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

2025-11-04

AI Summary

Current tabular foundation models suffer from deployment-oriented challenges—including heterogeneous preprocessing requirements, fragmented APIs, inconsistent fine-tuning protocols, and the absence of standardized evaluation metrics for calibration and fairness—severely hindering real-world adoption. To address these issues, we propose TabFoundation, the first unified framework for tabular foundation models. It integrates model-aware automatic preprocessing, architecture-decoupled design, and a modular evaluation system supporting zero-shot inference, meta-learning, full-parameter fine-tuning, and diverse parameter-efficient fine-tuning (PEFT) strategies. Crucially, TabFoundation introduces the first cross-model consistent evaluation protocol jointly assessing performance, calibration, and fairness. This significantly improves experimental reproducibility and deployment efficiency. By unifying preprocessing, training, and evaluation standards, TabFoundation advances tabular foundation models toward standardization and industrial scalability.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Humans and AI: Other Foundations of Human Computation & AI
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

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.

Problem

• Standardizing heterogeneous preprocessing pipelines for tabular foundation models
• Unifying fragmented APIs and inconsistent fine-tuning procedures
• Addressing lack of standardized evaluation for calibration and fairness metrics

Innovation

• Unified library standardizes tabular foundation model workflows
• Automates preprocessing and manages architectural heterogeneity internally
• Supports multiple adaptation strategies with integrated evaluation modules