Natural-Language-Guided Generator-Agnostic Shortlisting for Protein Binder Design
It is suggested that LLM-generated ranking policies can act as an interpretable post-generation decision layer for combining heterogeneous proxy metrics to prioritize binders from large candidate pools.