AnalysisAI ModelsSeptember 30, 2026

TabFM: 400M-parameter zero-shot foundation model for tabular data

Read original source →arxiv.org

TabFM is a 400M-parameter tabular foundation model that casts supervised tabular prediction as in-context learning, replacing per-dataset tree ensembles and AutoML searches. A companion paper targets the cost of conditioning on all training examples in every forward pass.

1 source

More stories today

Open the live feed