ROI-level Visual Question Answering on Chaoyang
66AccuracyPathPocket
Evaluation Results
| Method | Links | |
|---|---|---|
| PathPocket2026.06 | 66 | |
| Qwen3VLModel Size=8B2026.06 | 53.2 | |
| Qwen3VLModel Size=4B2026.06 | 35.8 | |
| LLaVA-Med2026.06 | 33.2 | |
| Quilt-LLaVA2026.06 | 30.6 | |
| Qwen3VLModel Size=2B2026.06 | 25.5 |