I'm not claiming that character level predictions result in superior overall performance. Not at all. My claim is merely that it's more difficult for models to

I'm not claiming that character level predictions result in superior overall performance. Not at all. My claim is merely that it's more difficult for models to reverse character strings specifically when their direct input is not individual characters. Not impossible, and sure you could fine-tune it for perfect results. But the whole reason large language models are interesting is that they don't require fine-tuning to perform an incredible range of tasks.