This is the most technically correct answer - language models are at their core trying to continue a string of text. If you had trained the model on knowing exa

This is the most technically correct answer - language models are at their core trying to continue a string of text. If you had trained the model on knowing exactly which letters represented the other letters in this particular puzzle, it might work.