FitAQA - Temporal Grounding: leaderboard
Metric: Mean temporal IoU (%; localizing the intervals where a described form error occurs; over FitAQA's fitness action-quality videos (30 bodyweight exercises, 38-error taxonomy); temporal grounding task). Source: arxiv.org. Saturation forecast: Around September 2027. 18 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 | 47.5 |
| 2 | Gemini 3.1 Pro (Preview) | 44 |
| 3 | Qwen 3.5 27B (Non-reasoning) | 37.6 |
| 4 | GPT-5.4 | 36.9 |
| 5 | Qwen 3.5 27B (Thinking) | 36.6 |
| 6 | Qwen 3.5 35B A3B (Thinking) | 31.7 |
| 7 | Qwen 3.5 35B A3B (Non-reasoning) | 30.5 |
| 8 | Qwen 3.5 9B (Thinking) | 30.4 |
| 9 | Qwen 3.5 9B (Non-reasoning) | 25.7 |
| 10 | Gemma 4 31B | 24.9 |
| 11 | Qwen 3 VL 8B | 10.1 |
Interactive version: theaggregate.ai/benchmark?slug=fitaqa-temporal-grounding · How It Works · Data refreshed daily, snapshot 2026-09-26.