aiqbench analysis
Decoupled Causal Discovery introduces a novel framework for inferring causal structures from observational data without relying on traditional conditional independence tests or restrictive functional assumptions. By directly identifying the Markov boundary through decoupling mechanisms, this method offers improved robustness in high-noise regimes and establishes theoretical guarantees for identifiability and completeness. This advancement addresses a core challenge in machine learning: moving beyond correlation to uncover underlying causal mechanisms essential for reliable AI systems.
Decoupled Causal Discovery
arXiv:2609.23535v1 Announce Type: new Abstract: Causal discovery from observational data is a fundamental yet challenging task in scientific research. While existing approaches are primarily based on conditional independence tests, structure scores, or restrictive functional assumptions, we propose Decoupled Causal Discovery (DCD), a novel decoupling-based perspective that does not rely on these methodologies. DCD directly identifies the Markov boundary (MB) by decoupling non-target variables via weighting functions, such that only variables within the MB preserve dependence with the target under the decoupled distribution. Building on this, DCD iteratively constructs the Completed Partially Directed Acyclic Graph (CPDAG) by exploiting structural asymmetries within the MBs. We establish the theoretical identifiability, soundness, and completeness of DCD. Empirical evaluations demonstrate that DCD achieves strong performance, particularly excelling in challenging noise regimes.
Abstract:Causal discovery from observational data is a fundamental yet challenging task in scientific research. While existing approaches are primarily based on conditional independence tests, structure scores, or restrictive functional assumptions, we propose Decoupled Causal Discovery (DCD), a novel decoupling-based perspective that does not rely on these methodologies. DCD directly identifies the Markov boundary (MB) by decoupling non-target variables via weighting functions, such that only variables within the MB preserve dependence with the target under the decoupled distribution. Building on this, DCD iteratively constructs the Completed Partially Directed Acyclic Graph (CPDAG) by exploiting structural asymmetries within the MBs. We establish the theoretical identifiability, soundness, and completeness of DCD. Empirical evaluations demonstrate that DCD achieves strong performance, particularly excelling in challenging noise regimes.
Machine Learning (cs.LG); Machine Learning (stat.ML)
arXiv:2609.23535 [cs.LG]
(or arXiv:2609.23535v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2609.23535 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
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 · 218 words
Related coverage
CausalMoE: A Billion-Scale Multimodal Foundation Model for Granger Causal Discovery with Pattern-Routed Heterogeneous Experts
arXiv · Jun 12, 2026
Noesis: Bidirectional Graph-RAG with Adaptive Parallelism and Cross-Knowledge-Base Semantic Discovery
arXiv · Aug 18, 2026
Self-discovering RL in the Era of Experience: Is Learning History an Asset or a Burden?
arXiv · Sep 30, 2026
SPEAR: Structure Property Explainability with Attention Regularization
arXiv · Aug 17, 2026