Reasoning for Mobile User Experience with Multimodal LLMs: Task, Benchmark, and Approach

2026-06-11

AI Summary

Existing approaches struggle to effectively evaluate user experience (UX) based on interface screenshots, and multimodal large language models often lack fine-grained reasoning capabilities for user interfaces. To address this, this work introduces UXBench, a benchmark comprising 2,000 visual question-answering samples and eight real-world UI tasks designed to assess models’ understanding of layout, visual hierarchy, and content consistency. We propose the UI-UX model, which incorporates a reward routing mechanism to dynamically balance perceptual understanding and logical reasoning, along with an asymmetric transition reward to suppress redundant or insufficient reasoning steps. Built upon Qwen3-VL-4B-Thinking and optimized via reinforcement learning, UI-UX achieves an accuracy of 0.7963 on UXBench, significantly outperforming Claude-4.5-Sonnet (0.6550) while demonstrating strong generalization and low inference latency.

Abstract

User experience (UX) centered on usability, perceived consistency, and functional clarity is fundamental to real-world user interfaces (UI). The application of multimodal large language models (MLLMs) in the field of user interfaces is evolving rapidly, such as visual element grounding, graphical user interface (GUI) agents, and design-to-code generation. However, research efforts on evaluating UX based on UI screenshots are still immature. To address this, we propose UXBench, a novel multimodal benchmark consisting of 2,000 VQA data samples designed to assess MLLMs' ability to perform UI-based reasoning. UXBench includes 8 tasks based on real-world UI screenshots that require fine-grained diagnosis of UX issues across layout relationships, visual hierarchy, and content consistency. Our extensive evaluation of mainstream MLLMs shows that they remain fundamentally limited in their capacity for UI-based reasoning. The results underscore the need for further advancements in this area. To bridge this gap, we propose UI-UX, an MLLM based on Qwen3-VL-4B-Thinking foundation model and enhanced via reinforcement learning with two key innovations: a reward routing mechanism that dynamically balances perceptual understanding and logical reasoning during inference, and an asymmetric transition reward that suppresses redundant or insufficient reasoning steps. Experiments demonstrate that UI-UX achieves state-of-the-art (SOTA) performance on UXBench, attaining an accuracy of 0.7963 -- surpassing Claude-4.5-Sonnet's 0.6550 -- while exhibiting strong generalization across diverse UI tasks and maintaining low inference latency.

Problem

• user experience
• multimodal LLMs
• UI reasoning
• UX evaluation
• benchmark

Innovation

• UXBench
• UI-UX
• reward routing mechanism
• asymmetric transition reward
• multimodal LLMs