List your company

Mathematical Reasoning on GSM CoT

83.2Accuracy (GSM CoT)χ2-DRTO
64.06469.0327478.968Mar 27, 2026

Evaluation results

MethodLinks
χ2-DRTOBackbone=Llama3-8B2026.0383.2
KL-DRTOBackbone=Llama3-8B2026.0379.6
DPOBackbone=Llama3-8B2026.0376.4
GRPOBackbone=Llama3-8B2026.0376
PPOBackbone=Llama3-8B2026.0375.2
RTOBackbone=Llama3-8B2026.0374.4
SFTBackbone=Llama3-8B2026.0364.8