I wonder if detecting incompatibilities in rendering docx formatted documents in LibreOffice and Collabora is something that AI models could actually help with. If you had a large enough dataset of various docx (or other Microsoft formatted files) you could have Word and LibreOffice render each of the files and then have the model check the two renderings for differences. If something shows up you could probably use a second model to highlight what elements of the relevant docx file was at fault. That way when something shows up consistently the tool could provide that data to the relevant teams to fix the problem. I think there will always be so many edge cases with problems like this that having humans test at this scale wouldn’t be a good use of engineering time.
I wonder if detecting incompatibilities in rendering docx formatted documents in LibreOffice and Collabora is something that AI models could actually help with. If you had a large enough dataset of various docx (or other Microsoft formatted files) you could have Word and LibreOffice render each of the files and then have the model check the two renderings for differences. If something shows up you could probably use a second model to highlight what elements of the relevant docx file was at fault. That way when something shows up consistently the tool could provide that data to the relevant teams to fix the problem. I think there will always be so many edge cases with problems like this that having humans test at this scale wouldn’t be a good use of engineering time.