Both. Atom swaps are the discontinuities you see, replica swaps are done by swapping the temperature (the color of the frames)
03.03.2026 23:08
π 0
π 0
π¬ 0
π 0
Both. Atom swaps are the discontinuities you see, replica swaps are done by swapping the temperature (the color of the frames)
Maybe less extreme than these, but in the data record archive.materialscloud.org/records/18tk... we have the structures we rejected because they had inconsistent DFT energies. these could be fantastic stress-tests for any DFT code
Well, thank you for leading the development of a code that is robust enough to converge > 50% of the structures with all elements from 1 to 102. Took long enough that most would have decayed, but that's another story ^_^
π’ We have been working on a new universal atomistic dataset that combines the principles of MAD with a meta-GGA level of theory, so we can all simulate water that does not freeze at 500K π§ , and have all our bases covered, with reference data for every isotope with a half-life above 24 hours β’οΈ
So let us show you just how *universal* #PET-MAD-1.5 can be. This is a movie of a parallel tempering simulation, with replicas from 300K to 3000K, of what we call a "Mendeleev cluster" - one atom each of every element from 1 to 102.
Snapshot of a ring-polymer instanton calculation for the reaction of methane with a H radical
New recipe just landed on the atomistic-cookbook.org π§βπ³π. Thanks to @yairlitman.bsky.social for explaining how to use ipi-code.org to perform ring-polymer instanton calculations of reaction rates that include quantum tunneling effects βοΈβ‘. Check it out π atomistic-cookbook.org/examples/rin...
π’ New #preprint is out! Investigating the many flavors of last-layer #UQ, Moritz and π§βπMatthias propose a practitioners' guide on "how to train a shallow ensemble". TL;DR? for good calibration use NLL, include force, and optimize the backbone, fine-tuning for speed! ππβ‘οΈ arxiv.org/html/2602.15...
No Install. No Setup. Just Chemical Shift Predictions.
At shiftml.materialscloud.io we host the latest ShiftML3 in the web and you can predict chemical shifts of organic crystals for free in the web!
That's def the case for BPNNs. For MACE it's more subtle - it has enough parameters but the heavy bias towards fast converging body order expansion ends up hurting when it has to learn that it does not converge fast.
The body ordered expansion, equations
Many #machinelearning potentials are built (or understood) in terms of "atomic cluster expansions" that link directly to a body-ordered energy decomposition that can be computed explicitly with a sequence of electronic structure calculations. But what kind of expansion do they learn in practice? Aπ§΅
If you want to learn about materials modeling, from DFT to MD, well marinated in a spicy ML sauce, don't miss out the @ictp.bsky.social / @nccr-marvel.bsky.social college. Details and application instructions here indico.ictp.it/event/11146/. See you in Miramare!
We are delighted to announce the launch of NCCR Separations β a new National Center of Competence in Research dedicated to breakthrough separation technologies for a sustainable future!
Read more: actu.epfl.ch/news/a-new-n...
#NCCRSeparations #EPFL #GreenChemistry #EnergyTransition #ClimateTech
Fantastic news from the @snf-fns.ch, who despite the budget cuts managed to fund six new NCCRs. Looking forward to doing some cool simulations to advance separation science! actu.epfl.ch/news/a-new-n...
PIMD Kusama
Table showing results of a few representative universal model on the matbench leaderboard
If you got curious by the PET-OAM results a week ago, you can learn more reading up arxiv.org/abs/2601.16195. Including some general considerations on how to train and use safely an unconstrained ML potential.
The fact that the background image was AI generated about 18 months ago tells you a lot of how fast the general field is progressing ^_^
We just kick-started our first workshop of the year, the CECAM Flagship workshop βA Roadmap for an Atomistic Machine Learning Software Ecosystemβ - laptops open, ideas humming, and the groove has begun.
www.cecam.org/workshop-det...
Screenshot of the matbench discovery leaderboard as of 14.01.2026, showing a PET based model in the top position
Not going to make a big deal out of a benchmark table, but PET just got the top spot on matbench-discovery.materialsproject.org. And don't be fooled by the huge parameters count, it's faster and can handle larger structures than eSEN-30M π. Kudos to π§βπ Filippo, Arslan and Paolo!
The Iranian regime has shut down the Internet. They do this every time there are mass protests as a prelude to massacre. That the international media is not properly covering all this is a disgrace. Don't look away. #Iran
We don't just give those freedoms for granted, we are giving them away because we are angry with water bottles with the cap attached, and school canteens offering a halal option.
Join the next IMX talk
ποΈ Thursday, 8 January 2026 at 15:00
ποΈ Dr. Christian Leinenbach, EMPA and EPFL
π From Powder to Performance: Microstructure Engineering in Metal Additive Manufacturing for Advanced Applications
Find out more: memento.epfl.ch/event/imx-ta...
Zooming in on a large-scale dataset to showcase the new adaptive resolution features in chemiscope
π’ chemiscope.org 1.0.0rc1 just dropped on pypi! We are making (a few) breaking changes to the interfaces, fixing a ton of bugs and introducing some exciting features (you can finally load datasets with > 100k points!). We'd be grateful if you test, break and report π github.com/lab-cosmo/ch...
The ACS system had a broken "alternative reviewers suggestion form" for like six months. Help me help you, for gawd's sake.
Playing around with humidity sensitive salts with the kiddos. Pink/blue is CoCl2, green/brown is CuCl2, light blue Cu(OH)2. Other ideas? Possibly not too toxic, and possibly that I can make with scrap metals, HCl, H2O2 and ammonia.
BTW same applies to the brilliant foresight of cutting the budget of the @snsf.ch, one of the best-run national science agencies I've ever interacted with, and one that has provided immense return on investment making π¨π a hub for innovation.
Teachers have been striking two days a week for a month in Vaud. As a parent, it's a pain, but they're right to strike. Cutting public services that work well when there is no budget crisis and no debt is idiotic and self-damaging. Taxes are so effing low in π¨π anyway!
π Congratulations to Prof. Dirk Grundler, head of the Laboratory of Nanoscale Magnetic Materials and Magnonics (LMGN) on his promotion to Full Professor of Materials Science in the School of Engineering!
More info: actu.epfl.ch/news/nominat...
Amazing news, well done!
Congrats to π§βπ Sergey Pozdnyakov who received a distinction (best 8% of theses at @materials-epfl.bsky.social) for his PhD thesis "Advancing understanding and practical performance of machine learning interatomic potentials". ΠΠΎΡΡ Π°Π»ΠΈ π! infoscience.epfl.ch/entities/pub...
Features reconstruction errors between the latent spaces of several universal MLIPs
No day goes by without a new universal #ML potential. But how different they really are? Sanggyu and Sofiia tried to give a quantitative answer by comparing the reconstruction errors between their latent-space features. If you are curious, check out the #preprint arxiv.org/html/2512.05...
