ROI-level Visual Question Answering on SMU-VL 6
79.3AccuracyPathPocket
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PathPocket2026.06 | 79.3 | 76.6 | |
| Qwen3-VL-8BModel Size=8B2026.06 | 75.8 | 73 | |
| Qwen3-VL-4BModel Size=4B2026.06 | 72.9 | 70 | |
| Qwen3-VL-2BModel Size=2B2026.06 | 66.7 | 63.5 | |
| Quilt-LLaVA2026.06 | 24.4 | 21.5 | |
| LLaVA-Med2026.06 | 19.2 | 16.5 |