ExoMiner++ expands automated validation of TESS planet candidates
NASA’s ExoMiner++ uses deep learning to help validate exoplanet candidates in data from NASA’s planet-hunting missions. Automated screening can prioritize review, while confirmation still depends on scientific checks and supporting observations.
Full Disclosure Administrator · Published August 7, 2025 · Updated October 17, 2026NASA’s 2026 science coverage describes ExoMiner++, an open-source deep-learning model trained on data from exoplanet-hunting missions and applied to TESS observations. The tool is designed to help assess whether signals in telescope data are likely to be real planets or false positives. It builds on earlier ExoMiner work that researchers used to validate hundreds of exoplanets.
Automated validation can help scientists triage large candidate lists, but a model’s output is not the same as an independent observation. Transit-like signals can be produced by eclipsing binary stars, instrumental effects, or other astrophysical sources. Researchers still need to examine model performance, training data, uncertainty, and follow-up evidence before treating a candidate as confirmed.
The broader lesson is that AI can make large research archives easier to search while leaving scientific responsibility with the researchers. A transparent workflow should preserve the candidate’s data provenance, provide a way to inspect why the model ranked it, and document how false positives were handled.
Source: NASA Artificial Intelligence research coverage, including ExoMiner++.