PACE: Plug-and-Play Contextual Embedding for Feature Screening with Pretrained Tabular Foundation Models
PACE (Plug-and-Play Contextual Embedding) inserts a frozen tabular foundation model (TFM) column encoder before an existing feature-scoring rule, expanding each feature into a higher-dimensional contextual representation, positioning pretrained column geometry as a reusable upstream primitive for tabular learning.
Qi Qin, Er-Bo Li, Ting Wei et al.
· 0 citations