Ziheng Lu retweeted
Added Prophet to Ouro yesterday to start testing and benchmarking it.
I've been waiting a long time for a challenger to CHGNet.
Now the community's rare-earth-free permanent magnet discovery research gets better magnetic moments, more accurate Curie temperatures, and can distinguish the energies of different magnetic states of the same structure, something that used to require DFT. Prophet does it all in one model.
It's not perfect yet (it prefers antiferromagnetism in Mn pnictides like MnBi) but a step in the right direction. Our first benchmarks are up on Ouro.
Try it out and see what you think.
Check out our new model.
Now you can work on magnetic materials like what you do with DFT, input any structure, any spin, converge, and get the results!
github.com/kairosmaterial/pr…
ELEMENTA is a large scale DFT dataset for UMLIP and “foundation model” training. The coolest thing we found is the sampling on the spin degree of freedom. This has long been neglected but turned out to be so important!
Kairos Materials releases ELEMENTA, a large-scale materials dataset featuring 210M DFT-labeled configurations, 9.7M+ polymorphs, 1.9M chemical compositions and 4.7M+ spin samples.
huggingface.co/datasets/kair…
github.com/kairosmaterial/EL…
kairosmaterials.com/papers/E…
#AI4Science #AI4Materials
some personal update: I left @MSFTResearch after 3 incredible years (where I met the best group of people!) @KairosMaterials is our new venture. We build cool AI, do cool simulations/inferences, and A LOT of cool experiments here. Many HCs open, globally.
Kairos Materials is sponsoring #ICML 2026 AI for science workshop @AI_for_Science, and we're honored to present the Dataset Competition award.
Our co-founder @YMiko8315 and ML specialist will be there.
We‘re hiring! kairosmaterials.com/contact/
Visit our booth!
Ziheng Lu retweeted
Excited to announce major #MatterSim updates!
👩🔬 Experimentally synthesized high thermal conductor identified by MatterSim
⚡️ 3-5x inference speed-up
💪 MatterSim-MT: a new multi-task foundation model for in silico materials characterization
⬇️ Details below (1/6)
Ziheng Lu retweeted
Thermal conductivity is critical in modern electronics, but in a post-Moore’s Law world, the need for novel structures that surpass the heat transfer properties of silicon is essential. Learn how AI is helping scientists discover these next-gen materials. msft.it/6013SnPAN
ALT Illustrated image of Hongxia Hao and Bing Lv. "Abstracts: A Microsoft Research Podcast” runs along the bottom.
New work in collab w/ great @nanophononics - Saying “hey, stop searching for that material, it does not exist!” is often harder than finding the material itself. With AI and some careful constraints, we try to probe the upper limit of heat transfer in matter
Using @MSFTResearch MatterSim model, we have explored the upper limits of bulk materials' thermal conductivity. While we found several highly conductive materials, none has a thermal conductivity as high as diamond. @ZNanotheory @luzihen @HongxiaHao
arxiv.org/abs/2503.11568
Checkout MatterSim, TamGen, Aurora on Azure AI Foundry!
We’re taking our latest AI research breakthroughs and putting them in the hands of devs everywhere, with Azure AI Foundry Labs. azure.microsoft.com/en-us/bl…
Ziheng Lu retweeted
The BioEmu-1 model and inference code are now public under MIT license!!!
Please go ahead, play with it and let us know if there are issues.
github.com/microsoft/bioemu
Super excited to preprint our work on developing a Biomolecular Emulator (BioEmu): Scalable emulation of protein equilibrium ensembles with generative deep learning from @MSFTResearch AI for Science.
#ML #AI #NeuralNetworks #Biology #AI4Science
biorxiv.org/content/10.1101/…
MatterGen is now in Nature! Try it out from the open source repo!
Microsoft researchers introduce MatterGen, a model that can discover new materials tailored to specific needs—like efficient solar cells or CO2 recycling—advancing progress beyond trial-and-error experiments. msft.it/6012U8zX8
Ziheng Lu retweeted
2024年もマテリアルズインフォマティクスの論文を読んできましたが、今年は汎用機械学習ポテンシャルの公開モデルが進展した年だったなと思います。
MetaさんのEquiformerV2やMicrosoftさんのMatterSimなど、様々な高精度モデルが無料で簡単に使えるすごい時代になりました。
matbench-discovery.materials…
Ziheng Lu retweeted
Very excited to talk about
#MatterGen and #MatterSim (on GitHub now!)
at the @ictpnews Total Energy and Force Methods Workshop in Trieste at 17 today!
Come if you're interested in generative models for materials and machine learning force fields :)
#compchem #ai #materials
Ziheng Lu retweeted
Super excited to preprint our work on developing a Biomolecular Emulator (BioEmu): Scalable emulation of protein equilibrium ensembles with generative deep learning from @MSFTResearch AI for Science.
#ML #AI #NeuralNetworks #Biology #AI4Science
biorxiv.org/content/10.1101/…
Yes, finally got MatterSim out for our community to use. Please checkout the codes (and weights!), have fun with it, interface it with your awesome pipelines/softwares, and accelerate your cool research!
GitHub repo: github.com/microsoft/matters…
Preprint: arxiv.org/abs/2405.04967
We're thrilled to announce the release of MatterSimV1-1M and MatterSimV1-5M on GitHub. These cutting-edge models are now available for researchers, developers, and innovators to explore, customize and build upon. msft.it/6013oBZLt
ALT Logo for MatterSim on a blue to green gradient background. The logo features a white molecular structure with six nodes connected by lines, positioned to the left of the text ‘MatterSim’ in white.
super cool group, extremely productive, dream postdoc position, join now!
PLEASE RETWEET: Come and spend 3.5 years as a research fellow in my group at @UoBChemistry researching wide band gap oxides, using @uob_bear.
jobs.ac.uk/job/DJH373/resear…
Turned out off-equilibrium structures are helpful to predict stable materials.
Great work from the team @Jielanli @xie_tian @zany_cloud @rpinsler and many more.
Exciting news! MatterSim has officialy claimed the #1 spot on the #MatBench Discovery leaderboard (matbench-discovery.materials…).
MatterSim is a SOTA #machinelearning force field for materials simulation, discovery and more (arxiv.org/abs/2405.04967).@luzihen
#AI4Science @MSFTResearch
There's a new version of this post
Put your favorite crystals in a pressure cooker, heat it up, and let your DL model learn a “universal” materials representation for broad tasks! Checkout MatterSim:
arxiv.org/abs/2405.04967
Excited to share our latest development, MatterSim: A Deep Learning Atomistic Model Across Elements, Temperatures and Pressures. Check it out 👉arxiv.org/abs/2405.04967 microsoft.com/en-us/research… #MicrosoftResearch #AI #DeepLearning #MaterialScience
@luzihen