List your company

Multiple-choice factual recall on POLYFACT High

73.15AccuracyGRPO
55.657260.198664.7469.2814Jun 4, 2026

Evaluation results

MethodLinks
GRPOBackbone=Qwen-2.5-7B2026.0673.15
BaselineBackbone=Qwen-2.5-7B2026.0666.69
GRPOBackbone=OLMo-2-1124-7B2026.0664.21
CPT + GRPOBackbone=Qwen-2.5-7B2026.0664.11
SFTBackbone=Qwen-2.5-7B2026.0663.25
CPTBackbone=Qwen-2.5-7B2026.0661.29
CPT + GRPOBackbone=OLMo-2-1124-7B2026.0661.26
CPT + SFTBackbone=OLMo-2-1124-7B2026.0659.2
CPT + SFTBackbone=Qwen-2.5-7B2026.0658.1
BaselineBackbone=OLMo-2-1124-7B2026.0657.93
CPTBackbone=OLMo-2-1124-7B2026.0657.89
SFTBackbone=OLMo-2-1124-7B2026.0656.33