Tag
The paper introduces Large Discovery Model (LDM), a recurrent architecture that couples generative models with Bayesian non-parametric surrogates to guide uncertainty-aware search in scientific domains like molecules and proteins, achieving significant performance gains over existing methods.