List your company

STEM Reasoning on GPQA Diamond (Accuracy avg@4)

55Accuracy (avg@4)PAPO
26.50433.90241.348.698Mar 27, 2026

Evaluation results

MethodLinks
PAPOModel=Qwen2.5-14B2026.0355
ORM(GRPO)Model=Qwen2.5-14B2026.0347
PAPOModel=Qwen3-4B-Base2026.0343
PAPOModel=Qwen2.5-7B2026.0342.4
ORM(GRPO)Model=Qwen2.5-7B2026.0340.7
ORM(DAPO)Model=Qwen3-4B-Base2026.0339
ORM(GRPO)Model=Qwen2.5-3B2026.0338.8
PAPOModel=Qwen2.5-3B2026.0336.3
BaseModel=Qwen3-4B-Base2026.0334.9
BaseModel=Qwen2.5-14B2026.0334.4
BaseModel=Qwen2.5-7B2026.0330.1
BaseModel=Qwen2.5-3B2026.0327.6