Comment by JohnKemeny

We shouldn’t anthropomorphize LLMs—they don’t “struggle.” A better framing is: why is the most likely next token, given the prior context, one that reinforces the earlier wrong turn?