Google unveils TabFM, a zero-shot foundation model for tabular data
burny_tech · x · 2026-09-19
Following TimesFM for time series, Google Research introduced TabFM, a zero-shot foundation model for tabular data aimed at simplifying classification and regression workflows.
- Context: Tabular data powers enterprise ML like churn prediction and fraud detection, long dominated by supervised tree models (XGBoost, AdaBoost, random forests).
- Pain point: Deploying these models requires far more than one .fit() — extensive hyperparameter tuning and domain-specific feature engineering.
- Approach: Borrowing in-context learning from LLMs, the pretrained model performs new tasks zero-shot without weight updates.
Weights and code are on Hugging Face and GitHub; commenters call it "intelligence in a spreadsheet."
More from Research
- Amazon: RAG must move beyond factual grounding to surface diverse opinions — _reachsumit · 2026-11-04
- ShiftRAG bypasses RAG's textual bottleneck by feeding retrieval signals as soft tokens — _reachsumit · 2026-11-04
- Meta's Atomic Creator Tokens beat Semantic IDs: +19.8pt Recall@15 and 2.7x lower latency — _reachsumit · 2026-11-04
- HiGenTrans transfers LLM knowledge into generative recommenders at input, representation and decision levels — _reachsumit · 2026-11-04
- SA-OPSD distills self-teacher knowledge at beam-search pruning boundaries for generative recommendation — _reachsumit · 2026-11-04
- MV-Thinker: Multi-View Thinking with Adaptive Generation for Generative Recommendation — _reachsumit · 2026-11-04