ROI-level Visual Question Answering on SMU-VL 1
77AccuracyPathPocket
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PathPocket2026.06 | 77 | 73.8 | |
| Qwen3-VL-8BModel Size=8B2026.06 | 76.6 | 73.4 | |
| Qwen3-VL-2BModel Size=2B2026.06 | 75.8 | 72.6 | |
| Qwen3-VL-4BModel Size=4B2026.06 | 70.8 | 67.3 | |
| Quilt-LLaVA2026.06 | 25.7 | 22.4 | |
| LLaVA-Med2026.06 | 23.7 | 20.5 |