Title: ICML Invited Talk Proxy objectives in reinforcement learning from human feedback
Description: No description
Keywords: No keywords
Text content: ICML Invited Talk Proxy objectives in reinforcement learning from human feedback    ICML 2023      Skip to yearly menu bar Skip to main content  Main Navigation                       ICML                                      Help/FAQ                                                    Contact ICML                                                    Downloads                                                    Code of Conduct                                                    Create Profile                                                    Journal To Conference Track                                                    Diversity & Inclusion                                                    Privacy Policy                                                    Future Meetings                                                    Press                                                    Careers                                                     My Stuff                                    Login                                 Select Year: (2023)            2025                     2024                     2023                     2022                     2021                     2020                     2019                     2018                     2017                     2016                     2015                     2014                     2013                     2012                     2011                     2010                     2009                     2008                     2007                     2006                     2005                     2004                     2002                     1996                     IMLS Archives                                                Getting Started                                                   Schedule                                                   Tutorials                                      Main Conference                                      Invited Talks                                                    Orals                                                    Awards                                                    Test of Time Award                                                    Papers                                                     Workshops                                      Community                                      Affinity Events                                                    Socials                                                    Mentorship                                                    Town Hall / Business Meeting                                                     Sponsors                                                   Organizers                                             Help                                      FAQ                                                    Presenters Instructions                                                    Moderators Instructions                                                    RocketChat Help                                                    RocketChat Desktop Client                                                                    Invited Talk                                                   Proxy objectives in reinforcement learning from human feedback                                               John Schulman                      Moderator                             :  Emma Brunskill  Exhibit Hall 2                                                                                [                              Abstract                          ]                                                                                                                                                                            Abstract: Proxy objectives are a fundamental concept in machine learning. That is, there's a true objective that we care about, but it's hard to compute or estimate, so instead we construct a locally-valid approximation and optimize that. I will examine reinforcement from human feedback with this lens, as a chain of approximations, each of which can widen the gap between the desired and achieved result.                  Chat is not available.                         Successful Page Load                                ICML uses cookies for essential functions only. We do not sell your personal                         information.                         Our Privacy Policy »    Accept                             Cookies                                                                            The ICML Logo above may be used on presentations. Right-click and choose                             download. It is a vector graphic and may be used at any scale.                                                           Useful links                            About ICML ICML Proceedings at PMLR Code of Conduct        Contact   1269 Law Street, San Diego CA 92109                                Email            ICML Proceedings at PMLR