ROI-level Visual Question Answering on SMU-VL 4
73.1AccuracyPathPocket
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PathPocket2026.06 | 73.1 | 70.1 | |
| Qwen3-VL-8BModel Size=8B2026.06 | 71.9 | 68.9 | |
| Qwen3-VL-4BModel Size=4B2026.06 | 69.4 | 66.3 | |
| Qwen3-VL-2BModel Size=2B2026.06 | 68.5 | 65.4 | |
| LLaVA-Med2026.06 | 22.1 | 19.3 | |
| Quilt-LLaVA2026.06 | 20.4 | 17.7 |