A 1.5B Model Learns to Skip Its Own Chain-of-Thought — and Gets 41% Cheaper Without Getting Dumber

A 1.5B Model Learns to Skip Its Own Chain-of-Thought — and Gets 41% Cheaper Without Getting Dumber | BedrockNews