Prompting or training a model to reason in explicit intermediate steps, trading more tokens for better answers.
← All terms