The corpus used in our ACL 2017 paper is available here: https://github.com/yinfeiy/PICO-data.
Note that we are very much still very much actively collecting more data as part of this corpus -- I will post updates here as we finalize them.