Tag
LongWebBench is a benchmark for evaluating long-horizon webpage generation from both structural and functional perspectives, using VLM-based metrics and DOM-augmented agent-based pipelines. Experiments show current VLMs struggle with long-range coherence and executable interactions.