ROI-level Visual Question Answering on SMU-VL 2
75.9AccuracyPathPocket
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PathPocket2026.06 | 75.9 | 0.728 | |
| Qwen3-VL-8BModel Size=8B2026.06 | 74.9 | 0.718 | |
| Qwen3-VL-2BModel Size=2B2026.06 | 73 | 0.698 | |
| Qwen3-VL-4BModel Size=4B2026.06 | 68.3 | 0.649 | |
| Quilt-LLaVA2026.06 | 22.9 | 0.199 | |
| LLaVA-Med2026.06 | 20.1 | 0.172 |