Causal Foundation Models: Disentangling Physics from Instrument Properties

TL;DR

This paper introduces a causally-motivated foundation model that disentangles physical phenomena from instrument effects in structured time series data, improving generalization and adaptation especially with limited data.

Contribution

It proposes a dual-encoder architecture with contrastive learning to separate physical and instrumental factors, enhancing model robustness and interpretability.

Findings

Outperforms traditional models on simulated astronomical data

Enables effective few-shot learning and adaptation

Demonstrates importance of causal structure in representation learning

Abstract

Foundation models for structured time series data must contend with a fundamental challenge: observations often conflate the true underlying physical phenomena with systematic distortions introduced by measurement instruments. This entanglement limits model generalization, especially in heterogeneous or multi-instrument settings. We present a causally-motivated foundation model that explicitly disentangles physical and instrumental factors using a dual-encoder architecture trained with structured contrastive learning. Leveraging naturally occurring observational triplets (i.e., where the same target is measured under varying conditions, and distinct targets are measured under shared conditions) our model learns separate latent representations for the underlying physical signal and instrument effects. Evaluated on simulated astronomical time series designed to resemble the complexity of variable stars observed by missions like NASA's Transiting Exoplanet Survey Satellite (TESS), our method significantly outperforms traditional single-latent space foundation models on downstream prediction tasks, particularly in low-data regimes. These results demonstrate that our model supports key capabilities of foundation models, including few-shot generalization and efficient adaptation, and highlight the importance of encoding causal structure into representation learning for structured data.