Visual Grounding on VIEW2SPACE
85.67mIoUHuman-max
Evaluation results
| Method | Links | |
|---|---|---|
| Human-max2026.03 | 85.67 | |
| Human-avg2026.03 | 81.62 | |
| Human-min2026.03 | 77.53 | |
| Ours (Grounded CoT)Model Type=Fine-tuned Model, Backbone=Qwen3VL-4B2026.03 | 69.34 | |
| Qwen3-VL-4B2026.03 | 16.53 | |
| GPT-52026.03 | 8.18 | |
| Random (frequency)2026.03 | 2.01 | |
| Random (chance)2026.03 | 0.22 |