Text-demo In-context Learning on Demo-ICL-Bench
84Accuracy (Demo)Human
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Human2026.02 | 84 | — | — | |
| Gemini-2.5-Pro2026.02 | 54.4 | — | — | |
| GPT-4o2026.02 | 48.8 | — | — | |
| Qwen2.5-VLSize=72B, Frame=322026.02 | 45 | 24.2 | 20.8 | |
| Demo-ICLSize=7B, Frame=322026.02 | 43.4 | 29.4 | 14 | |
| Demo-ICL (SFT)Size=7B, Frame=322026.02 | 38.4 | 27.8 | 10.6 | |
| VideoChat-R1Size=7B, Frame=322026.02 | 34.4 | 27 | 7.4 | |
| Video-R1Size=7B, Frame=322026.02 | 33.6 | 27.6 | 6 | |
| Qwen2.5-VLSize=7B, Frame=322026.02 | 32.8 | 26 | 6.8 | |
| OlaSize=7B, Frame=322026.02 | 32.2 | 23 | 9.2 | |
| InternVL-3Size=8B, Frame=322026.02 | 31.4 | 26.6 | 4.8 | |
| Ola-Video (Base)Size=7B, Frame=322026.02 | 31.4 | 22.8 | 8.6 | |
| LLaVA-VideoSize=7B, Frame=322026.02 | 31 | 25.4 | 5.6 | |
| Qwen2-VLSize=7B, Frame=322026.02 | 29 | 21.8 | 7.2 |