We introduce a framework for generating highly multimodal datasets with explicitly calculable mutual information (MI) between modalities. This enables the construction of benchmark datasets that provide a novel testbed for systematic studies of mutual information estimators and multimodal self-supervised learning (SSL) techniques. Our framework constructs realistic datasets with known MI using a flow-based generative model and a structured causal framework for generating correlated latent variables. We benchmark a suite of MI estimators on datasets with varying ground truth MI values and verify that regression performance improves as the MI increases between input modalities and the target value. Finally, we describe how our framework can be applied to contexts including multi-detector astrophysics and SSL studies in the highly multimodal regime.
@article{arxiv.2510.21686,
title = {Multimodal Datasets with Controllable Mutual Information},
author = {Raheem Karim Hashmani and Garrett W. Merz and Helen Qu and Mariel Pettee and Kyle Cranmer},
journal= {arXiv preprint arXiv:2510.21686},
year = {2026}
}
Comments
16 pages, 7 figures, 2 tables. Our code is publicly available at https://github.com/RKHashmani/MmMi-Datasets. Datasets generated based on Figure 1 can be found at https://huggingface.co/datasets/RKHashmani/mmmi-dag1-2modalities-cifar10