Search papers, labs, and topics across Lattice.
The paper introduces Vision2Web, a new hierarchical benchmark for evaluating coding agents on visual website development, encompassing UI-to-code generation, frontend reproduction, and full-stack development. It features 193 tasks across 16 categories derived from real-world websites, along with a workflow-based agent verification paradigm using a GUI agent verifier and a VLM-based judge. Experiments with various VLMs show significant performance gaps, particularly in full-stack development, highlighting the challenges in end-to-end website creation.
Today's best visual coding agents still struggle to build complete websites from scratch, as revealed by a new benchmark spanning UI generation to full-stack development.
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.
Tsinghua AI