BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Understanding the Interplay between LLMs' Utilisation of Parametri
 c and Contextual Knowledge  - Prof Isabelle Augenstein (University of Cope
 nhagen)
DTSTART:20260501T110000Z
DTEND:20260501T120000Z
UID:TALK244069@talks.cam.ac.uk
CONTACT:Suchir Salhan
DESCRIPTION:Language Models (LMs) acquire parametric knowledge from their 
 training process\, embedding it within their weights. The increasing scala
 bility of LMs\, however\, poses significant challenges for understanding a
  model's inner workings and further for updating or correcting this embedd
 ed knowledge without the significant cost of retraining. Moreover\, when u
 sing these language models for knowledge-intensive language understanding 
 tasks\, LMs have to integrate relevant context\, mitigating their inherent
  weaknesses\, such as incomplete or outdated knowledge. Nevertheless\, stu
 dies indicate that LMs often ignore the provided context as it can be in c
 onflict with the pre-existing LM's memory learned during pre-training. Con
 flicting knowledge can also already be present in the LM's parameters\, te
 rmed intra-memory conflict. This underscores the importance of understandi
 ng the interplay between how a language model uses its parametric knowledg
 e and the retrieved contextual knowledge.\n\nIn this talk\, I will aim to 
 shed light on this important issue by presenting our research on evaluatin
 g the knowledge present in LMs\, diagnostic tests that can reveal knowledg
 e conflicts\, as well as on understanding the characteristics of successfu
 lly used contextual knowledge.\n\n\n\n**Bio:** Isabelle Augenstein is a Pr
 ofessor at the University of Copenhagen\, Department of Computer Science\,
  where she heads the Natural Language Processing section. Her main researc
 h interests are fair and accountable NLP\, including challenges such as ex
 plainability\, factuality and bias detection. Prior to starting a faculty 
 position\, she was a postdoctoral researcher at University College London\
 , and before that a PhD student at the University of Sheffield. In October
  2022\, Isabelle Augenstein became Denmark’s youngest ever female full p
 rofessor. She currently holds a prestigious ERC Starting Grant on 'Explain
 able and Robust Automatic Fact Checking’\, and her research has been rec
 ognised by a Karen Spärck Jones Award\, as well as a Hartmann Diploma Pri
 ze. She is a member of the Royal Danish Academy of Sciences and Letters\, 
 and co-leads the Danish Pioneer Centre for AI.
LOCATION:FW11 Hybrid (In-Person + Online). Here is the Google Meet Link: h
 ttps://meet.google.com/cru-hcuo-rhu
END:VEVENT
END:VCALENDAR