BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CAIL//Events//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:CAIL Events
BEGIN:VEVENT
UID:2026-04-10-greg-durrett@cail.columbia.edu
DTSTAMP:20260823T060647Z
DTSTART:20260410T150000Z
DTEND:20260410T160000Z
SUMMARY:ML Seminar: Greg Durrett - LLM Reasoning Beyond Scaling
LOCATION:School of Social Work\, Room C03
DESCRIPTION:Agentic large language models can write and debug complex code
 \, solve competition-level math problems\, and conduct in-depth literature
  review. These reasoning capabilities are enabled by scaling of data: pre-
 training data to learn vast knowledge\, fine-tuning data to learn natural 
 language reasoning\, and RL environments to refine that reasoning. In this
  talk\, I will investigate the current LLM reasoning paradigm\, its bounda
 ries\, and the future of LLM reasoning beyond scaling. First\, I will desc
 ribe the state of reasoning models and where I think scaling will lead to 
 additional successes. I will then shift to discussing issues which are not
  resolved by pure scaling. First\, I will describe our work on calibrating
  models' decisions through better understanding of their environments. We 
 find that explicitly telling an LLM its likelihood to succeed or fail at t
 asks allows it to reason about cost-benefit tradeoffs in its action space.
  Then\, I will describe our new benchmark CREATE\, which tests LLMs' capab
 ilities for associative creativity. I will highlight limitations of LLMs a
 pplied to creative tasks like scientific ideation and where I see future w
 ork making progress in these areas.\n\nhttps://cail.columbia.edu/events/20
 26-04-10-greg-durrett
URL:https://cail.columbia.edu/events/2026-04-10-greg-durrett
END:VEVENT
END:VCALENDAR