BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CAIL//Events//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:CAIL Events
BEGIN:VEVENT
UID:2026-09-24-sharon-levy@cail.columbia.edu
DTSTAMP:20260921T212455Z
DTSTART:20260924T200000Z
DTEND:20260924T210000Z
SUMMARY:NLP Seminar: Sharon Levy - Beyond Surface-Level Safety: Uncovering
  Implicit Biases
LOCATION:CSB 453
DESCRIPTION:Current safety alignment techniques effectively mitigate expli
 cit biases but can fail on more subtle implicit biases. First\, I will dis
 cuss our methodology to discover implicit biases using logic puzzles. This
  provides automatic generation and evaluation and can be easily tailored t
 o various demographic attributes. Next\, I will highlight the issue of sel
 ective refusal bias\, where models may refuse to generate harmful content 
 targeting some demographic groups and not others. Together\, this research
  highlights areas of improvement for model alignment.\n\nhttps://cail.colu
 mbia.edu/events/2026-09-24-sharon-levy
URL:https://cail.columbia.edu/events/2026-09-24-sharon-levy
END:VEVENT
END:VCALENDAR