The labelling doesn't seem to be entirely consistent to me, but I think the idea is that 51 is inviting you to brainstorm, while 68 is asking a question that just happens to be open ended.
Hey! Worked on this here at Databricks: the blog post goes into the dataset collection design a bit (https://www.databricks.com/blog/2023/04/12/dolly-first-open-...). In summary, you're right - brainstorming and GeneralQA will have overlap because the taxonomy naturally has some overlap
