Generalization Can Emerge in Tabular Foundation Models From a Single Table
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866909899736219648 |
|---|---|
| author | Ma, Junwei Shaheen, Nour Labach, Alex Mhedhbi, Amine Hutter, Frank Caterini, Anthony L. Thomas, Valentin |
| author_facet | Ma, Junwei Shaheen, Nour Labach, Alex Mhedhbi, Amine Hutter, Frank Caterini, Anthony L. Thomas, Valentin |
| contents | Deep tabular modelling increasingly relies on in-context learning where, during inference, a model receives a set of $(x,y)$ pairs as context and predicts labels for new inputs without weight updates. We challenge the prevailing view that broad generalization here requires pre-training on large synthetic corpora (e.g., TabPFN priors) or a large collection of real data (e.g., TabDPT training datasets), discovering that a relatively small amount of data suffices for generalization. We find that simple self-supervised pre-training on just a \emph{single} real table can produce surprisingly strong transfer across heterogeneous benchmarks. By systematically pre-training and evaluating on many diverse datasets, we analyze what aspects of the data are most important for building a Tabular Foundation Model (TFM) generalizing across domains. We then connect this to the pre-training procedure shared by most TFMs and show that the number and quality of \emph{tasks} one can construct from a dataset is key to downstream performance. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_09665 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Generalization Can Emerge in Tabular Foundation Models From a Single Table Ma, Junwei Shaheen, Nour Labach, Alex Mhedhbi, Amine Hutter, Frank Caterini, Anthony L. Thomas, Valentin Machine Learning Deep tabular modelling increasingly relies on in-context learning where, during inference, a model receives a set of $(x,y)$ pairs as context and predicts labels for new inputs without weight updates. We challenge the prevailing view that broad generalization here requires pre-training on large synthetic corpora (e.g., TabPFN priors) or a large collection of real data (e.g., TabDPT training datasets), discovering that a relatively small amount of data suffices for generalization. We find that simple self-supervised pre-training on just a \emph{single} real table can produce surprisingly strong transfer across heterogeneous benchmarks. By systematically pre-training and evaluating on many diverse datasets, we analyze what aspects of the data are most important for building a Tabular Foundation Model (TFM) generalizing across domains. We then connect this to the pre-training procedure shared by most TFMs and show that the number and quality of \emph{tasks} one can construct from a dataset is key to downstream performance. |
| title | Generalization Can Emerge in Tabular Foundation Models From a Single Table |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2511.09665 |