List your company

Multiple-choice factual recall on POLYFACT Low

56.93AccuracyGRPO
47.15449.69252.2354.768Jun 4, 2026

Evaluation results

MethodLinks
GRPOBackbone=Qwen-2.5-7B2026.0656.93
GRPOBackbone=OLMo-2-1124-7B2026.0654.48
CPT + GRPOBackbone=OLMo-2-1124-7B2026.0654.41
BaselineBackbone=Qwen-2.5-7B2026.0652.26
CPT + SFTBackbone=OLMo-2-1124-7B2026.0652.02
CPTBackbone=OLMo-2-1124-7B2026.0651.82
BaselineBackbone=OLMo-2-1124-7B2026.0651.8
CPTBackbone=Qwen-2.5-7B2026.0651.09
SFTBackbone=OLMo-2-1124-7B2026.0650.04
CPT + GRPOBackbone=Qwen-2.5-7B2026.0649.33
SFTBackbone=Qwen-2.5-7B2026.0648.88
CPT + SFTBackbone=Qwen-2.5-7B2026.0647.53