Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TabTune: A Unified Library for Inference and Fine-Tuning Tabular Foundation Models

Aditya Tanna, Pratinav Seth, Mohamed Bouadi, Utsav Avaiya, Vinay Kumar Sankarapu

TL;DR

TabTune introduces a unified, extensible library that standardizes preprocessing, adaptation strategies, and evaluation for tabular foundation models. By supporting seven TFMs and multiple tuning paradigms (zero-shot, meta-learning, SFT, PEFT), it enables consistent benchmarking and deployment-oriented assessment of accuracy, calibration, and fairness. Comprehensive experiments across TALENT, OpenML-CC18, and TabZilla reveal how model–strategy pairings influence performance and reliability, offering domain-specific guidance for medical and financial settings. The work emphasizes practical impact: practitioners can rapidly compare configurations, ensure trustworthy predictions, and move TFMs toward real-world, responsible deployment in structured data domains.

Abstract

Tabular foundation models represent a growing paradigm in structured data learning, extending the benefits of large-scale pretraining to tabular domains. However, their adoption remains limited due to heterogeneous preprocessing pipelines, fragmented APIs, inconsistent fine-tuning procedures, and the absence of standardized evaluation for deployment-oriented metrics such as calibration and fairness. We present TabTune, a unified library that standardizes the complete workflow for tabular foundation models through a single interface. TabTune provides consistent access to seven state-of-the-art models supporting multiple adaptation strategies, including zero-shot inference, meta-learning, supervised fine-tuning (SFT), and parameter-efficient fine-tuning (PEFT). The framework automates model-aware preprocessing, manages architectural heterogeneity internally, and integrates evaluation modules for performance, calibration, and fairness. Designed for extensibility and reproducibility, TabTune enables consistent benchmarking of adaptation strategies of tabular foundation models.

TabTune: A Unified Library for Inference and Fine-Tuning Tabular Foundation Models

TL;DR

TabTune introduces a unified, extensible library that standardizes preprocessing, adaptation strategies, and evaluation for tabular foundation models. By supporting seven TFMs and multiple tuning paradigms (zero-shot, meta-learning, SFT, PEFT), it enables consistent benchmarking and deployment-oriented assessment of accuracy, calibration, and fairness. Comprehensive experiments across TALENT, OpenML-CC18, and TabZilla reveal how model–strategy pairings influence performance and reliability, offering domain-specific guidance for medical and financial settings. The work emphasizes practical impact: practitioners can rapidly compare configurations, ensure trustworthy predictions, and move TFMs toward real-world, responsible deployment in structured data domains.

Abstract

Tabular foundation models represent a growing paradigm in structured data learning, extending the benefits of large-scale pretraining to tabular domains. However, their adoption remains limited due to heterogeneous preprocessing pipelines, fragmented APIs, inconsistent fine-tuning procedures, and the absence of standardized evaluation for deployment-oriented metrics such as calibration and fairness. We present TabTune, a unified library that standardizes the complete workflow for tabular foundation models through a single interface. TabTune provides consistent access to seven state-of-the-art models supporting multiple adaptation strategies, including zero-shot inference, meta-learning, supervised fine-tuning (SFT), and parameter-efficient fine-tuning (PEFT). The framework automates model-aware preprocessing, manages architectural heterogeneity internally, and integrates evaluation modules for performance, calibration, and fairness. Designed for extensibility and reproducibility, TabTune enables consistent benchmarking of adaptation strategies of tabular foundation models.

Paper Structure

This paper contains 65 sections, 1 equation, 2 figures, 13 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Traditional Machine Learning and Deep Learning for Tabular Data
  4. Tabular Foundation Models
  5. Representation Learning and Alternative Approaches
  6. Tabular ML Toolkits and Frameworks
  7. Adaptation Techniques: Meta-Learning, Fine-Tuning, and PEFT
  8. Calibration, Fairness, and Responsible Deployment
  9. System Design
  10. Design Principles
  11. Modular Architecture
  12. Supported Models
  13. Implementation
  14. Model-Aware Preprocessing
  15. Flexible Tuning Strategies
  16. ...and 50 more sections

Figures (2)

  • Figure 1: Overview of the TabTune architecture. Raw tabular data are processed through a model-aware pre-processing engine and modular tuning pipeline, coordinated by a central orchestration engine with integrated evaluation assessment modules.
  • Figure 2: Modular architecture of TabTune. The TabularPipeline orchestrates the workflow by chaining the model-aware DataProcessor, the encapsulated TFM, and the adaptive TuningManager, forming a cohesive and control framework.