Thank you for your question.
When dealing with unsigned documents in Document Intelligence Studio, you should create a label for the 'signed' field even if the document is unsigned, as this helps the model understand that the absence of content is significant. Annotate the empty signature area to ensure the model recognizes it as intended for a signature. Include a diverse training dataset with both signed and unsigned documents to help the model distinguish between the two scenarios. After training, evaluate the model’s performance on a validation set to ensure it accurately identifies empty signature fields. The field name 'dec_sign' likely serves as a placeholder for the signature area.
If you are still seeing similar results, please feel free to submit feedback from DI studio using the smiley icon on the top right corner about your results. Thank you.