Hi,
I am trying to use PubMED QA dataset, when I am using long_answer and try to preprocessing the data
most of end/stop position the answers are taking 0 as position and that is because the long-answer are not necessarily located in the context
any thoughts about how that can be solved
thanks