NeSyAM Enables Neurosymbolic Action Model Learning Under Partial Observability

NeSyAM introduces a variational framework for learning action models when state information is incomplete or observations are missing. This addresses a critical gap in existing neurosymbolic approaches that require fully observable image traces.