A Causal Foundation Model for Structure and Outcome Prediction

About

We introduce TabPFN-CFM, a causal foundation model that can handle multiple causal problems. TabPFN-CFM predicts both causal structure and outcomes from observational data, supports queries on all three levels of Pearl's Causal Hierarchy and uses known graph structure when available to improve predictions. TabPFN-CFM is trained on synthetic datasets, and generalises to real datasets, demonstrating improved performance over both structural and outcome prediction baselines.

Max Zhu, Martino Mansoldo, Ching-Hao Wang, Stefan Groha · 2026

Related benchmarks

Task	Dataset	Result
Interventional Outcome Prediction	SEMs in-distribution n=1024 (fit samples)	MSE0.551	7
Interventional Outcome Prediction	SEMs n=1024 fit samples (out-of-distribution)	MSE0.739	7
Interventional Prediction	Law School Admissions Interventional (test)	MSE0.896	7
Observational Outcome Prediction	SEMs out-of-distribution n=1024 fit samples	MSE0.537	7
Outcome Prediction	Amazon Sales Interventional (test)	MSE0.455	7
Observational Outcome Prediction	SEMs in-distribution n=1024 (fit samples)	MSE0.433	7
Outcome Prediction	Law School Admissions Observational (test)	MSE0.872	7
Outcome Prediction	Amazon Sales Observational (test)	MSE0.012	7
Adjacency Matrix Prediction	Law School Admissions	AUROC0.886	6
Ancestral Matrix Prediction	Law School Admissions	AUROC91.2	6