Visual Grounding on CrowdHuman (val)
72.8AP50ChatRex-7B + LGSCs
Evaluation results
| Method | Links | |
|---|---|---|
| ChatRex-7B + LGSCsSetting=Proposed method2026.04 | 72.8 | |
| ChatRex-7BSetting=Naive finetuning2026.04 | 67.4 | |
| ChatRex-7BSetting=Baseline2026.04 | 52 | |
| Groma-7B2026.04 | 32.1 | |
| Qwen2.5-VL-7B2026.04 | 15.3 | |
| InternVL3-8B2026.04 | 9.9 | |
| Ferret-7B2026.04 | 7.2 |