The prompt may request decomposition or provide worked reasoning examples before the final answer. Intermediate text can improve some tasks but is not guaranteed to be correct or faithful, so verification remains necessary.
Chain-of-thought prompting encourages a model to use intermediate reasoning steps for a complex task.
The prompt may request decomposition or provide worked reasoning examples before the final answer. Intermediate text can improve some tasks but is not guaranteed to be correct or faithful, so verification remains necessary.