Large Scale Transfer Learning for Tabular Data via Language Modeling

TL;DR

This paper introduces TabuLa-8B, a large language model trained on extensive tabular data, demonstrating significant improvements in zero-shot and few-shot prediction accuracy over existing models.

Contribution

It presents a novel dataset extraction process and a fine-tuning scheme for LLMs to excel in tabular data prediction tasks, bridging the gap in transfer learning for tabular data.

Findings

TabuLa-8B achieves over 15 pp higher zero-shot accuracy than existing models.

In few-shot settings, TabuLa-8B outperforms models trained on much larger datasets.

The model and dataset are publicly released for further research.

Abstract

Tabular data -- structured, heterogeneous, spreadsheet-style data with rows and columns -- is widely used in practice across many domains. However, while recent foundation models have reduced the need for developing task-specific datasets and predictors in domains such as language modeling and computer vision, this transfer learning paradigm has not had similar impact in the tabular domain. In this work, we seek to narrow this gap and present TabuLa-8B, a language model for tabular prediction. We define a process for extracting a large, high-quality training dataset from the TabLib corpus, proposing methods for tabular data filtering and quality control. Using the resulting dataset, which comprises over 2.1B rows from over 4M unique tables, we fine-tune a Llama 3-8B large language model (LLM) for tabular data prediction (classification and binned regression) using a novel packing and attention scheme for tabular prediction. Through evaluation across a test suite of 329 datasets, we find that TabuLa-8B has zero-shot accuracy on unseen tables that is over 15 percentage points (pp) higher than random guessing, a feat that is not possible with existing state-of-the-art tabular prediction models (e.g. XGBoost, TabPFN). In the few-shot setting (1-32 shots), without any fine-tuning on the target datasets, TabuLa-8B is 5-15 pp more accurate than XGBoost and TabPFN models that are explicitly trained on equal, or even up to 16x more data. We release our model, code, and data along with the publication of this paper.