At a high level, we tried to see if language models can help evaluate the strength of evidence found in randomized controlled trials. We break down the task of identifying placebos and intervention adherence to see if language models (and eventually Elicit) can answer these questions without being too black-box.
We are excited about the progress we made while also wishing we got a lot further. This is what Andreas, our CEO, had to say about the experience. Maybe some of you are trudging through the valley of research despair right now and need to hear it :)