Work

A Benchmark for the Generation and Evaluation of Scientific Architecture Diagrams in HTML/SVG

Year: 2026

Type: conference-paper

Abstract: While Large Language Models (LLMs) can now automate many parts of scientific writing, architectural figure generation faces a critical modality bottleneck because image outputs are inherently difficult to edit, verify, and integrate into text-native LLM workflows.To enable fully text-driven synthesis, we propose a new task: generating architectural figures as pure HTML/SVG source code.Furthermore, we introduce FigureCode, a novel benchmark that evaluates structural fidelity and visual properties through code-based evaluation.Our benchmark improves agreement with human judgments by 0.35 Kendall's τ over prior VLM-based methods.Moreover, the figure-ascode paradigm improves edit-instruction success rate by 30% and retrieval F1 score by 25.3% over baselines, while achieving 66.1% WCAG/ARIA accessibility coverage.

Source: Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)

DOI: 10.18653/2026.findings-emnlp.2060