AI Dynamics

Global AI News Aggregator

About

Invalid reasoning yields 80-90% of chain-of-thought gains

Building on @sewon__min et al.'s find that multiple-choice examples with random answers barely harm performance, @BoshiWang2 et al. find you can achieve 80-90% the gains of chain-of-thought using only examples with invalid reasoning.

→ View original post on X — @goodside