Multimodal Understanding on RealWorldQA (mean@8 acc %)
71.63Mean@8 AccuracyRAPO_D
Evaluation Results
| Method | Links | |
|---|---|---|
| RAPO_DBase Model=Qwen2-VL-8B Instruct, Temperature=0.62026.05 | 71.63 | |
| PAPO_DBase Model=Qwen2-VL-8B Instruct, Temperature=0.62026.05 | 71.37 | |
| DAPOBase Model=Qwen2-VL-8B Instruct, Temperature=0.62026.05 | 68.3 | |
| Base (Qwen2-VL-8B Instruct)Temperature=0.62026.05 | 67.2 | |
| VPPO_DBase Model=Qwen2-VL-8B Instruct, Temperature=0.62026.05 | 66.8 |