UI-UX: UX Defect Diagnosis with Multimodal LLMs

Upload a mobile UI screenshot and select a diagnostic task. The model will analyze the screenshot and identify potential UX defects with chain-of-thought reasoning.

Examples — click to load

UI Screenshot	Diagnostic Task	Additional Content (required for Mismatch tasks)

Supported Tasks

Dimension	Task	What it checks
Usability	PopupNoClose	Modal popup without close control
Usability	BubbleOcclT	Text overlay occluding page text
Usability	BubbleOcclBtn	Text overlay occluding buttons
Efficiency	PopupBlock	Popup blocking mini-program close
Efficiency	PopupStack	Multiple modal dialogs stacked
Trustworthiness	MismatchBadge	Badge/functionality inconsistency
Trustworthiness	MismatchContent	Service description mismatch
Trustworthiness	MismatchFunc	Ad/landing page inconsistency

Model: UI-UX-4B (Qwen3.5-4B + GRPO) | Benchmark: 79.63% on UXBench | Paper: CVPR 2026 Findings