Problem
Research questions and friction points this paper is trying to address.
website development
visual-to-code
agent evaluation
full-stack development
benchmark
Vision2Web: A Hierarchical Benchmark for Visual Website Development with Agent Verification
🤖 AI Summary
This work addresses the lack of systematic evaluation benchmarks for end-to-end vision-driven web development, particularly in complex full-stack scenarios. To bridge this gap, we propose Vision2Web—the first hierarchical benchmark for visual website generation—structured across three tiers: static UI generation, interactive multi-page frontend replication, and long-horizon full-stack development. Built on real-world website data, Vision2Web comprises 193 tasks, 918 prototype images, and 1,255 test cases. It introduces a novel dual-component evaluation paradigm that combines a GUI agent validator with a vision-language model (VLM) judge to reliably assess both functional correctness and semantic fidelity of generated outputs. Experimental results reveal significant performance bottlenecks in current models when tackling full-stack development tasks.
📝 Abstract
Recent advances in large language models have improved the capabilities of coding agents, yet systematic evaluation of complex, end-to-end website development remains limited. To address this gap, we introduce Vision2Web, a hierarchical benchmark for visual website development, spanning from static UI-to-code generation, interactive multi-page frontend reproduction, to long-horizon full-stack website development. The benchmark is constructed from real-world websites and comprises a total of 193 tasks across 16 categories, with 918 prototype images and 1,255 test cases. To support flexible, thorough and reliable evaluation, we propose workflow-based agent verification paradigm based on two complementary components: a GUI agent verifier and a VLM-based judge. We evaluate multiple visual language models instantiated under different coding-agent frameworks, revealing substantial performance gaps at all task levels, with state-of-the-art models still struggling on full-stack development.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Vision2Web
hierarchical benchmark
agent verification
visual website development
VLM-based judge
🔎 Similar Papers
No similar papers found.
