Retrieval & Fine-Tuning for In-Context Tabular Models

TL;DR

This paper introduces LoCalPFN, a method combining retrieval and fine-tuning to adapt transformer-based tabular models to local data subsets, significantly improving performance on diverse datasets.

Contribution

The paper proposes a novel retrieval and fine-tuning scheme for transformer-based tabular models, achieving state-of-the-art results on 95 datasets and advancing deep learning for tabular data.

Findings

LoCalPFN outperforms previous models on 95 datasets.

Retrieval and fine-tuning significantly boost model performance.

The approach surpasses tuned tree-based models.

Abstract

Tabular data is a pervasive modality spanning a wide range of domains, and the inherent diversity poses a considerable challenge for deep learning. Recent advancements using transformer-based in-context learning have shown promise on smaller and less complex datasets, but have struggled to scale to larger and more complex ones. To address this limitation, we propose a combination of retrieval and fine-tuning: we can adapt the transformer to a local subset of the data by collecting nearest neighbours, and then perform task-specific fine-tuning with this retrieved set of neighbours in context. Using TabPFN as the base model -- currently the best tabular in-context learner -- and applying our retrieval and fine-tuning scheme on top results in what we call a locally-calibrated PFN, or LoCalPFN. We conduct extensive evaluation on 95 datasets curated by TabZilla from OpenML, upon which we establish a new state-of-the-art with LoCalPFN -- even with respect to tuned tree-based models. Notably, we show a significant boost in performance compared to the base in-context model, demonstrating the efficacy of our approach and advancing the frontier of deep learning in tabular data.