possible datasets

[Enter-tainer]

• https://huggingface.co/datasets/yuntian-deng/im2latex-100k originally used to train ocr models
• full arxiv: https://info.arxiv.org/help/bulk_data_s3.html
• we can also find similar things like oiwiki, where we can extract equations from markdown files

[Enter-tainer]

We may also generate synthesised data using a) random things, like sqlsmith b) large language models. But they might not make a lot of sense

[Enter-tainer]

also

• https://github.com/mathjax/MathJax-Tests
• https://github.com/KaTeX/KaTeX/blob/main/test/katex-spec.js