MM-Spectrum: Multimodal Multi-spectral Molecular Structural Elucidation with a Stable MoE Framework
MM-Spectrum uses modality-aware MoE routing to keep heterogeneous spectral signals from degrading structure prediction.
The paper targets molecular structural elucidation from multiple spectroscopic inputs, where simple sequence concatenation can underperform because modalities are imbalanced. Its router is given both token representations and spectral identity, so experts can specialize by modality and cross-modal interaction. The framework reports consistent gains across full-modality, bimodal, and missing-modality settings, with ablations and interpretability analysis supporting the design. Accepted to ICML 2026. ArXiv · AI/CL/LG's note
The paper targets molecular structural elucidation from multiple spectroscopic inputs, where simple sequence concatenation can underperform because modalities are imbalanced. Its router is given both token representations and spectral identity, so experts can specialize by modality and cross-modal interaction. The framework reports consistent gains across full-modality, bimodal, and missing-modality settings, with ablations and interpretability analysis supporting the design. Accepted to ICML 2026. ArXiv · AI/CL/LG's note
score 4