Video Referring Description on VideRefer-Bench-D Single-Frame
2.35SCElysium-7B
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Elysium-7Bmode=Zero-shot, frame_protocol=single-frame, max_num_tiles=162025.10 | 2.35 | 0.3 | 0.02 | 3.59 | 1.57 | |
| LLaVA-OneVison-7Bmode=Zero-shot, frame_protocol=single-frame, max_num_tiles=162025.10 | 2.62 | 1.58 | 2.19 | 2.07 | 2.12 | |
| Qwen2-VLSize=7B2025.12 | 2.97 | 2.24 | 2.03 | 2.31 | 2.39 | |
| Qwen2-VL-7Bmode=Zero-shot, frame_protocol=single-frame, max_num_tiles=162025.10 | 2.97 | 2.24 | 2.03 | 2.31 | 2.39 | |
| FerretSize=7B2025.12 | 3.08 | 2.01 | 1.54 | 2.14 | 2.19 | |
| Ferret-7Bmode=Zero-shot, frame_protocol=single-frame, max_num_tiles=162025.10 | 3.08 | 2.01 | 1.54 | 2.14 | 2.18 | |
| OspreySize=7B2025.12 | 3.19 | 2.16 | 1.54 | 2.45 | 2.34 | |
| Osprey-7Bmode=Zero-shot, frame_protocol=single-frame, max_num_tiles=162025.10 | 3.19 | 2.16 | 1.54 | 2.45 | 2.34 | |
| GPT-4o2025.12 | 3.34 | 2.96 | 3.01 | 2.5 | 2.95 | |
| GPT-4omode=Zero-shot, frame_protocol=single-frame, max_num_tiles=162025.10 | 3.34 | 2.96 | 3.01 | 2.5 | 2.95 | |
| InternVL2Size=26B2025.12 | 3.55 | 2.99 | 2.57 | 2.25 | 2.84 | |
| InternVL2-26Bmode=Zero-shot, frame_protocol=single-frame, max_num_tiles=162025.10 | 3.55 | 2.99 | 2.57 | 2.25 | 2.84 | |
| VideoReferSize=7B2025.12 | 4.41 | 3.27 | 3.03 | 2.97 | 3.42 | |
| GAR-1Bmode=Zero-shot, frame_protocol=single-frame, max_num_tiles=162025.10 | 4.41 | 2.98 | 1.09 | 2.4 | 2.72 | |
| GAR-8Bmode=Zero-shot, frame_protocol=single-frame, max_num_tiles=162025.10 | 4.41 | 2.96 | 1.58 | 2.45 | 2.75 | |
| UniPixelSize=7B2025.12 | 4.45 | 3.32 | 3.05 | 3.04 | 3.47 | |
| UFVideoSize=7B2025.12 | 4.53 | 3.48 | 3.26 | 3.09 | 3.59 |