GOTabPFN: From Feature Ordering to Compact Tokenization for Tabular Foundation Models on High-Dimensional Data

In three linesGOTabPFN introduces a feature compression method for tabular foundation models in high-dimensional, low-sample-size regimes. Graph-guided Ordering with Local Refinement (GO-LR) orders features, then Neuro-Inspired Subunit Compression pools them into meta-features. Results: improved stability and accuracy under tight token budgets on tabular benchmarks.

Summary generated by Claude — human-verified