BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Value Reasoning and Test-Time Verification for Trustworthy LLMs - 
 Prof. Lu Wang (University of Michigan)
DTSTART:20250619T130000Z
DTEND:20250619T140000Z
UID:TALK231049@talks.cam.ac.uk
CONTACT:Shun Shao
DESCRIPTION:Abstract:\nDespite their impressive capabilities\, large langu
 age models (LLMs) continue to face significant limitations in complex real
 -world settings\, particularly when navigating high-stakes moral reasoning
  or when efficient and trustworthy test-time behavior is required. This ta
 lk explores two complementary directions that address these challenges: ev
 aluating limitations in value reasoning and scaling verification through e
 fficient process supervision.\n\nFirst\, I introduce CLASH\, a new benchma
 rk that examines how well LLMs reason about dilemmas involving conflicting
  values. CLASH enables a structured analysis of decision ambivalence\, psy
 chological discomfort\, and value shifts over time. The benchmark reveals 
 the difficulty LLMs have in representing nuanced human value reasoning\, e
 specially in ambiguous or temporally dynamic contexts.\n\nSecond\, I prese
 nt ThinkPRM\, a generative process reward model that enables step-by-step 
 verification using long chain-of-thought reasoning. Unlike traditional dis
 criminative PRMs that require extensive labeled supervision\, ThinkPRM is 
 trained on only a fraction of the process data by leveraging LLMs’ inher
 ent reasoning abilities to generate and verify each step in a solution. Th
 is approach supports more scalable and efficient test-time oversight\, out
 performing strong baselines in various domains.\n\nBio:\nLu Wang is an Ass
 ociate Professor in Computer Science and Engineering at University of Mich
 igan\, Ann Arbor. Previously\, she was an Assistant Professor in Khoury Co
 llege of Computer Sciences at Northeastern University. She received her Ph
 .D. in Computer Science from Cornell University. Her research focuses on b
 uilding trustworthy large language models that produce factual\, accurate\
 , and safe content. She has been working on problems of summarization\, re
 asoning\, evaluation\, as well as applications in AI for education and com
 putational social science. Lu has received paper awards at ACL\, CHI\, and
  SIGDIAL. She won the NSF CAREER award in 2021.\n
LOCATION:https://cam-ac-uk.zoom.us/j/97599459216?pwd=QTRsOWZCOXRTREVnbTJBd
 XVpOXFvdz09
END:VEVENT
END:VCALENDAR