MOSAIC:多模态流匹配模型实现分子结构与光谱任意互转
Every Measurement, Every Direction, All at Once: Multimodal Flow Matching for Molecules and Spectra
一篇面向科学生成建模的论文:MOSAIC 用随机掩码让分子结构和 UV/IR/Raman 光谱互为条件与目标,一个模型跑通 any-to-any 生成,QM9S 上结构恢复率 88.8%,比最强基线高 23.2%。做分子设计、光谱预测或数据稀缺场景生成建模的研究者值得看它的掩码训练思路。
论文提出 MOSAIC,一个覆盖分子 3D 结构与 UV、IR、Raman 光谱的多模态流匹配 transformer,可在 QM9S 数据集上达到 88.8% 的分子图恢复率,比最强光谱到结构基线在结构生成上高出 23.2%,同时在光谱生成上超过专用的 structure-to-spectra 模型。其核心机制是把单次实验产生的多种耦合测量视为统一训练信号:通过随机掩码,任一模态既可作为条件输入也可作为预测目标,从而从一次实验中派生出多个监督去噪任务,避免了传统生成模型只用单一方向的监督浪费。由此得到单一 any-to-any 模型,支持从任意模态子集生成其余不相交模态子集。
消融实验显示,any-to-any 训练目标在多个模型规模上都优于仅用结构训练的版本。模型还能处理部分可用的测量子集,并通过把生成光谱与实验参考谱匹配来实现分子鉴定。对实验数据昂贵、每条数据含多类测量的科学生成建模场景,这种把耦合测量全部利用起来的思路提供了更高的数据效率。
Every Measurement, Every Direction, All at Once: Multimodal Flow Matching for Molecules and Spectra
Generative modeling in science is often data-limited, where experiments are expensive. However, each experiment may yield several coupled measurements of the same physical entity. Standard generative models often train only on a single direction, discarding supervision from the rest. We argue that each experiment contains more training signal than this asymmetry allows the model to exploit: random masking enables measurements to serve as either conditioning inputs or prediction targets, yielding multiple supervised denoising tasks from a single experiment. We investigate this premise with MOSAIC, a multimodal flow matching transformer over molecular 3D structure and UV, IR, and Raman spectra. MOSAIC is a single model that is capable of any-to-any generation: generating any subset of modalities from any disjoint subset of the others. On QM9S, MOSAIC achieves 88.8% molecular graph recovery, outperforming the strongest spectra-to-structure baseline by 23.2% in structure generation, while simultaneously outperforming a structure-to-spectra specialist model in spectral generation. Ablations show that the any-to-any objective improves structure generation over structure-only training across model sizes. Furthermore, MOSAIC supports arbitrary subsets of available measurements and enables molecular identification by matching generated spectra to experimental references.
This report is categorized under AITOP's "Papers" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include MOSAIC, flow matching, multimodal, QM9S.
Source attribution: curated from arXiv cs.LG (@Thorben Prein, Elton Pan, Rafael Gómez-Bombarelli, Santiago Miret).