🤖 AI Summary
This work addresses the poor accessibility and lack of semantic structure in math-intensive academic papers on arXiv when rendered as PDFs by proposing an automated method to convert TeX/LaTeX source code into high-quality HTML5. The key innovation lies in the first large-scale integration of MathML 4 Intent annotations within an academic publishing platform, substantially enhancing both accessibility and semantic expressiveness of mathematical content. The system employs a Rust-based rewrite of the conversion engine, combined with LaTeXML, a community-driven error feedback mechanism, and an automated pipeline. This approach achieves a 75% error-free conversion rate (with a target of 90%), resolves approximately half of user-reported issues, and reduces computational overhead while accelerating preview generation.
📝 Abstract
We report on the ongoing development of arXiv's HTML Papers offering, available on every new TeX/LaTeX submission since its initial release in 2023.
The main highlights from 2025 and early 2026 are:
(i) community-driven improvements to HTML fidelity and service health, with roughly half of 6,000 user reports resolved;
(ii) corpus-scale conversion work aimed at 90% error-free HTML (currently 75%);
(iii) initial MathML 4 Intent annotations for accessible speech output;
(iv) an in-progress Rust port of LaTeXML, reducing compute costs and enabling faster previews on submission.
The arXiv HTML Papers project remains experimental, but is gradually maturing as we better understand the needs of arXiv's readers and the technical opportunities presented by new standards and by advances in programming languages and AI.