ROI-level Visual Question Answering on SMU-VL 5
75.3AccuracyQwen3-VL-8B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3-VL-8BModel Size=8B2026.06 | 75.3 | 72.5 | |
| PathPocket2026.06 | 74.9 | 72 | |
| Qwen3-VL-4BModel Size=4B2026.06 | 70.1 | 67.1 | |
| Qwen3-VL-2BModel Size=2B2026.06 | 69.9 | 66.9 | |
| Quilt-LLaVA2026.06 | 24.2 | 21.4 | |
| LLaVA-Med2026.06 | 19.5 | 16.9 |