Counting on SuperClevr Counting
91.38Avg@8 AccuracyPAPO_D
Evaluation results
| Method | Links | |
|---|---|---|
| PAPO_DBackbone=Qwen2.5-VL, Model Size=7B2025.07 | 91.38 | |
| DAPOBackbone=Qwen2.5-VL, Model Size=7B2025.07 | 90.05 | |
| PAPO_GBackbone=Qwen2.5-VL, Model Size=7B2025.07 | 89.81 | |
| PAPO_DBackbone=Qwen2.5-VL, Model Size=3B2025.07 | 83.56 | |
| PAPO_GBackbone=Qwen3-VL (thinking), Model Size=2B2025.07 | 80.63 | |
| GRPOBackbone=Qwen3-VL (thinking), Model Size=2B2025.07 | 80.13 | |
| DAPOBackbone=Qwen2.5-VL, Model Size=3B2025.07 | 74.25 | |
| GRPOBackbone=Qwen2.5-VL, Model Size=7B2025.07 | 73.94 | |
| PAPO_GBackbone=Qwen2.5-VL, Model Size=3B2025.07 | 62.56 | |
| GRPOBackbone=Qwen2.5-VL, Model Size=3B2025.07 | 55.81 |