Egocentric Deictic Visual Grounding on EgoPoint-Ground Hybrid
97.9Precision@0.3Human Study
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Human Study2026.03 | 97.9 | 91.4 | 84.6 | |
| SV-CoTBase Model=Qwen3-vl2026.03 | 79.5 | 75.6 | 71.8 | |
| Qwen3-VL-4B2026.03 | 74.3 | 70.2 | 62.2 | |
| Qwen2.5-VL-3B2026.03 | 73.1 | 67.1 | 59.2 | |
| Qwen3-VL-8B2026.03 | 72.6 | 67.6 | 59.8 | |
| Qwen2.5-VL-7B2026.03 | 72.1 | 65.3 | 56.3 | |
| LLaVA-OneVision-4B2026.03 | 67.9 | 47.5 | 14.5 | |
| InternVL-3.5-8B2026.03 | 55.4 | 41.3 | 27 | |
| GroundingGPT2026.03 | 44.7 | 35.8 | 21.9 | |
| DeepSeek-VL22026.03 | 34.4 | 29.3 | 25.9 | |
| LLaVA-v1.5-13B2026.03 | 33.8 | 19.8 | 7.9 | |
| Ferret-7B2026.03 | 28.2 | 22.1 | 18.1 | |
| Florence2-Large2026.03 | 24.1 | 18.8 | 14.9 | |
| Florence2-Base2026.03 | 14.9 | 11 | 8 |