I suspect that regex is very much like mathematics for these models: they're just bad at it. If you provided the model failure cases for each failed attempt, yo

I suspect that regex is very much like mathematics for these models: they're just bad at it. If you provided the model failure cases for each failed attempt, you might get better output. I've also seen folks say that requiring reasoning makes answers more robust as well.