List your company

Commonsense Question Answering on CSQA (OOD)

63.8AccuracyR1 Distill -> GRPO
46.01650.63355.2559.867Dec 3, 2025

Evaluation results

MethodLinks
R1 Distill -> GRPOtraining=SFT + GRPO2025.1263.8
BOLT -> GRPOtraining=SFT + GRPO2025.1262.8
RL-Onlytraining=RL2025.1262.6
SkillFactory -> GRPOtraining=SFT + GRPO2025.1260.8
STaR -> GRPOtraining=SFT + GRPO2025.1260.5
R1 Distilltraining=SFT2025.1256.6
Qwen2.5 1.5B Instructtraining=None2025.1255.7
STaRtraining=SFT2025.1255.4
SkillFactorytraining=SFT2025.1247.1
BOLTtraining=SFT2025.1246.7