The results of the table recognition for the example PDF are incorrect.

[pJahad]

I ran DocumentConverter on the PDF file featured in the example (https://arxiv.org/pdf/2408.09869) with TableFormerMode.ACCURATE and (do_cell_matching = False or do_cell_matching = True). Results of Table 1(item.export_to_html()) : Does this reflect the accuracy of the current model? Or is my usage incorrect?