It sounds like they are describing a regex filter being applied to the model's beam search. LLMs generate the most probable words, but they are frequently track

It sounds like they are describing a regex filter being applied to the model's beam search. LLMs generate the most probable words, but they are frequently tracking several candidate phrases at a time and revising their combined probability. It lets them self correct if a high probability word leads to a low probability phrase. I think they are saying that if highest probability phrase fails the regex, the LLM is able to substitute the next most likely candidate.