A new study demonstrates that a committee of weaker reasoning models can outperform stronger models by effectively selecting correct solutions through a critic-comparator mechanism. This approach shows promise in improving AI performance, particularly in language models, by leveraging existing capabilities within weaker models.