Posterior Inference of Hamiltonian Parameters from RIXS Spectroscopy

The paper reports the first use of simulation‑based inference for resonant inelastic X‑ray scattering (RIXS) spectroscopy, applying truncated marginal neural ratio estimation and conditional flow matching to infer full posterior distributions of Hamiltonian parameters for two Ni$^{2+}$ compounds. A vision‑transformer encoder tailored to the RIXS map’s physical layout produces sharper, better‑covered posteriors than generic image encoders. The validated method, applied to experimental data, uncovers parameter correlations invisible to point estimators and yields posterior predictive distributions that closely match observed spectra, enabling new analyses such as nuisance‑marginalized uncertainty quantification, multi‑measurement posterior fusion, and active experimental design.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Statistics ML.

• llms
• diffusion
• computer-vision
• nlp