UI2App - Executability (Self-Debug): leaderboard
UI2App (Hong Kong University of Science and Technology, 2026) gives a vision-language model uncaptioned screenshots of a real multi-route web application, 45 sets and 327 screenshots in all, and asks it to generate a runnable React application. This board counts an application as executable when it builds and renders after up to three rounds of feeding build errors back to the model, in percent; higher is better.
Metric: EXEC@3 (%; share of applications that build and render after up to three rounds of feeding the build errors back to the model; 45 screenshot sets (327 screenshots) of real multi-route web applications; the model must emit a runnable React application from the screenshots alone). Source: arxiv.org. Saturation forecast: Estimated already saturated. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 100 | #94 |
| 2 | Claude Sonnet 4.6 | 100 | #161 |
| 3 | Kimi K2.5 (Thinking) | 86.7 | #256 (Kimi K2.5) |
| 4 | GPT-5.4 | 82.2 | #123 |
| 5 | Qwen 3.5 397B A17B | 77.8 | #246 |
| 6 | GLM-4.6V | 35.6 | #701 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=ui2app-executability-self-debug · How It Works · Data refreshed daily, snapshot 2026-11-13.