Interventional Time Series Priors for Causal Foundation Models

TL;DR

This paper introduces CausalTimePrior, a framework for generating synthetic time series data with interventional targets to advance causal foundation models in temporal data analysis.

Contribution

It presents a novel method for creating synthetic temporal causal models with interventional data, enabling training of causal foundation models for time series.

Findings

PFNs trained on CausalTimePrior can estimate causal effects in new TSCMs.

The framework supports complex causal graph structures and intervention types.

It bridges the gap in synthetic data generation for time series causal inference.

Abstract

Prior-data fitted networks (PFNs) have emerged as powerful foundation models for tabular causal inference, yet their extension to time series remains limited by the absence of synthetic data generators that provide interventional targets. Existing time series benchmarks generate observational data with ground-truth causal graphs but lack the interventional data required for training causal foundation models. To address this, we propose \textbf{CausalTimePrior}, a principled framework for generating synthetic temporal structural causal models (TSCMs) with paired observational and interventional time series. Our prior supports configurable causal graph structures, nonlinear autoregressive mechanisms, regime-switching dynamics, and multiple intervention types (hard, soft, time-varying). We demonstrate that PFNs trained on CausalTimePrior can perform in-context causal effect estimation on held-out TSCMs, establishing a pathway toward foundation models for time series causal inference.