Dr. St❦fan Gugler

Normalize academics talking about their failures. I'll go first: We got rejected at Berghain on Saturday night.

@eigengenesis you learned on n_shells (principal quantum number), block (orbital angular mom), and valence number (electron config). how can you claim you told it nothing about electrons, orbitals and QM when you told it basically everything ?

i trained my model on the first 86 elements of the periodic table and told it nothing about electrons, orbitals, or quantum mechanics. it rediscovered the structure of the periodic table and predicted the physical properties of unseen elements 87-118 with <10% error.

@pmddomingos There is some prelim evidence that force fields may achieve sota performance without explicitly encoding equivariance or energy conservation. But yeah, exploiting sym is still a great bet. arxiv.org/abs/2503.01431 arxiv.org/abs/2412.11569

The #1 problem in AI is poor generalization, and the #1 solution is exploiting symmetries.

I'll be at Pacifichem this week, hit me up if you wanna have a coffee! ☕

I'm in San Diego this week for NeurIPS, hit me up if you wanna have a coffee ☕

@MarkNeumannnn Me too!

I'll be in San Diego this week for Neurips - let me know if you want to meet up at the conference! Would be happy to chat about forcefields, molecular generative models, inverse folding, mutational scanning etc etc.

I'm in Oslo this week for WATOC, hit me up if you wanna have a coffee ☕

@no_earthquake The true story of _italics_ c o v i d 1 9

how is this real

@marceldotsci stop him for his own good 🥲

can stefan be stopped? it’s unclear. congrats!

@MarioKrenn6240 @uni_tue Super cool, congratulations Mario :)

Very happy to announce that I'll be joining the University of Tübingen (@uni_tue) as a Full Professor (W3) of "Machine Learning for Science" in the Computer Science Department within the Faculty of Science and the #ExcellenceCluster ML4Science. Tübingen is an exceptionally vibrant place for AI/ML research with many many great groups and initiatives, and with strong applications in the natural sciences. Couldn't be happier; can't wait to start and meet all future colleagues, and building new and more powerful AI tools for understanding the universe. We have numerous open positions (also within my @ERC_Research StG), for post-docs and PhDs. If you are excited to build AI-driven discovery of physics experiments for new ways to observe the universe (super-fast physics simulators, AI-driven exploration of large complex spaces, etc) and agent-based idea generations, send me an e-Mail.

Post on previous work: x.com/stevain/status… and x.com/stevain/status…

9/ Check it yourself: 🔗: github.com/khaledkah/tv-s… 📄: arxiv.org/abs/2502.08598 Thanks to Khaled, Winnie, Oliver, Klaus, and Shin for the cool collaboration as well as @bifoldberlin @ml_tuberlin @TUBerlin @RIKEN_AIP_EN @RIKEN_AIP

We have a new paper on diffusion!📄 Faster diffusion models with total variance/signal-to-noise ratio disentanglement! ⚡️ Our new work shows how to generate stable molecules in sometimes as little 8 steps and match EDM’s image quality with a uniform time grid. 🧵