We're excited to share that our paper on the approach to building the System Graph, “Queryable Evidence from Scientific Papers at Scale: A Statistical Graph and API for Health Research,” has been accepted to the GenAI4Health Workshop at NeurIPS 2026.

Scientific knowledge is immensely valuable, but is contained in discrete papers that are individually retrieved, read, and compared. We demonstrate a pipeline that does that at scale, turning extracted statistical findings from scientific literature into structured evidence that is linked back to its source. Instead of reading papers one-by-one, researchers can directly query findings, compare evidence across studies, surface unstudied pairs of concepts for hypothesis generation, and assemble the system of evidence around a problem.

This capability is publicly accessible for research and non-commercial use as the System Public Graph. 

We will share the full paper in the coming weeks, and look forward to presenting in Sydney this December.