FineDiving-QA (Video-STaR) - Accuracy: leaderboard
Metric: Correct-answer rate (%) on held-out FineDiving videos converted to QA, including numerical execution-score questions; GPT-3.5-Turbo evaluates answers against reference labels following the Video-ChatGPT protocol; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Video-STaR | 20.2 | |
| 2 | Video-LLaVA-Plus (Video-STaR paper) | 19.1 | |
| 3 | Video-LLaVA | 17.6 | |
| 4 | Vid-LLaVA-Gemini (Video-STaR paper) | 16.3 |
Interactive version: theaggregate.ai/benchmark?slug=finediving-qa-video-star-accuracy · How It Works · Data refreshed daily, snapshot 2026-10-12.