List your company

Multimodal Understanding on RealWorldQA (mean@8 acc %)

71.63Mean@8 AccuracyRAPO_D
66.606867.910969.21570.5191May 10, 2026

Evaluation results

MethodLinks
RAPO_DBase Model=Qwen2-VL-8B Instruct, Temperature=0.62026.0571.63
PAPO_DBase Model=Qwen2-VL-8B Instruct, Temperature=0.62026.0571.37
DAPOBase Model=Qwen2-VL-8B Instruct, Temperature=0.62026.0568.3
Base (Qwen2-VL-8B Instruct)Temperature=0.62026.0567.2
VPPO_DBase Model=Qwen2-VL-8B Instruct, Temperature=0.62026.0566.8