COSMO Lab

Cluster highlights in chemiscope 1.0 RC3

Release candidate 3 of chemiscope 1.0 is out, with class and range based highlighting of points. Try it, break it, report it on github.com/lab-cosmo/ch...

A new national research programme recognizes EPFL's expertise The Swiss Confederation launches six new National Centres of Competence in Research (NCCRs). The NCCR “Separations”, which aims to accelerate research in separation sciences - the quest for chemical a...

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...

Pareto front for PET-OMAT models

If you're scared by the 700M parameters (you shouldn't be) there's a whole set of models from 🐁 to 🦣. You can find them all on github.com/lab-cosmo/upet !

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.

You can fetch the model here github.com/lab-cosmo/upet, as easy as `pip install upet`, and then, for the ASE interface, `from upet.calculator import UPETCalculator;
calculator = UPETCalculator(model="pet-oam-xl", version="1.0.0", device="cuda")` Have fun and go break it!

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 🚀. Kudos to 🧑‍🚀 Filippo, Arslan and Paolo!

Zooming in on a large-scale dataset to showcase the new adaptive resolution features in chemiscope

📢 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 🐛 github.com/lab-cosmo/ch...

A cheeshire cat sitting on a tree overlooking a winter landscape, with snapshots of the many classes of materials that the PET-MAD universal interatomic potential can be used for

Hope y'all are getting a great start of 2026. Here we're taking some time to add the 2025 winter card to the archives www.epfl.ch/labs/cosmo/i... 🎅=🧑‍🚀

Ah, a fine example of a `for regressor in sklearn.supervised_learning:` paper. It's an underappreciated genre.

example of a streamlit app integrating a chemiscope viewer

📢 New chemiscope.org release just landed! To make it even easier to integrate ⚗️🔭 into your workflow, we added a @streamlit.bsky.social component, so you can run analyses and show you atomistic data in a web app by just writing a few lines of python! try it, break it, report it!

Congrats to 🧑‍🚀 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". Поїхали 🚀! 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...

Oh. My. Gawd. ☝️☝️☝️☝️ 👀

PET-MAD as a lightweight universal interatomic potential for advanced materials modeling - Nature Communications PET-MAD is a fast and lightweight universal machine-learning potential, trained on a small but diverse dataset, that delivers near-quantum accuracy in atomistic simulations for both organic and inorga...

📢 PET-MAD is here! 📢 It has been for a while for those who read the #arXiv, but now you get it preciously 💸 typeset by @natcomms.nature.com Take home: unconstrained architecture + good train set choices give you fast, accurate and stable universal MLIP that just works™️ www.nature.com/articles/s41...

Developed in collaboration with the THEOS group with support from the @nccr-marvel.bsky.social, and stored on the #materialscloud here archive.materialscloud.org/records/c4en..., you can also take a look at the data as a #chemiscope 🔭⚛️ chemiscope.materialscloud.io?load=https%3...

Massive Atomic Diversity: a compact universal dataset for atomistic machine learning - Scientific Data Scientific Data - Massive Atomic Diversity: a compact universal dataset for atomistic machine learning

📢 Let us (re)introduce to you our Massive Atomic Diversity dataset for universal MLIPs. MAD includes molecules, clusters, surfaces and plenty of bulk configs, we cover a lot of ground with fewer than 100k structures, using highly consistent DFT settings. Read more 📑 www.nature.com/articles/s41...

Machine learning for atomic-scale simulations: balancing speed and physical laws - ΑΙhub

‪In this blog post, Filippo Bigi, Marcel Langer (@labcosmo.bsky.social‬) and @micheleceriotti.bsky.social write about the need to balance speed and physical laws when using ML for atomic-scale simulations
aihub.org/2025/10/10/m...

Machine learning for atomic-scale simulations: balancing speed and physical laws - ΑΙhub

A primer for non conservative (& rotationally unconstrained) MLIPs, and how to use them safely. Thanks @aihub.org for the space! aihub.org/2025/10/10/m...

Learning Inter-Atomic Potentials without Explicit Equivariance Accurate and scalable machine-learned inter-atomic potentials (MLIPs) are essential for molecular simulations ranging from drug discovery to new material design. Current state-of-the-art models enforc...

Looks like @ox.ac.uk forbids their researchers to do any kind of literature search, though it seems that thankfully they can still submit to the arxiv arxiv.org/abs/2510.00027 🤷

However, this seems to damage the transferability of highly-preconditioned models such as MACE - less so for more expressive unconstrained models such as PET. Does this match your experience?

This doesn't matter much as most of the fragments that make up the body-order decomposition as deranged soups of highly-correlated electrons. Models with sufficient expressive power *can* learn if presented with the fragments ...

TL;DR: not really. ML potentials learn whatever they want, as long as it allows them good accuracy on the train set. We note in particular that MACE is strongly preconditioned to learn a fast-decaying body-order expansion, whether it decays fast or not.

📝 We have been told (& been telling) that ML potentials are linked quite directly to the expansion of the atomic energy into pairs, triples, and so on. But is this actually true 🤔? Go read the latest from the 🧑‍🚀 team (w/QM help from Joonho's team at Harvard) to find out more arxiv.org/html/2509.14...

GitHub - lab-cosmo/flashmd: A universal ML model to predict molecular dynamics trajectories with long time steps A universal ML model to predict molecular dynamics trajectories with long time steps - lab-cosmo/flashmd

Bragging time - ⚡ FlashMD⚡ was accepted as a spotlight paper at #NeurIPS25. if you still haven't checked it out, it's already on the #arxiv arxiv.org/abs/2505.19350, the code is at flashmd.org and the 🧑‍🍳📖 is here atomistic-cookbook.org/examples/fla.... Congrats to Filippo, Sanggyu and Augustinus!

Michele Parrinello giving the ICTP Colloquium (he speaks about catalysis) as part of the conference celebrating his 80th birthday. Amazing creativity throughout a long career!

An Adapted Similarity Kernel and Generalised Convex Hull for Molecular Crystal Structure Prediction We adapted an existing approach to identifying stabilisable crystal structures from prediction sets - the Generalised Convex Hull (GCH) - to improve its application to molecular crystal structures. Th...

I'm very pleased to say my first preprint, with @graemeday.bsky.social and @micheleceriotti.bsky.social is now online!

This is the main work of my PhD, adapting a similarity kernel to be more suited for exploring molecular CSP landscapes

#compchemsky #chemsky #compchem

doi.org/10.26434/che...

error plots for the PET-MAD-DOS model on different datasets

Anticipating 🧑‍🚀 Wei Bin's talk at #psik2025 (noon@roomA), 📢 a new #preprint using PET and the MAD dataset to train a universal #ml model for the density of states, giving band gaps for solids, clusters, surfaces and molecules with MAE ~200meV. Go to the talk, or check out arxiv.org/html/2508.17...!

With funding from a @snf-fns.ch Sinergia, the @nccr-marvel.bsky.social and @erc.europa.eu, and computing time from @cscsch.bsky.social !

Reaction energies of pristine and reconstructed surfaces with water

The reconstructed surface contains different sites with different reactivity. Despite the higher stability, for some sites the disordered surface is *more* reactive with water, one of the main contaminants affecting the stability of LPS batteries. Useful to design better stabilization strategies!

Surface energy diagram of LPS before and after reconstruction

Wulff shape of LPS particles based on the computed surface energies

Reconstructed surfaces become lower in energy, and the surface energy less orientation dependent - and so the Wulff shape of particles become more spherical.