BEGIN:VCALENDAR
VERSION:2.0
PRODID:researchseminars.org
CALSCALE:GREGORIAN
X-WR-CALNAME:researchseminars.org
BEGIN:VEVENT
SUMMARY:Moritz Münchmeyer (University of Wisconsin-Madison)
DTSTART:20260217T193000Z
DTEND:20260217T203000Z
DTSTAMP:20260423T022741Z
UID:nhetc/140
DESCRIPTION:Title: <a href="https://researchseminars.org/talk/nhetc/140/">
 AI Reasoning in Theoretical Physics with TPBench and MadEvolve</a>\nby Mor
 itz Münchmeyer (University of Wisconsin-Madison) as part of NHETC Seminar
 \n\n\nAbstract\nLarge-language models are now powerful enough to assist ph
 ysicists with mathematical reasoning at research level. In this talk\, I w
 ill first present our dataset TPBench (tpbench.org)\, which was constructe
 d to benchmark and improve AI models specifically for theoretical physics.
  I will then discuss how test-time scaling and symbolic verification can b
 e used to improve their performance and reliability. In the second part of
  my talk I will present MadEvolve\, our new LLM-based system to iterativel
 y improve scientific algorithms. I will show that MadEvolve can set state-
 of-the-art results for algorithms in computational cosmology\, such as the
  reconstruction of initial conditions of the universe. Finally\, I will br
 iefly present preliminary work on reinforcement learning fine-tuning of re
 asoning for theoretical physics.\n
LOCATION:https://researchseminars.org/talk/nhetc/140/
END:VEVENT
END:VCALENDAR