The paper, “A11YRepair: Bridging Web Accessibility Barriers via Knowledge-Enhanced Divide-and-Conquer Repair,” introduces an LLM-based framework for automatically repairing web accessibility violations. A11YRepair groups related accessibility issues, decomposes them according to their underlying causes, and selectively incorporates knowledge from the Web Content Accessibility Guidelines (WCAG) to support more focused and consistent repairs.

The work was evaluated using A11YBench, a benchmark comprising 60 real-world web projects, and demonstrates improvements in repair effectiveness and cost compared with existing approaches. The paper is authored by Kai Huang, Ling Zhu, Jian Zhang, Xiaofei Xie, and Chunyang Chen and will be presented at ASE 2026 in Munich. More information about the paper is available here.