The Year in AI Papers
The year/Independent research

Paper 2605.10616

MulTaBench: Benchmarking Multimodal Tabular Learning with Text and Image

Alan Arazi, Eilam Shapira, Shoham Grunblat, Mor Ventura, Elad Hoffer, Gioia Blayer, David Holzmüller, Lennart Purucker, Gaël Varoquaux, Frank Hutter, Roi Reichart

Published
May 2026
Research lab
Independent
Citations
2
GitHub
Not linked

01 In brief

Summary

MulTaBench is a benchmark of 40 datasets (20 image-tabular, 20 text-tabular) designed for Multimodal Tabular Learning (MMTL).

It addresses the limitation of existing benchmarks that focus on modality co-occurrence, which masks the benefits of task-specific tuning.

The authors propose two criteria for dataset inclusion: Joint Signal (each modality provides complementary predictive information) and Task-awareness (task-agnostic embeddings lose critical information, necessitating Target-Aware Representations, TAR).

A curation pipeline evaluates candidate datasets using 5 tabular learners (LightGBM, CatBoost, TabM, TabPFNv2, TabPFN-2.5) across 4 conditions (Unimodal Structured, Unimodal Unstructured, Joint Frozen, Joint TAR).

TAR is implemented by fine-tuning the last 3 layers of the encoder (e5-v2-small for text, DINO-v3-small for images) with LoRA on the prediction target.

Only 41% of text-tabular and 31% of image-tabular candidates passed the criteria.

Experiments show TAR consistently outperforms frozen embeddings across all learners, encoder scales (small vs.

large), and embedding dimensions (15, 30, 60 PCA components, and no PCA).

The benchmark includes datasets from healthcare, e-commerce, and other domains, and is the largest image-tabular benchmark to date.

The authors argue that current architectures (TFMs, LLMs/VLMs, joint models) are suboptimal for MMTL, and MulTaBench enables the development of true Multimodal Tabular Foundation Models that combine in-context learning with target-aware representations.

02 From the paper

Abstract

Tabular Foundation Models have recently established the state of the art in supervised tabular learning, by leveraging pretraining to learn generalizable representations of numerical and categorical structured data. However, they lack native support for unstructured modalities such as text and image, and rely on frozen, pretrained embeddings to process them. On established Multimodal Tabular Learning benchmarks, we show that tuning the embeddings to the task improves performance. Existing benchmarks, however, often focus on the mere co-occurrence of modalities; this leads to high variance across datasets and masks the benefits of task-specific tuning. To address this gap, we introduce MulTaBench, a benchmark of 40 datasets, split equally between image-tabular and text-tabular tasks. We focus on predictive tasks where the modalities provide complementary predictive signal, and where generic embeddings lose critical information, necessitating Target-Aware Representations that are aligned with the task. Our experimental results demonstrate that the gains from target-aware representation tuning generalize across both text and image modalities, several tabular learners, encoder scales, and embedding dimensions. MulTaBench constitutes the largest image-tabular benchmarking effort to date, spanning high-impact domains such as healthcare and e-commerce. It is designed to enable the research of novel architectures which incorporate joint modeling and target-aware representations, paving the way for the development of novel Multimodal Tabular Foundation Models.