What's your evidence that it will result in worse translations? I'm skeptical that such a thing as a universally optimal translation exists in cases beyond the

What's your evidence that it will result in worse translations? I'm skeptical that such a thing as a universally optimal translation exists in cases beyond the trivial. But if it does, I see no reason to think LLMs are anywhere close to it, so I think nobody will be able to tell the difference with watermarking. That's certainly true for code. LLM code is at best mediocre. There is oceans of room to subtly watermark generated code without practical impact.