ROI-level Visual Question Answering on SMU-VL 3
77.1AccuracyPathPocket
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PathPocket2026.06 | 77.1 | 74.4 | |
| Qwen3-VL-8BModel Size=8B2026.06 | 76.1 | 73.4 | |
| Qwen3-VL-2BModel Size=2B2026.06 | 74.3 | 71.4 | |
| Qwen3-VL-4BModel Size=4B2026.06 | 69.1 | 66.1 | |
| Quilt-LLaVA2026.06 | 24.8 | 22 | |
| LLaVA-Med2026.06 | 23.3 | 20.5 |