Paper Club 7 - Gaël Varoquaux - Probabl

Are tabular foundation models the next big thing? How do they work?

These are the questions we’ll explore at our 7th Paper Club, featuring Gaël Varoquaux, co-creator of scikit-learn, Chief Science Officer at Probabl, and Research Director at INRIA.

You’ve probably heard of gradient boosting and random forests. But over the past two years, a new family of models has been gaining traction: tabular foundation models (TFMs). Gaël Varoquaux and his team are among the leading research groups in this field, developing state-of-the-art open-source models such as TabICL and, more recently, TabICLv2.

During this one-hour session, he’ll introduce us to this new family of models, explain how they work, walk us through recent research at the cutting edge of the field, and share his perspective on where tabular machine learning is headed.

Want to dive into the topic beforehand? Here are a few recommended reads:

An accessible introduction: https://blog.probabl.ai/demystifying-tfms

The TabICL paper: https://arxiv.org/pdf/2502.05564

The TabICLv2 paper: https://arxiv.org/pdf/2602.11139