TabFM: A zero-shot foundation model for tabular data
research.google
June 30, 2026
4 min read
🔥🔥🔥🔥🔥
49/100
Summary
TabFM is a new foundation model designed for tabular data, simplifying classification and regression workflows. It utilizes a "zero-shot" logic approach, similar to that of TimesFM, to enhance the handling of enterprise data infrastructure.
Key Takeaways
- TabFM is a new foundation model designed for tabular data classification and regression, utilizing zero-shot prediction to simplify workflows.
- The model eliminates the need for manual training, hyperparameter tuning, and complex feature engineering by framing tabular prediction as an in-context learning problem.
- TabFM is trained entirely on hundreds of millions of synthetic datasets, addressing the scarcity of high-quality, diverse tabular datasets for pre-training.
- The model processes tabular data as a unified prompt, allowing it to learn relationships between columns and rows directly at inference time.
