Putting Andrew Ng's OCR models to the test

Why LLMs still have problems with OCR