BEGIN:VCALENDAR
VERSION:2.0
CALSCALE:GREGORIAN
PRODID:UW-Madison-Physics-Events
BEGIN:VEVENT
SEQUENCE:2
UID:UW-Physics-Event-7796
DTSTART:20220727T160000Z
DTEND:20220727T171500Z
DTSTAMP:20260414T113558Z
LAST-MODIFIED:20220725T142812Z
LOCATION:Online Seminar: Please sign up for our mailing list at www.ph
 ysicsmeetsml.org for zoom link
SUMMARY:Minerva: Solving Quantitative Reasoning Problems with Language
  Models\, Physics ∩ ML Seminar\, Guy Gur-Ari\, Google Brain
DESCRIPTION:Quantitative reasoning tasks which can involve mathematics
 \, science\, and programming are often challenging for machine learnin
 g models in general and for language models in particular. We show tha
 t transformer-based language models obtain significantly better perfor
 mance on math and science questions when trained in an unsupervised wa
 y on a large\, math-focused dataset. Performance can be further improv
 ed using prompting and sampling techniques including chain-of-thought 
 and majority voting. Minerva\, a model that combines these techniques\
 , achieves SOTA on several math and science benchmarks. I will describ
 e the model\, its capabilities and limitations.
URL:https://www.physics.wisc.edu/events/?id=7796
END:VEVENT
END:VCALENDAR