The paper explores the ability of large language models (LLMs) to judge the quality of their own generations. It introduces a reasoning with refinement strategy called ART, which improves performance on multistep reasoning tasks by asking necessary questions and ranking refinements. The approach achieves a 5-point gain over self-refinement baselines while using a smaller model as the decision maker.

https://arxiv.org/abs//2311.07961

YouTube: https://www.youtube.com/@ArxivPapers

TikTok: https://www.tiktok.com/@arxiv_papers

Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016

Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers