[ainch]

On tabular datasets less than ~250k samples, tabular foundation models now outperform boosting. Of course it remains to be seen how they'll scale to significantly larger datasets as the models improve.

https://huggingface.co/spaces/TabArena/leaderboard