Work

From Georgetown to GPT: Six Decades of Machine Translation, the Multilingual NMT Ceiling, and Where Large Language Models Win - and Where They Do Not