VideoInstruct (Correctness of Information): leaderboard
VideoInstruct benchmark (Video-ChatGPT) — judge-scored video instruction following; correctness-of-information dimension.
Metric: Correctness of Information. Source: huggingface.co. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | PPLLaVA-7B-dpo | 3.85 | |
| 2 | VLM-RLAIF | 3.63 | |
| 3 | PLLaVA-34B | 3.6 | |
| 4 | IG-VLM-GPT4v | 3.4 | |
| 5 | VideoChat2_HD_mistral | 3.4 | |
| 6 | PPLLaVA-7B | 3.32 | |
| 7 | VideoGPT+ | 3.27 | |
| 8 | ST-LLM-7B | 3.23 | |
| 9 | CAT-7B | 3.08 | |
| 10 | LLaMA-VID-13B (2 Token) | 3.07 |
Interactive version: theaggregate.ai/benchmark?slug=videoinstruct-correctness-of-information · How It Works · Data refreshed daily, snapshot 2026-10-20.