We develop the Less Greedy Equivalence Search algorithm for learning causal structure from observational and interventional data with prior knowledge.

We propose a percolation-based differentiable d-separation framework that bridges constraint-based causal discovery methods with differentiable DAG structure learning.