QA performance by Gemini-2.5-Pro based on captions on World-Sense (test)
34World-Sense QA ScoreASID-Captioner
Evaluation Results
| Method | Links | |
|---|---|---|
| ASID-CaptionerSize=7B2026.02 | 34 | |
| Gemini-2.5-ProSize=not specified2026.02 | 33.8 | |
| ASID-CaptionerSize=3B2026.02 | 32.3 | |
| Gemini-2.5-FlashSize=not specified2026.02 | 31 | |
| AVoCaDOSize=7B2026.02 | 25.7 | |
| video-SALMONN-2Size=7B2026.02 | 18.2 | |
| Qwen3-Omni-CaptionerSize=30B-A3B2026.02 | 14.1 | |
| Qwen3-Omni-InstructSize=30B-A3B2026.02 | 12.7 | |
| UGC-VideoCaptionerSize=3B2026.02 | 11.2 | |
| ARC-Hunyuan-VideoSize=7B2026.02 | 8.7 | |
| Qwen2.5-OmniSize=7B2026.02 | 8.6 | |
| MiniCPM-o-2.6Size=8B2026.02 | 7.2 | |
| HumanOmniV2Size=7B2026.02 | 6.6 |