List your company

Mathematical Reasoning on GSM8K (EM %)

63.1EM (%)TUR-DPO
58.52459.71260.962.088Apr 30, 2026

Evaluation results

MethodLinks
TUR-DPOEvaluation Protocol=Human eval2026.0463.1
TUR-DPOEvaluation Protocol=LLM judge2026.0462.8
PPOEvaluation Protocol=Human eval2026.0462
DPOEvaluation Protocol=Human eval2026.0459.2
IPOEvaluation Protocol=LLM judge2026.0458.9
DPOEvaluation Protocol=LLM judge2026.0458.7