MIT researchers develop self-learning language models that outperform larger counterparts

This innovative approach to language modeling challenges the conventional belief that smaller models possess limited capabilities.