Self-Refine: Iterative Refinement with Self-Feedback

The authors reveal in their paper that LLMs can generate feedback on their work, to repeatedly improve the output.