VideoVIBE - Workflow Validity: leaderboard
The Workflow Validity subdimension (semantic and logical consistency) of VideoVIBE (2026): video MLLMs diagnose which failure a recorded session of a one-shot generated website shows, from the video and page source. Accuracy in percent; higher is better.
Metric: Failure-diagnosis accuracy (%; questions of the Workflow Validity subdimension of Semantic and Logical Consistency; zero-shot single-choice diagnosis from the interaction video and the page source code (Video+HTML), about 1.7K questions built from 6,338 verified failures of one-shot generated web applications recorded under human operation; free-form answers mapped deterministically to an option). Source: arxiv.org. Saturation forecast: Around February 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Flash Lite | 72.83 | #364 |
| 2 | Gemini 2.5 Flash | 69.57 | #453 |
| 3 | Gemini 3 Flash | 66.3 | #164 |
| 4 | Qwen 3.5 35B A3B | 66.3 | #482 |
| 5 | Gemini 3.1 Pro (Preview) | 58.7 | #94 |
| 6 | GLM-4.6V | 52.17 | #701 |
| 7 | Qwen 3 VL 8B Instruct | 48.91 | #981 |
| 8 | Qwen 3 VL 8B (Thinking) | 48.91 | #887 |
| 9 | Qwen 3 VL 30B A3B (Thinking) | 47.83 | #712 |
| 10 | Qwen 3 VL 30B A3B Instruct | 43.48 | #871 |
| 11 | MiMo-V2.5 | 42.39 | #354 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=videovibe-workflow-validity · How It Works · Data refreshed daily, snapshot 2026-11-13.