Clarification on support for multiple harmful categories in a single prompt

[yonglixiang]

Thank you for the excellent work on this project.

There is a question regarding harmful content classification. When a prompt contains multiple types of harmful content, does the model support returning multiple harmful categories, or is it designed to provide only a single category? (From my experiments so far, the model only produces one harmful label per input.)

Thank you for your time and clarification.

[deshpandeneeraj]

+1