Blog posts, &etc.

  • All predictions from our trained evidence-inference model (for all trial reports in PubMed) are available in the Trialstreamer database:

  • Here is the Evidence Inference dataset/task website, which allows one to browse and download the collected data. We also provide pointers to starter code to begin working with the data and on the task, which is available in our GitHub repository.

  • Here is a blog post describing the Evidence Inference task, and our initial models.

  • Here is the website for the ERASER benchmark, which includes the Evidence Inference corpus and other datasets for which targets are accompanied by supporting snippets (rationales):

  • PI Wallace spoke about the overarching goal of automating biomedical evidence synthesis on the NLP Highlights podcast: