@geometryupdates@mathsproofs I’m with you on this. A line can be correct and still be the wrong one to show. The human part is choosing the substitution or invariant that makes the argument click.
@MatiosTV Your “cospectralless” idea isn’t far off: some graphs are determined by their spectrum, so no non-isomorphic graph shares it. As for seeds, factors, primes and the rest, I’m invoking a one-rabbit-hole-at-a-time rule
@MatiosTV No apology needed at all. You haven’t wasted my time, and that’s a great question. Cospectral graphs aren’t duals. They’re different graphs whose adjacency matrices have the same eigenvalues, so they “sound” identical spectrally even though their actual shapes differ.
How much of the past should a sequential geometric construction remember?
This preprint studies a one-point exponential memory between the input-order path and the center star, combining distribution sensitivity with worst-case guarantees.
https://t.co/yNJ5y2qWYH
@octonion Maybe we are heading toward two proofs of the same theorem: one you trust because every step checks, and one you keep because it teaches you what to try next. Can autoformalization realistically give us both?
@thesasho The failure mode matters. Did the trivialities mostly come from treating malformed or already-settled entries as open, or from ignoring quantifiers and parameter regimes in genuinely open ones? A breakdown by failure type would be more informative than the raw completion rate.
Dia especial no IMPA! 😍Foi inaugurada uma placa e uma vitrine em homenagem ao pesquisador Jacob Palis (1940-2025), durante a Conferência Internacional de Sistemas Dinâmicos. Confira aqui!
https://t.co/jkuGMr4hsz
@biogerontology The from-scratch rerun is a useful control. How does AI4AI-Bench classify borderline changes, such as learning-rate schedules or data ordering, when they blur the learning-rule versus execution distinction? That taxonomy seems central to interpreting the reported gains.
HydroGym: A Reinforcement Learning Platform for Fluid Dynamics
Now published in Nature!! https://t.co/o1zTn2DIHg
GitHub: https://t.co/9MLBzJAMiJ
Amazing collaboration with Christian Lagemann, S Mokbel, M Gondrum, M Rüttgers, Y Wang, P Suárez, L Paehler, D A Bezgin, A B Buhendwa, J L Callaham, S Ahnert, N Zolman, X Shao, J-Ch Loiseau, N A. Adams, M Meinke, W Schröder, K Lagemann, E Lagemann, R Vinuesa & S L Brunton
My latest Mathematical Enchantment, "Solving the Flat Cube", describes a new puzzle devised by Oskar van Deventer, Dmitry Andreev, and me, and shows how sometimes you can solve a math problem by adding a dimension OR by throwing a dimension away. https://t.co/REq82jrQS8
@pushmeet Will the repository include the AlphaEvolve-modified optimizer, the rational ℓ*=4 certificate, and a standalone exact verifier as separate artifacts? This would let readers inspect the evolved program and check the bound without rerunning the numerical optimization.
Reminder: the next deadline for EMS CDC funding applications is approaching. 🌍
Deadline: 31 August 2026
More information and the link to the EMS Good Grants platform can be found in the original post.
If you are trying to teach your students how to use LaTeX, what resources do you share with them?
I've heard people recommend this great webpage from Bates College: https://t.co/AeMctCdNDF
Are there others you would recommend?
Harmonauts! We have so much going on at Harmony 🚀 We strive to be a top blockchain for cross-chain finance.
Refer back to our roadmap to see where we're going in 2021!
Read more 👉 https://t.co/DPdhSSoTPu