BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20251027T230626EDT-4381IUhoFp@132.216.98.100
DTSTAMP:20251028T030626Z
DESCRIPTION:Yifan Feng\n\nNUS Business School\, Singapore\n\nA Mallows-type
  Model for Preference Learning from Ranked Choices\n\nDate: Friday\, Octob
 er 31\, 2025\n	Time: 10:00 - 11:00 am\n	Location: Bronfman Building\, Room 0
 46\n\n\nAbstract\n\nThe first part of the talk introduces a novel distance
 -based (“Mallows-type”) ranking model built on a new distance function\, t
 he reverse major index (RMJ). The model is motivated by the problem of lea
 rning customer preferences from a general feedback structure we call 'rank
 ed choices\,' where each participant selects and ranks their top-k items f
 rom a displayed assortment. The RMJ-based ranking model yields surprisingl
 y parsimonious\, closed-form choice probability distributions over arbitra
 ry display sets\, which makes estimation both efficient and scalable in pr
 actice. We establish model identifiability and provide strong consistency 
 guarantees for the estimator\, even under model misspecification and low v
 ariation in assortments.\n\nIf time permits\, the second part of the talk 
 explores how (ranked) choice modeling can enhance LLM alignment. Current a
 lignment methods typically rely on pairwise preference optimization. While
  simple\, this approach overlooks richer forms of human feedback. To addre
 ss this\, we propose Ranked Choice Preference Optimization (RCPO). It is a
  unified framework that connects preference optimization with (ranked) cho
 ice modeling through maximum likelihood estimation. The framework is versa
 tile\, accommodating both utility-based and rank-based choice models. We e
 mbed the Mallows–RMJ model into RCPO as an illustration\, showing how appr
 opriate feedback structures\, combined with the right model\, can yield mo
 re effective alignment.\n
DTSTART:20251031T140000Z
DTEND:20251031T150000Z
LOCATION:Room 046\, Bronfman Building\, CA\, QC\, Montreal\, H3A 1G5\, 1001
  rue Sherbrooke Ouest
SUMMARY:Management Science Research Centre (MSRC) Seminar: Yifan Feng
URL:/bensadoun-school/channels/event/management-scienc
 e-research-centre-msrc-seminar-yifan-feng-368559
END:VEVENT
END:VCALENDAR