NextFin News - Astronomers at the University of Warwick have utilized a sophisticated artificial intelligence pipeline to validate 118 new exoplanets, effectively mining a "hidden haul" of worlds from data previously collected by NASA’s Transiting Exoplanet Survey Satellite (TESS). The discovery, published this week in the Monthly Notices of the Royal Astronomical Society, marks a significant shift in how deep-space data is processed, moving away from manual verification toward automated, high-precision machine learning. By applying the newly developed RAVEN pipeline to observations of over 2.2 million stars, the research team identified not only the 118 confirmed planets but also more than 2,000 high-quality candidates, nearly half of which were entirely unknown to science until now.
The sheer volume of data generated by modern space telescopes has long outpaced the capacity of human researchers to analyze it. TESS, which monitors the subtle dimming of starlight caused by planets passing in front of their host stars, produces millions of light curves that contain "noise" from stellar activity or instrumental artifacts. Traditionally, distinguishing a true planetary transit from a false positive required painstaking manual vetting. The RAVEN pipeline changes this calculus by automating the search through Full Frame Images (FFIs), focusing on planets with short orbital periods of less than 16 days. This efficiency allowed the team to process four years of TESS data with a level of precision that Marina Lafarga Magro, the study’s lead author, noted resulted in uncertainties up to ten times smaller than previous methods.
Among the 118 validated worlds are several rare "ultra-short-period" planets that complete a full orbit in less than 24 hours, as well as residents of the "Neptunian desert"—a region close to stars where mid-sized planets are theoretically expected to be scarce. The discovery of these outliers provides critical data points for planetary formation theories. Furthermore, the AI identified previously unknown planetary pairs orbiting the same star, offering a clearer picture of multi-planet system architectures. The findings suggest that approximately 9% to 10% of Sun-like stars host a close-in planet, a statistic that aligns with earlier Kepler mission data but carries much higher statistical weight due to the refined AI vetting process.
The success of the RAVEN project signals a broader transition in the field of astrophysics, where the bottleneck is no longer the collection of data, but the intelligence applied to it. As U.S. President Trump’s administration continues to emphasize American leadership in space and technology, the integration of AI into NASA-funded research has become a strategic priority. This automated approach is particularly vital as the scientific community prepares for the Nancy Grace Roman Space Telescope, which is expected to capture tens of thousands of exoplanet transits. Without the deployment of tools like RAVEN or NASA’s own ExoMiner++, the vast majority of that data would likely remain unexamined in digital archives.
The economic and scientific implications of this "AI-first" astronomy are substantial. By reducing the time required to validate a planet from months to hours, researchers can more quickly direct high-demand resources, such as the James Webb Space Telescope, toward the most promising candidates for atmospheric characterization. This targeted approach maximizes the return on billions of dollars in public and private investment in space infrastructure. The University of Warwick’s breakthrough demonstrates that the next frontier of discovery lies in the algorithms that can see what the human eye—and even previous generations of software—simply missed.
Explore more exclusive insights at nextfin.ai.
