view article
Figure 4
A Lab-in-the-Loop approach combining AI and expert input for iterative machine learning. The operation of recursive Lab-in-the-Loop cycles for iterative ML is presented. The AXIS-foundation model is applied by CRIMS to assign crystal probabilities to all crystallization images produced at the local facility. AXIS scores along with the images are presented to expert crystallographers using the facility, via CRIMS web interfaces, and they can introduce their own annotations. Discrepancies between ML and human scores are automatically collected, curated and assembled into a new training dataset to produce improved AI models. This cycle can be operated as many times as necessary to achieve optimal performance and continuous adaptation to the changing conditions at any particular facility.

IUCrJ
Volume 13| Part 3| May 2026| Pages 235-248
ISSN: 2052-2525
https://doi.org/10.1107/S2052252526001399