Data Field Theory formalizes machine learning as geometry on Riemannian manifolds: data as a field on curved space, learning as movement along that geometry. It is the furthest upstream work in the lab and the base of the foundational model now in development at DFT Labs.
The paper is published open access in Frontiers in Big Data.
Read the full paper at Frontiers (DOI 10.3389/fdata.2026.1752468) →