one-to-many-translation

This paper proposes a discrete diffusion framework for efficient one-to-many machine translation, achieving sublinear latency scaling with the number of target languages and supporting zero-shot transfer to unseen source languages.