1. Finding papers / claims / data across an academic literature which is ballooning in size.
2. Using these raw materials to to answer questions in a reliable manner.
#2 is where the bulk of the tricky ML work is, and where vanilla language models often fall short because of limited context windows and hallucination.
We're also working to expand Elicit to help academics with other parts of their research, like surfacing critiques, suggesting related prior work, brainstorming related research questions, identifying risks of bias, …
