madaan / self-refine

LLMs can generate feedback on their work, use it to improve the output, and repeat this process iteratively.
https://selfrefine.info
Apache License 2.0
596 stars 48 forks source link