VEBench - Technique Recognition: leaderboard
Metric: Technique Recognition accuracy (%): five-option multiple choice on the editing technique (J-cut, L-cut, cutaway, match, jump, invisible and smash cuts) used in a real edited video, each model at its default frame sampling; higher is better. Source: arxiv.org. Saturation forecast: Around 2030. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro | 34.65 |
| 2 | Qwen 3 VL 32B Instruct | 28.99 |
| 3 | Qwen 3 VL 8B Instruct | 28.64 |
| 4 | Qwen 3 VL 4B Instruct | 28.6 |
| 5 | Qwen 3 VL 30B A3B Instruct | 28.3 |
| 6 | Qwen 2.5 VL 7B Instruct | 27.47 |
| 7 | InternVL3-14B | 27.08 |
| 8 | InternVL3-8B | 25.21 |
| 9 | GPT-4o | 24.68 |
Interactive version: theaggregate.ai/benchmark?slug=vebench-technique-recognition · How It Works · Data refreshed daily, snapshot 2026-10-07.