MIT's Crystal-CLIP Fuses 3D Atomic Structure With Text, Outperforms Text-Only Baselines on Materials Discovery

MIT researchers unveiled a multimodal model that fuses 3D crystalline structures with natural-language descriptions, outperforming text-only baselines when identifying candidate materials. Ju Li framed the result as 'structure is really at the heart of the materials science field'; the team ran it on solid-state lithium-ion electrolytes and recovered known top candidates, suggesting applicability across the >90% of inorganic solids that are crystalline.