When should a reasoning model quit a problem it probably can't solve? Conformal Thinking sets stop/continue thresholds for test-time compute by distribution-free risk control, holding the error rate under a target you pick. It adds a lower threshold that gives up early when confidence isn't climbing fast enough, saving tokens on likely-unsolvable problems.

https://benjaminhan.net/posts/20260624-conformal-thinking/?utm_source=mastodon&utm_medium=social

#ConformalPrediction #LLMs #Reasoning #AI