This paper introduces Contrastive Preference Learning (CPL), a new algorithm for optimizing behavior from human feedback without learning reward functions. CPL is fully off-policy and can be applied to arbitrary MDPs, making it scalable and simpler than prior methods.

https://arxiv.org/abs//2310.13639

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