List your company

Mathematical Reasoning on Minerva-Math (@4, @16, @64)

26.5Accuracy @4Q-RM
0.0846.94213.820.658Apr 14, 2026

Evaluation results

MethodLinks
Q-RMBase Model=Qwen3-8B Base2026.0426.52725.7
IPVRMBase Model=Qwen3-8B Base2026.0425.42526.8
Implicit PRMBase Model=Qwen3-8B Base2026.0424.624.326.5
DPO-RMBase Model=Qwen3-8B Base2026.0424.324.626.8
EndoRMBase Model=Qwen3-8B Base2026.0419.522.821.7
DPO-RMBase Model=Qwen3-0.6B Base2026.0410.79.911
Implicit PRMBase Model=Qwen3-0.6B Base2026.0410.311.411
IPVRMBase Model=Qwen3-0.6B Base2026.0410.31111.4
Q-RMBase Model=Qwen3-0.6B Base2026.049.212.110.7
IPVRMBase Model=Llama3.2-3B2026.0478.57.7
EndoRMBase Model=Qwen3-0.6B Base2026.046.38.87.7
DPO-RMBase Model=Llama3.2-3B2026.046.39.97.7
Q-RMBase Model=Llama3.2-3B2026.045.99.98.1
EndoRMBase Model=Llama3.2-3B2026.045.94.85.5
Implicit PRMBase Model=Llama3.2-3B2026.045.99.28.5
DPO-RMBase Model=Llama3.2-1B2026.043.33.72.9
Implicit PRMBase Model=Llama3.2-1B2026.042.92.63.7
Q-RMBase Model=Llama3.2-1B2026.042.22.63.3
IPVRMBase Model=Llama3.2-1B2026.041.52.64.4
EndoRMBase Model=Llama3.2-1B2026.041.11.82.2