Running document intelligence to extract Layout but the output given is from a diffeent model (general-document)

Question

Hi, while using the Python SDK to request the analysis of a PDF document from the Document Intellicence service, we are getting different result between docs tested on document intelligence studio end the results coming back from the SDK call.

More details:
when we test the Layout (modelid="prebilt-layout") model to extract tables in the Azure Document Intelligence Studio we get the tables extracted correctly. From the SDK, the tables are extracted with some missing columns, the weird thing is that the result is the same as if the model used is the General Documents (modelid="prebuilt-document"). Is it possible that the Prebuilt-layout model is not yet supported or that there is some parameter I am missing?

Runtime Python3.11
API version used: tested all of them

Answer 1

Encountering a little bit different, but basically the same bug. There is a discrepancy between document intelligence studio output and SDK output on .NET as well. Same API version, same model, same extra features enabled, same image, but confidence score from every single read word is different. I am having values read wrong from SDK and their confidence score in unacceptable levels whereas in document intelligence studio I am having amazing results.

We need an urgent fix for this.