List your company

Correction on CriticBench (test)

75.38Math ScoreCIPO
67.330469.420271.5173.5998May 14, 2026

Evaluation results

MethodLinks
CIPOBase Model=Qwen3-4B, Training=+CIPO2026.0575.3848.7292.7274.4772.82
Qwen3-32BModel=Qwen3-32B2026.0574.235597.2176.9575.85
Qwen2.5-72B-InstructModel=Qwen2.5-72B-Instruct2026.0573.9355.2792.2677.374.69
GRPOBase Model=Qwen3-4B, Training=+GRPO2026.0570.7148.8989.7872.3470.43
InitialBase Model=Qwen3-4B2026.0567.6449.0786.6971.6368.75