Direct discovery · Chemistry · Materials science
CRESt discovers an eight-element catalyst through multimodal active learning and robotic tests
An experimentally tested eight-element formate-oxidation catalyst with improved cost-specific performance.
Summary
CRESt explored more than 900 catalyst chemistries and conducted 3,500 electrochemical tests in three months. The campaign identified a catalyst in the Pd–Pt–Cu–Au–Ir–Ce–Nb–Cr space with a reported 9.3-fold improvement in cost-specific performance over palladium. Human feedback and camera-based anomaly monitoring remained part of the workflow.
AI role
Multimodal models combined literature, compositions and microscopy with Bayesian optimization to select robotic synthesis and testing experiments.
Narrative role
Fills a 2025 physical-discovery gap with a measured material outcome from an integrated experimental loop, extending coverage beyond computational screening and generic laboratory launches.
Caveat
The 9.3-fold figure is cost-specific catalyst performance, not research speed. The result concerns formate oxidation and does not establish commercial durability or a causal AI effect on scientific productivity. Full main-text access was limited.