First, I don't want to waste time feeding something into an LLM that should be commented in the first place.

Second, not at all. An LLM can tell you how the regex works (hopefully). It can't tell you what each piece means in terms of the program's logic. Or at least not always and not reliably.