self-reward Best-of-N2026.0529.80.262
PhiVariant=Distilled, Sampling=Temperature, Evaluation Protocol=All completions2026.0529.20.284
QwenVariant=Distilled, Sampling=Temperature, Evaluation Protocol=Self-reward Best-of-N2026.0529.10.201
QwenVariant=Base, Sampling=Power, Evaluation Protocol=Self-reward Best-of-N2026.0528.70.118
QwenVariant=Distilled, Sampling=Temperature, Evaluation Protocol=All completions2026.0528.50.21
QwenVariant=Base, Sampling=Power, Evaluation Protocol=All completions2026.0528.30.118
Qwen-MathVariant=Distilled, Sampling=Standard, Evaluation Protocol=Self-reward Best-of-N2026.0528.10.113
QwenVariant=Distilled, Sampling=Standard, Evaluation Protocol=All completions2026.05280.437
Qwen-MathVariant=Base, Sampling=Power, Evaluation Protocol=Self-reward Best-of-N2026.0527.90.087
QwenVariant=Distilled, Sampling=Standard, Evaluation Protocol=Self-reward Best-of-N2026.0527.80.426
Qwen-MathVariant=Base, Sampling=Power, Evaluation Protocol=All completions2026.0527.70.088
Qwen-MathVariant=Distilled, Sampling=Temperature, Evaluation Protocol=All completions2026.0527.70.149
Qwen-MathVariant=Distilled, Sampling=Temperature, Evaluation Protocol=Self-reward Best-of-N2026.0527.70.109
Qwen-MathVariant=Distilled, Sampling=Standard, Evaluation Protocol=All completions2026.0527.50.165
PhiVariant=Distilled, Sampling=Standard, Evaluation Protocol=All completions2026.0526.80.321
PhiVariant=Distilled, Sampling=Standard, Evaluation Protocol=Self-reward Best-of-N2026.0526.70.319
QwenVariant=Base, Sampling=Standard, Evaluation Protocol=Self-reward Best-of-N2026.0524.51.527
QwenVariant=Base, Sampling=Standard, Evaluation Protocol=All completions2026.0524.41.531
PhiVariant=Base, Sampling=Standard, Evaluation Protocol=All completions2026.0522.30.802
PhiVariant=Base, Sampling=Standard, Evaluation Protocol=Self-reward Best-of-N2026.0522.30.8
Qwen-MathVariant=Base, Sampling=Standard, Evaluation Protocol=Self-reward Best-of-N2026.0510.30.675
Qwen-MathVariant=Base, Sampling=Standard, Evaluation Protocol=All completions2026.05100.675