Google Research
8/10 signal
Introducing TabFM: A zero-shot foundation model for tabular data
contextreasoning
What happened
Google Research introduced TabFM, a zero-shot foundation model designed specifically for tabular classification and regression. Integrated directly into BigQuery ML, TabFM frames tabular prediction as an in-context learning (ICL) problem, allowing it to perform predictions on new datasets without explicit model training, hyperparameter optimization, or manual feature engineering. This challenges traditional supervised tree-based approaches like XGBoost.
Why it matters
It brings the zero-shot, in-context learning paradigm of LLMs to tabular data, potentially replacing tedious manual feature engineering pipelines.
The take
Tabular data has long been the last stronghold of traditional ML (XGBoost, Random Forests). Applying the LLM paradigm of in-context learning directly to tabular data via a foundation model is a massive shift. If TabFM can match tree-based performance without the tedious feature engineering and tuning cycles, it will fundamentally change enterprise data science pipelines.
Do this
If you manage tabular data pipelines in Google Cloud, explore TabFM in BigQuery ML to benchmark its zero-shot performance against your existing XGBoost/Random Forest models.
Don't read this site daily. Get it in your inbox.
The daily brief and Sunday deep dive — distilled, scored, and opinionated. For builders only.