CAVSR: leaderboard
CAVSR (Shandong University and collaborators, 2026) is a benchmark of 605 real AI-generated short videos from TikTok, RedNote and YouTube that child-education experts annotated for risks to children. This board scores multi-label recognition over 26 fine-grained risk labels as Micro-F1 in percent, higher is better.
Metric: Micro-F1 (%; multi-label child-risk recognition over 26 fine-grained risk labels on 605 expert-annotated AI-generated videos from TikTok, RedNote and YouTube; the model reviews each video directly). Source: arxiv.org. Saturation forecast: Around February 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 50.12 | #164 |
| 2 | GPT-4o | 38.98 | #741 |
| 3 | Qwen 2.5 VL 7B Instruct | 12.46 | #1833 |
| 4 | Qwen 2 VL 7B | 10.28 | #2261 |
| 5 | Qwen 2 VL 2B | 5.46 | #6284 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=cavsr · How It Works · Data refreshed daily, snapshot 2026-11-13.