List your company

Mathematical Reasoning on Olympiad Bench (accuracy, Average)

44AccuracyBRIDGE
21.6427.44533.2539.055Sep 8, 2025

Evaluation results

MethodLinks
BRIDGEBackbone=Qwen3-8B-Base2025.094449.9
LUFFYBackbone=Qwen3-8B-Base2025.0943.144
RLBackbone=Qwen3-8B-Base2025.0942.442.9
CHORDBackbone=Qwen3-8B-Base2025.0942.245.9
SRFTBackbone=Qwen3-8B-Base2025.094039.8
SFT→RLBackbone=Qwen3-8B-Base2025.0939.645.5
SFT+RLBackbone=Qwen3-8B-Base2025.0939.240.1
SFTBackbone=Qwen3-8B-Base2025.0929.837.6
BaseBackbone=Qwen3-8B-Base2025.0922.526.6