1147 page LLM proof!?
Statement by the Yang-Mills Working Group
Michael R. Douglas, David J. Gross (Chair), Arthur Jaffe,
Igor R. Klebanov and Nikita Nekrasov
kitp.ucsb.edu/sites/default/…
And there is one opening for a postdoc/staff scientist yet at the group of Fundamental Physics and Algebraic Combinatorics. The group currently has 3 postdocs (2 mathematicians, 1 physicist).
inspirehep.net/jobs/3216098
The applications will be reviewed on arrival from mid November
pilk retweeted
“The ultimate hidden truth of the world is that it is something we make, and could just as easily make differently”
— David Graeber (1961-2020)
Link: tinyurl.com/y9a24tv9
pilk retweeted
when colleagues say a book making fun of academia plays into the hands of conservatives
pilk retweeted
"Our initial evaluation suggested that GPT-6-Astra's score was in the late eighties followed closely by Opus 5.5. But, as we started to investigate the mistakes that the models might have made, the models, especially Astra, insisted that they were right and that it was the expert-supplied answers that were erroneous. When we went back to the experts it turned out, in multiple instances, that that the experts found some subtle errors buried deep inside their calculations. As they revised their answers, Astra's score started to inch up and it finally ended up with 98% (49/50)."
pilk retweeted
👀 Ok this is very interesting, per Maxim Kontsevich:
"The AI agent then proposed me a better definition of what I'm doing, so I'm shocked"
youtube.com/watch?v=c_VFKWs4…
This paper of Lloyd Nick Trefethen once again finds uses in a problem I care about.
I love how simple and powerful rational approximations are :)
The paper is also a joy to read! Definitely one of my favorite papers out there: arxiv.org/pdf/1803.00973
pilk retweeted
Let us wait a year and see what happens, before waxing metaphoric. The previous guild that was replaced was the coders, a year and a half after vibe coding became a thing, we see the slop and the bloated code, but not the incredible code and progress this was supposed to unlock.
Matthew Schwartz has apparently spent ~3 months (on sabbatical from his Harvard professorship to as a visiting researcher at Anthropic, it seems). I've only spent ~30 minutes looking diagonally at it, and have not converged on what I think about it.
This paper is the closest to me! Anyways, it looks quite ok. I was mainly looking at the part of twisted (co)homology. It's not surprising that twisted cohomology should be generous and give you a double copy here.
What is interesting is that the low-energy expansion of
these integrals (if one believes the computation...) *also* give multiple zeta values :) when it doesn't need to be the case!
The paper does remark the computation, and then looks at other mechanisms, like the single valued map
bootloops.ai/stringintegral.…