Our deep learned exchange-correlation functional Skala is finally available to try out 🎉 Tell us what works and where skala can be improved!

Interesting study around scaling and equivariance arxiv.org/abs/2510.09768

We asked the same question: how can we combine the strengths of continuous and discrete approaches? Similar to CDCD, in our work, Purrception, we extend Variational FM to model VQ latents through continuous-discrete transport for image generation :D 👉 arxiv.org/abs/2510.01478

Presenting our second work in ‘Equivariance vs scale’ discussion with 🔹 Platonic Transformers. Safe to say, Equivariance is not dead. 😄 To learn more check this 👇

What if you could equip transformers with symmetry without slowing them down? 💡 Our new paper extends RoPE transformers to symmetry groups. The key breakthrough: It achieves this with zero computational overhead, improving performance even on non-equivariant tasks.

🚨 New preprint! Love using Transformers and scaling? Love equivariance and inductive bias? They needn’t be at odds!! Introducing the Platonic Transformer, a geometric Transformer framework that hijacks RoPE to achieve E(3) equivariance at no additional cost ⚡️ We are as fast

I will be presenting this at MIT Jameel Clinic for AI & Health’s #MoML2025 this Wednesday, 22nd Oct. Come by our poster to chat about how you can incorporate PlatoFormers into your pipeline TODAY

MIT Jameel Clinic for AI & Health I think it's absolutely remarkable that we can achieve E(3) equivariance without any gimmicky modules or expensive operations commonly seen in their geometric GNN counterparts We hope inductive bias can finally enter its scaling era 🔥

Recent discussions have largely focused on scaling versus geometry. Another perfect example showcasing that geometry could be made scalable, if we as GDL people start to take scaling seriously. Lets take the best of both worlds 🦾🦾

Time to scale up equivariant architectures!❤️Absolutely wonderful work Max! I wouldn't go as far as saying efficiency is everything, bc equivariance is ;), but for sure this seems like an indisputably winning combo! Scale+equiv ftw imo😁 & great to see Triton paying off like this

✨CAMERA READY UPDATE✨ with new cool plots in which we show how we can use our Equivariant Neural Eikonal Solver for path planning in Riemannian manifolds Check our paper here arxiv.org/pdf/2505.16035 And see you at NeurIPS 🥰