The paper proposes a framework called PIT that uses human preference data to implicitly train large language models (LLMs) to improve their response quality. PIT outperforms prompting-based methods in experiments on real-world and synthetic datasets.

https://arxiv.org/abs//2310.00898

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