Deep Scientific Reasoning under Physical Constraints: Structure-Aware Spectrum Prediction

Yingheng Wang, Tao Yu, Shufeng Kong, Francesco Ricci, John M Gregoire, Carla P Gomes
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:131963-131988, 2026.

Abstract

Structured scientific spectra encode rich physical information while obeying hard constraints, such as conservation laws and characteristic spectral geometry. Accurate prediction of these spectra is central to materials discovery, yet existing methods often treat them as unconstrained sequences and therefore fail to enforce the underlying physical structure. Taking electronic density of states (eDOS) as the prototypical example, we introduce DeepSciReasoner, a general paradigm for predicting scientific spectra under physical constraints. The framework combines structure-aware spectrum decoding with constraint-preserving physical reasoning, allowing predictions to capture rich spectral structure while respecting the underlying physics. We evaluate DeepSciReasoner on eDOS, phonon density of states (phDOS), X-ray absorption near-edge structure (XANES), and Raman spectra, where it substantially improves prediction accuracy while maintaining physical consistency. These results establish DeepSciReasoner as a reusable blueprint for structured scientific spectrum prediction under hard physical constraints.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-wang26ka, title = {Deep Scientific Reasoning under Physical Constraints: Structure-Aware Spectrum Prediction}, author = {Wang, Yingheng and Yu, Tao and Kong, Shufeng and Ricci, Francesco and Gregoire, John M and Gomes, Carla P}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {131963--131988}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/wang26ka/wang26ka.pdf}, url = {https://proceedings.mlr.press/v306/wang26ka.html}, abstract = {Structured scientific spectra encode rich physical information while obeying hard constraints, such as conservation laws and characteristic spectral geometry. Accurate prediction of these spectra is central to materials discovery, yet existing methods often treat them as unconstrained sequences and therefore fail to enforce the underlying physical structure. Taking electronic density of states (eDOS) as the prototypical example, we introduce DeepSciReasoner, a general paradigm for predicting scientific spectra under physical constraints. The framework combines structure-aware spectrum decoding with constraint-preserving physical reasoning, allowing predictions to capture rich spectral structure while respecting the underlying physics. We evaluate DeepSciReasoner on eDOS, phonon density of states (phDOS), X-ray absorption near-edge structure (XANES), and Raman spectra, where it substantially improves prediction accuracy while maintaining physical consistency. These results establish DeepSciReasoner as a reusable blueprint for structured scientific spectrum prediction under hard physical constraints.} }
Endnote
%0 Conference Paper %T Deep Scientific Reasoning under Physical Constraints: Structure-Aware Spectrum Prediction %A Yingheng Wang %A Tao Yu %A Shufeng Kong %A Francesco Ricci %A John M Gregoire %A Carla P Gomes %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-wang26ka %I PMLR %P 131963--131988 %U https://proceedings.mlr.press/v306/wang26ka.html %V 306 %X Structured scientific spectra encode rich physical information while obeying hard constraints, such as conservation laws and characteristic spectral geometry. Accurate prediction of these spectra is central to materials discovery, yet existing methods often treat them as unconstrained sequences and therefore fail to enforce the underlying physical structure. Taking electronic density of states (eDOS) as the prototypical example, we introduce DeepSciReasoner, a general paradigm for predicting scientific spectra under physical constraints. The framework combines structure-aware spectrum decoding with constraint-preserving physical reasoning, allowing predictions to capture rich spectral structure while respecting the underlying physics. We evaluate DeepSciReasoner on eDOS, phonon density of states (phDOS), X-ray absorption near-edge structure (XANES), and Raman spectra, where it substantially improves prediction accuracy while maintaining physical consistency. These results establish DeepSciReasoner as a reusable blueprint for structured scientific spectrum prediction under hard physical constraints.
APA
Wang, Y., Yu, T., Kong, S., Ricci, F., Gregoire, J.M. & Gomes, C.P.. (2026). Deep Scientific Reasoning under Physical Constraints: Structure-Aware Spectrum Prediction. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:131963-131988 Available from https://proceedings.mlr.press/v306/wang26ka.html.

Related Material