aiqbench analysis
This research investigates whether tabular foundation models like TabPFN can generalize to non-tabular tasks such as image and language recognition by reformulating inputs as simple row-based tables. The study demonstrates that without explicit spatial or sequential architecture, these models achieve competitive accuracy on MNIST, Tiny ImageNet, and language identification benchmarks using only context samples. This finding suggests a potential paradigm shift where universal tabular representations might simplify the deployment of foundation models across diverse data modalities.
Tabular foundation models for non-tabular tasks
arXiv:2608.22594v1 Announce Type: new Abstract: Tabular foundation models (TFMs) have recently emerged as a promising paradigm for machine learning on tabular data, offering the ability to generalize across datasets without task-specific training. Since many machine learning datasets can be represented as tables, this raises the question: does TFM capability extend beyond tasks traditionally regarded as tabular? We address this question by using TabPFN v3 on three non-tabular classification problems: handwritten digit recognition on MNIST, language identification of French and German words, and image classification on Tiny ImageNet. In each case, the original data are represented as rows of a table and classification is formulated as prediction of a missing label. We evaluate performance as a function of the number of context samples provided to the pretrained model, with no additional training or fine-tuning. Despite having no explicit access to the spatial or sequential structure characterizing the data, TabPFN v3 in some cases achieves accuracies comparable with that of models or methods geared specifically toward the corresponding tasks.
Abstract:Tabular foundation models (TFMs) have recently emerged as a promising paradigm for machine learning on tabular data, offering the ability to generalize across datasets without task-specific training. Since many machine learning datasets can be represented as tables, this raises the question: does TFM capability extend beyond tasks traditionally regarded as tabular? We address this question by using TabPFN v3 on three non-tabular classification problems: handwritten digit recognition on MNIST, language identification of French and German words, and image classification on Tiny ImageNet. In each case, the original data are represented as rows of a table and classification is formulated as prediction of a missing label. We evaluate performance as a function of the number of context samples provided to the pretrained model, with no additional training or fine-tuning. Despite having no explicit access to the spatial or sequential structure characterizing the data, TabPFN v3 in some cases achieves accuracies comparable with that of models or methods geared specifically toward the corresponding tasks.
Machine Learning (cs.LG)
arXiv:2608.22594 [cs.LG]
(or arXiv:2608.22594v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2608.22594 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 256 words
Related coverage
Towards Unified and Data-Efficient Prognostics and Health Management with Tabular Foundation Models
arXiv · Jun 6, 2026
GraphPFN: A Prior-Data Fitted Graph Foundation Model
arXiv · Aug 21, 2026
Tabular Foundation Models for Clinical Survival Analysis via Survival-Aware Adaptation
arXiv · Jun 12, 2026
Build Your Own Transaction Foundation Model for Financial Intelligence
NVIDIA · Jun 16, 2026