Tag
The paper introduces MolEmb, a lightweight framework that adapts multimodal large language models for general molecular embedding, enabling context-aware representations and cross-modal retrieval, along with a diagnostic benchmark MolCAR.