microsoft

/

Phi-3-vision-128k-instruct

like 741

First thing first, thanks for this amazing model.

I am encountering some weird resistance when asking to extract text in markdown format, maybe due to some overfitting. It generally says "The image provided does not contain a table. <provide to summarize what's written> Since there is no table present, I cannot convert any content into markdown format as a table. However, I can provide the text in markdown format as a list or paragraph if that would be helpful."

Then I say: yes, please proceed.

And it generates the markdown just fine.

Any ideas on how to get the answer at the first try?