A JEPA Recipe for Tabular Foundation Models
Tabular foundation models learn to predict cell values in context, whereas world-model self-supervision asks for prediction in representation space (LeCun, 2022; Assran et al., 2023). On a tabular foundation-model prior, the latent term of a joint-embedding predictive architecture (JEPA) collapsed in our earlier runs a...
Mingyu Jeon, Suwan Cho, Jae-Young Suh
· 0 citations