Demonstration Selection on Demo-ICL-Bench
76Overall AccuracyHuman
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Human2026.02 | 76 | 88 | |
| Gemini-2.5-Pro2026.02 | 26 | — | |
| GPT-4o2026.02 | 24.5 | — | |
| Demo-ICLSize=7B, Frame=322026.02 | 24 | 58 | |
| Demo-ICL (SFT)Size=7B, Frame=322026.02 | 21.5 | 54.5 | |
| LLaVA-VideoSize=7B, Frame=322026.02 | 20.5 | 44.5 | |
| VideoChat-R1Size=7B, Frame=322026.02 | 18.5 | 52 | |
| Qwen2.5-VLSize=7B, Frame=322026.02 | 18 | 46 | |
| Qwen2.5-VLSize=72B, Frame=322026.02 | 18 | 54 | |
| Ola-Video (Base)Size=7B, Frame=322026.02 | 18 | 48 | |
| Video-R1Size=7B, Frame=322026.02 | 17.5 | 48 | |
| InternVL-3Size=8B, Frame=322026.02 | 16.5 | 44 | |
| OlaSize=7B, Frame=322026.02 | 16 | 43 | |
| Qwen2-VLSize=7B, Frame=322026.02 | 14.5 | 38 |