Causal Foundation Models: Disentangling Physics from Instrument Properties
This addresses a fundamental challenge in time series analysis for domains like astronomy, enabling better generalization across heterogeneous instruments, though it appears incremental as it builds on existing foundation model concepts with causal enhancements.
The paper tackles the problem of foundation models for structured time series data conflating physical phenomena with instrument distortions, which limits generalization. Their causally-motivated model with dual-encoder architecture and contrastive learning significantly outperforms traditional models on simulated astronomical time series, especially in low-data regimes.
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.