When Do Learned Blocks Preserve Causal Meaning? Block Identifiability and Constructive Causal Abstractions

arXiv:2610.18155v1 Announce Type: cross Abstract: Standard identifiability theory assumes one-dimensional latent factors, but learned representations can encode latent information across multidimensional blocks whose internal coordinates are not uniquely deter

← Back to all developments · Signals are a development reported by more than one source. See the methodology.