Reproduce “Robust Inference-Time Steering of Protein Diffusion Models via Embedding Optimization” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Minhuan Li, Jiequn Han, Pilar Cossio, Luhuan Wu
EmbedOpt steers pretrained protein structure models toward experimental evidence by updating conditioning embeddings rather than noisy coordinates, without retraining weights. Across 77 systems, its advantage is concentrated on harder targets. On six real cryo-EM targets, it improves map fit over DPS at the tested setting. Its guarantee is local: surrogate reward improves, but the method cannot steer beyond the model’s structural support.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.