BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CAIL//Events//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:CAIL Events
BEGIN:VEVENT
UID:2025-12-12-jason-weston@cail.columbia.edu
DTSTAMP:20260823T060943Z
DTSTART:20251212T160000Z
DTEND:20251212T170000Z
SUMMARY:ML Seminar: Jason Weston - Self-Improvement of LLMs
LOCATION:School of Social Work\, Room C03
DESCRIPTION:Classically\, learning algorithms were designed to improve the
 ir performance by updating their parameters (weights)\, while keeping othe
 r components\, such as the training data\, loss function\, and algorithm\,
  fixed. We argue that fully intelligent systems will be able to self-impro
 ve across all aspects of their makeup. We describe recent methods that ena
 ble large language models (LLMs) to self-improve in various ways\, increas
 ing their performance on tasks relevant to human users. In particular\, we
  describe methods whereby models are able to create their own training dat
 a (self-challenging)\, train on this data using themselves as their own re
 ward model (self-rewarding)\, and train themselves to better provide their
  own rewards (meta-rewarding). We then discuss the future of self-improvem
 ent for AI and key challenges that remain unresolved.\n\nhttps://cail.colu
 mbia.edu/events/2025-12-12-jason-weston
URL:https://cail.columbia.edu/events/2025-12-12-jason-weston
END:VEVENT
END:VCALENDAR