SafeMol: Dual-Modality Safety Alignment for Molecular Multimodal Models
This work proposes SafeMol, a parameter-efficient safety alignment framework that jointly optimizes lightweight modules across text-only and graph-conditioned inputs, uses MMD for distribution-level representation alignment to reduce modality-induced discrepancies, and explicitly models molecular hazardousness and harm...
Xin-Miao Wang, Rui-Jie Wang, Meng-Hui Wang et al.
· 0 citations