NASA has open sourced its proven AI system ExoMiner++ as it scans TESS data for new worlds, enabling global researchers to accelerate exoplanet discovery through transparent, collaborative science.
NASA has open sourced ExoMiner++, its deep learning–based artificial intelligence system now analysing data from the Transiting Exoplanet Survey Satellite (TESS) to identify new exoplanets, marking a major step towards transparent and collaborative space science.
ExoMiner++ has already demonstrated its effectiveness by validating 370 exoplanets in earlier missions, establishing its reliability in separating genuine planetary signals from false positives. In its initial analysis of TESS data, the AI flagged approximately 7000 potential exoplanet candidates, each requiring follow-up observations for confirmation.
Designed to automate one of the most time-intensive aspects of planet hunting, ExoMiner++ significantly accelerates exoplanet discovery by reviewing thousands of signals quickly and accurately. The system uses deep learning to analyse stellar light curves, detecting the subtle dips in brightness that occur when a planet transits its host star.
Crucially, ExoMiner++ distinguishes real planets from misleading signals such as eclipsing binary stars and other astrophysical noise, improving the efficiency and accuracy of NASA’s detection pipeline.
By making ExoMiner++ open-source, NASA enables researchers, astronomers, and developers worldwide to study, audit, and build upon the model. This open access allows the global scientific community to replicate results, refine detection algorithms, and adapt the technology for other astronomical datasets, strengthening collaborative planetary science and expanding understanding of distant planetary systems.
