AI RESEARCH
BiMol-Diff: A Unified Diffusion Framework for Molecular Generation and Captioning
arXiv CS.CL
•
ArXi:2604.24089v1 Announce Type: new Bridging molecular structures and natural language is essential for controllable design. Autoregressive models struggle with long-range dependencies, while standard diffusion processes apply uniform corruption across positions, which can distort structurally informative tokens. We present BiMol-Diff, a unified diffusion framework for the paired tasks of text-conditioned molecule generation and molecule captioning.