RL Doesn't Just Sharpen What a Model Already Knows — It Teaches It to Combine Skills Into New Ones

RL Doesn't Just Sharpen What a Model Already Knows — It Teaches It to Combine Skills Into New Ones | BedrockNews