BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Making Better Use of (Large) Language and Translation Models with 
 Simple Inference Improvements. - Rico Sennrich\, University of Zurich
DTSTART:20240222T110000Z
DTEND:20240222T120000Z
UID:TALK212590@talks.cam.ac.uk
CONTACT:Panagiotis Fytas
DESCRIPTION:In a field where the state of the art is often advanced by sca
 le - building larger models on more data - I will make the argument that a
  surprising amount of progress can be achieved with simple modifications t
 o inference algorithms. In this talk\, I will focus on machine translation
 \, where massively multilingual models and large language models have been
  shown to handle many translation directions\, but which still suffer from
  problems such as hallucinations or translations in the wrong language. I 
 will show how these issues can be reduced massively with contrastive decod
 ing methods that pair each input with appropriate contrastive inputs. I wi
 ll also discuss Minimum Bayes Risk (MBR) Decoding\, a decoding method that
  has received renewed interest because it avoids common pitfalls in machin
 e translation\, but which suffers from a major increase in computational c
 ost. However\, I will show how the computational complexity of MBR decodin
 g can be reduced from quadratic to linear to the number of samples by usin
 g reference aggregation.
LOCATION:https://cam-ac-uk.zoom.us/j/97599459216?pwd=QTRsOWZCOXRTREVnbTJBd
 XVpOXFvdz09
END:VEVENT
END:VCALENDAR