Hierarchical Identifiability Criteria for Causal Abstractions

This work establishes a formal hierarchy of identifiability notions for causal effects within collections of compatible graphs. It proves that causal effects can be identified across an entire collection of graphs even if no single causal derivation applies uniformly to every graph.