A self-learning scientific agent for X-ray diffraction
Published in arXiv preprint, 2026
GanJiang is a self-learning scientific agent for powder X-ray diffraction, built on XMatcher, XQueryer, XDecomposer and WPEM. It turns analytical experience into reusable skills by diagnosing failures, revising instructions and code, and validating changes before reuse, without retraining the language model or altering the physical models. The system supports phase identification, multiphase decomposition and physics-constrained whole-pattern modelling. Evaluations demonstrate improved refinement across FullProf, GSAS-II and PyWPEM, alongside leading results among evaluated methods on DeltaXRDbench.
Recommended citation: Cao, B., Zhou, H., Yang, R., Li, J., Sun, S., Song, Y., Gao, H., Yu, Z., Zhang, T.-Y., & Wang, J. (2026). A self-learning scientific agent for X-ray diffraction. arXiv preprint arXiv:2610.07862.
Download Paper