List your company

Mathematical Reasoning on AIME24 (accuracy, Average)

16.7AccuracySFT→RL
2.7646.3821013.618Sep 8, 2025

Evaluation results

MethodLinks
SFT→RLBackbone=Qwen3-8B-Base2025.0916.745.5
BRIDGEBackbone=Qwen3-8B-Base2025.0916.749.9
SFTBackbone=Qwen3-8B-Base2025.0913.337.6
CHORDBackbone=Qwen3-8B-Base2025.0913.345.9
RLBackbone=Qwen3-8B-Base2025.091042.9
SFT+RLBackbone=Qwen3-8B-Base2025.091040.1
LUFFYBackbone=Qwen3-8B-Base2025.091044
SRFTBackbone=Qwen3-8B-Base2025.096.739.8
BaseBackbone=Qwen3-8B-Base2025.093.326.6