Large language models can effectively perform few-shot relation extraction tasks with the CoT-ER approach, outperforming fully-supervised methods.
