Evaluating Communicative Success in Machine Translation

A three-layer checklist-and-judge framework evaluates machine-translation interpreter agents across semantic, pragmatic, and cultural-social dimensions rather than relying on sentence-level fidelity metrics.