not in a way that it is effectively used - in real life all of the papers using CoT compare against a weak baseline and the benefits level off extremely quickly.

nobody except for recent deepmind research has shown test time scaling like o1