Self-Alignment of Large Language Models
Jason Weston
Meta AI
The NLP Reading Group is happy to have Jason Weston who will be giving a talk about “Self-Alignment of Large Language Models”.
Talk Description
We describe some recent methods for LLMs whereby they can self-learn how to perform better at tasks relevant to human users. In particular we describe the methods of Iterative DPO (https://arxiv.org/abs/2312.16682), Self-Rewarding LLMs (https://arxiv.org/abs/2401.10020), Iterative Reasoning Preference Optimization (https://arxiv.org/abs/2404.19733) and Following Length Constraints in Instructions (https://arxiv.org/abs/2406.17744).
Speaker Bio
Jason Weston is a research scientist at Meta AI, USA and a Visiting Research Professor at NYU. He earned his PhD in machine learning at Royal Holloway, University of London and at AT&T Research in Red Bank, NJ (advisors: Alex Gammerman, Volodya Vovk and Vladimir Vapnik) in 2000. From 2002 to 2003 he was a research scientist at the Max Planck Institute for Biological Cybernetics, Tuebingen, Germany. From 2003 to 2009 he was a research staff member at NEC Labs America, Princeton. From 2009 to 2014 he was a research scientist at Google, NY. Jason has published over papers primarily in the fields of machine learning and NLP, including best paper awards at ICML and ECML, and a Test of Time Award for his work “A Unified Architecture for Natural Language Processing: Deep Neural Networks with Multitask Learning”, ICML 2008 (with Ronan Collobert). He was part of the YouTube team that won a National Academy of Television Arts & Sciences Emmy Award for Technology and Engineering for Personalized Recommendation Engines for Video Discovery.
Logistics
Date: July 2nd
Time: 11:00 AM
Location: F01 or via Zoom (See email)