aiqbench analysis
This research advances causal discovery in linear Gaussian models by addressing inconsistencies in standard scoring criteria for singular statistical models. The authors introduce a differentiable search framework grounded in singular learning theory, utilizing WBIC and Automatic Differentiation Variational Inference to efficiently explore graph structures under unobserved confounding. Experimental results on synthetic and real-world benchmarks demonstrate improved performance over existing BIC-based baselines, offering a robust method for inferring causal graphs where traditional approaches fail.
Differentiable Causal Discovery for Singular Linear Models under Confounding
arXiv:2601.01368v2 Announce Type: replace Abstract: Score-based causal discovery in the presence of unobserved confounders requires both a consistent scoring criterion and an efficient search over graph structures. Linear causal models with correlated errors are naturally represented by acyclic directed mixed graphs (ADMGs), whose induced model families include singular statistical models for which the standard Bayesian Information Criterion is inconsistent. We develop a score-based framework for causal discovery over linear Gaussian ADMGs grounded in singular learning theory. We first verify that the conditions for consistency of the Widely Applicable Bayesian Information Criterion (WBIC) are satisfied by the linear causal model class. We then propose a scalable approximation of the WBIC via Automatic Differentiation Variational Inference (ADVI). Finally, we introduce a differentiable search over ADMGs through Gumbel-Softmax relaxations of both directed and bidirected adjacency. Experiments on synthetic and real-world benchmarks demonstrate improvements over the BIC-scored baseline on both singular and generic ADMGs under unobserved confounding.
Abstract:Score-based causal discovery in the presence of unobserved confounders requires both a consistent scoring criterion and an efficient search over graph structures. Linear causal models with correlated errors are naturally represented by acyclic directed mixed graphs (ADMGs), whose induced model families include singular statistical models for which the standard Bayesian Information Criterion is inconsistent. We develop a score-based framework for causal discovery over linear Gaussian ADMGs grounded in singular learning theory. We first verify that the conditions for consistency of the Widely Applicable Bayesian Information Criterion (WBIC) are satisfied by the linear causal model class. We then propose a scalable approximation of the WBIC via Automatic Differentiation Variational Inference (ADVI). Finally, we introduce a differentiable search over ADMGs through Gumbel-Softmax relaxations of both directed and bidirected adjacency. Experiments on synthetic and real-world benchmarks demonstrate improvements over the BIC-scored baseline on both singular and generic ADMGs under unobserved confounding.
Machine Learning (cs.LG)
arXiv:2601.01368 [cs.LG]
(or arXiv:2601.01368v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2601.01368 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 240 words
Related coverage
J++ Lens: Jacobian Filtering Enables More Faithful Workspace Lenses
LessWrong · Oct 8, 2026
PIRAMID: Progress and Plans
LessWrong · Jul 27, 2026
Differentiable Structure Learning for Cyclic Linear Gaussian Models with Latent Confounders
arXiv · Oct 1, 2026
Generalizing Identifiability Results in Causal Discovery from Complete Data to Missing Data Settings
arXiv · Oct 15, 2026