Steered generation for protein optimization: On datasets with ~10^2 measurements, steering a discrete diffusion model outperforms RL on a protein language model for generating improved variants.

@jsunn-y.bsky.social @francesarnold.bsky.social @yisongyue.bsky.social

arxiv.org/abs/2505.15093

Happy to share that our work on Active Learning-Assisted Directed Evolution is now published in @naturecomms.bsky.social!

Paper: www.nature.com/articles/s41...

Check out the original thread here: x.com/jsunn_y/stat...

Come visit our CARE benchmarks poster today at 11-2pm! (West Ballroom #5205)

I’m at NeurIPS this week presenting two of my recent projects!

Enzyme function classification (CARE) benchmarks: Friday 11-2pm West Ballroom #5205 @caltech.edu

Conditional generation from protein language models (ProCALM): Sunday at MLSB Workshop @Profluent

Please come say hi!