Excited to share the new NetSurfP 3, an ultra-fast webserver for protein solvent accessibility, disorder and secondary structure prediction. We apply a protein language model for alignment free predictions up to 600x the speed of previously available tools doi.org/10.1093/nar/gk…

Excited to share our new work (led by Tobias Höppe) on diffusion models for videos! 📽️ We train conditional diffusion models for arbitrary video completion tasks, achieving SOTA on video prediction. paper: arxiv.org/abs/2206.07696 website: sites.google.com/view/video-dif… 1/6

Looking for a postdoc? Interested in deep learning methods development and solving difficult problems in enzyme function prediction? Want to part part of mlls.dk and aicentre.dk? Then consider applying here dtu.dk/om-dtu/job-og-…

Can large language models reason about medical questions? We investigated whether GPT-3.5 could answer and reason about challenging medical questions (e.g., USMLE) using Chain-of-Thought (CoT) prompting. (1/6) TL;DR: Yes, close to the human level. Paper: arxiv.org/abs/2207.08143

Do you like machine learning, proteins and Copenhagen? Check out our open PhD position: jobportal.ku.dk/phd/?show=1581…. Deadline Jan 22. Please RT.

My work on large language models appears in the Danish scientific media! We showed that GPT-3/Codex virtually passes the USMLE (60.2%), but a large gap remains with experienced doctors. Nonetheless, I believe next-gen AI will play a major role in supporting future doctors.

I'm very happy to present DiscoTope-3.0 - Improved B-cell epitope prediction using AlphaFold2 modeling and inverse folding latent representations Paper: biorxiv.org/content/10.110… Web server: services.healthtech.dtu.dk/services/Disco…

GeoConfirmed And now with the TASS statements on charges?

🔥 New preprint: “DiffEnc: Variational Diffusion with a Learned Encoder” We extend diffusion models by generalizing the diffusion process with a learned drift parameterized by a time-dependent NN encoder. This achieves SOTA likelihood on CIFAR ✨ 📜 arxiv.org/abs/2310.19789 1/3

Ok, it’s time for a deeper explainer of our paper on generalizing diffusion models. By the way, if you're heading to New Orleans for NeurIPS, we're looking forward to discussing at our poster during the Diffusion Models Workshop! diffusionworkshop.github.io Here we go 🧵 1/

How well do🧬 DNA LMs perform on prediction tasks that are biologically relevant, are naturally very imbalanced and require reasoning over long contexts? We present BEND, a benchmark for DNA LMs on the human genome arxiv.org/abs/2311.12570

Happy to announce that our paper BEND (Benchmarking DNA Language Models on Biologically Meaningful Tasks) has been accepted for ICLR 2024. See you in Vienna!

Our paper on extending diffusion models by learning an encoder that parameterizes the forward diffusion drift was accepted at #ICLR2024 🎉 Congrats Beatrix M. G. Nielsen and team! Stay tuned for new experiments incl. SOTA on ImageNet ✨ Looking forward to many discussions in Vienna! 🇦🇹

If you’re at ICLR, come by our poster this afternoon at 4.30 pm in Halle B (poster number 50) and let’s chat about diffusion!

We're at #ICLR2024 presenting BEND in the poster session this afternoon - come chat with us about DNA LMs! Frederikke M @ ICLR2024

Elon Musk deleted this. You know what to do.