BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CAIL//Events//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:CAIL Events
BEGIN:VEVENT
UID:2026-04-17-danqi-chen@cail.columbia.edu
DTSTAMP:20260823T060708Z
DTSTART:20260417T150000Z
DTEND:20260417T160000Z
SUMMARY:ML Seminar: Danqi Chen - From Needle-in-a-Haystack to Long-Horizon
  Agents: A Retrospective on Long-Context Language Models
LOCATION:School of Social Work\, Room C03
DESCRIPTION:Language models' context sizes have rapidly increased from tho
 usands to millions of tokens\, reshaping how we build and use these models
 . In this talk\, I will trace this evolution along three dimensions: (1) h
 ow we think about training long-context language models from data (and arc
 hitecture) perspectives\, (2) how our evaluation and applications have shi
 fted — from synthetic retrieval tests to test-time scaling and long-hori
 zon agents\, and (3) how we should rethink inference and scaffolding to ma
 ke better use of long context\, beyond naively filling the context window.
  I will draw on recent work from our group on long-context model training\
 , evaluation\, and effective context management for long-horizon agentic t
 asks.\n\nhttps://cail.columbia.edu/events/2026-04-17-danqi-chen
URL:https://cail.columbia.edu/events/2026-04-17-danqi-chen
END:VEVENT
END:VCALENDAR