Referring Expression Comprehension on D-REC
95.2Precision @ IoU 0.3Human Study
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Human Study2026.03 | 95.2 | 92.6 | 81.9 | |
| SV-CoTBackbone=Qwen3-vl2026.03 | 81.4 | 75 | 69.4 | |
| Qwen3-VL-4BParameters=4B2026.03 | 73.7 | 68.4 | 60.5 | |
| Qwen2.5-VL-7BParameters=7B2026.03 | 70.3 | 63.8 | 54.1 | |
| Qwen3-VL-8BParameters=8B2026.03 | 70.1 | 64.8 | 56.1 | |
| Qwen2.5-VL-3BParameters=3B2026.03 | 66.5 | 59.2 | 49.1 | |
| LLaVA-OneVision-4BParameters=4B2026.03 | 58.6 | 34.7 | 8.8 | |
| InternVL-3.5-4BParameters=4B2026.03 | 53.5 | 40.2 | 27.2 | |
| LLaVA-OneVision-8BParameters=8B2026.03 | 50.5 | 25.8 | 4.9 | |
| InternVL-3.5-8BParameters=8B2026.03 | 45.6 | 32.5 | 21 | |
| DeepSeek-VL2-SmallSize=Small2026.03 | 44.3 | 39.3 | 34.1 | |
| Ferret-7BParameters=7B2026.03 | 42.8 | 35 | 26.1 | |
| GroundingGPT2026.03 | 37.8 | 26.1 | 12.9 |