I ran the review recipe with some multiple choice datasets. And I wasn’t able to select any other option. I also tested the recipe with ner_manual annotations, where I was able to manipulate the annotation - so I think not being able to do the same for multiple choice annotations is a bug.
Thanks for the report – I’ll have a look at this. Do you have an example of the multiple choice review examples it created? You can run Prodigy with PRODIGY_LOGGING=verbose and it should output the examples in the stream. (Feel free to remove/replace the actual text – I’m mostly interested in the IDs and format it generated.)