Jake Silberg

@jakublala My recollection is that GANs are harder to train because you are training toward a saddle point rather than a global minimum

Generative AI models, like Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs), analyze molecular structures and medical images to suggest potential drugs for effective treatment. For instance, Insilico Medicine has successfully explored the advantages of quantum GANs in generative chemistry, enhancing the efficiency and accuracy of drug design. #overview" target="_blank" rel="nofollow noopener">aws.amazon.com/startups/learn…

🚨 New paper! We introduce a planner-aware training tweak to diffusion language models. ⚡ One-line-of-code change to the loss 💡 Fixes training–inference mismatch 📈 Strong gains in protein, text, and code generation arxiv.org/abs/2509.23405 (1/n)

High-Affinity Protein Binder Design via Flow Matching and In Silico Maturation - Propose PPIFlow, a flow-matching model that achieves picomolar to nanomolar affinities across diverse targets, including 7/8 high-affinity VHHs, fully in silico - Combine Pairformer and Invariant Point Attention modules, trained through a four-stage curriculum: monomer and motif scaffolding, binder design, scFv design, VHH design - Develop an in silico maturation pipeline: (i) Identify interface residues with interaction energy < −5 REU across designed sequences (ii) Merge into a consensus set of anchor rotamers (iii) Fix anchors, apply noise (t = 0.6), and perform partial flow refinement to regenerate unconstrained backbone (iv) Redesign sequences Final candidates are filtered with pTM > 0.8 and ipTM > 0.5 with AF3score github.com/Mingchenchen/P…

Join our reading group session now about "Calibrating Generative Models to Distributional Constraints" arxiv.org/abs/2510.10020 :) On zoom: portal.valencelabs.com/starklyspeaking

Among the wonderful things about this story is that @brendanruberry interviewed DeBlasio through the Ring doorbell of his home in Huntington Station, while he was in Florida. semafor.com/article/10/29/…

Tired to go back to the original papers again and again? Our monograph: a systematic and fundamental recipe you can rely on! 📘 We’re excited to release 《The Principles of Diffusion Models》— with @DrYangSong, @gimdong58085414, @mittu1204, and @StefanoErmon. It traces the core ideas that shaped diffusion modeling and explains how today’s models work, why they work, and where they’re heading. 🧵You’ll find the link and a few highlights in the thread. We’d love to hear your thoughts and join some discussions! ⚡ Stay tuned for our markdown version, where you can drop your comments!