Somewhere in four years of starlight recorded by NASA’s planet-hunting telescope, 114 worlds were hiding in plain sight. They had already been photographed, their host stars already cataloged, their data already run through standard processing pipelines. But it took an artificial intelligence tool built specifically to sift signal from noise to pull them out.
The tool is called RAVEN, short for RAnking and Validation of ExoplaNets. In a peer-reviewed study published in Monthly Notices of the Royal Astronomical Society, a research team describes how the pipeline combed through calibrated images from NASA’s Transiting Exoplanet Survey Satellite (TESS), scanning roughly 2.2 million stars across 55 sectors of sky. The result: 114 statistically validated exoplanets and more than 2,000 additional candidates still awaiting confirmation. The NASA Exoplanet Archive, the definitive catalog maintained by Caltech under NASA contract, has accepted all 114 and characterized the batch as the largest mass confirmation of TESS planets to date.
Why this batch matters
To appreciate the scale, consider the bottleneck RAVEN was designed to break. TESS, launched in 2018, surveys nearly the entire sky by staring at successive strips of stars and watching for the tiny, periodic dips in brightness that occur when a planet crosses in front of its host star. The telescope has generated an enormous backlog of candidate signals, but turning a candidate into a confirmed planet traditionally requires follow-up observations from ground-based telescopes, a resource that is chronically oversubscribed. As of early 2025, TESS had contributed roughly 500 confirmed planets to the Archive. RAVEN’s 114 additions represent a jump of more than 20 percent in a single study.
For historical comparison, the largest single batch of validated exoplanets in any mission’s history came from Kepler in 2016, when Timothy Morton and colleagues used a similar statistical approach to confirm 1,284 worlds at once. RAVEN operates on the same philosophical ground: rather than waiting for a second telescope to independently verify each signal, the pipeline uses probability to do the heavy lifting.
How RAVEN works
RAVEN is not a simple yes-or-no classifier. According to a separate methods paper, the system is a Bayesian framework trained on both simulated planet transits and several categories of astrophysical imposters: eclipsing binary stars, background stellar blends, and instrumental glitches that can all mimic the signature of a transiting planet. For every candidate signal, RAVEN weighs competing explanations and assigns a probability that the dip is genuinely planetary in origin.
This is what astronomers call “statistical validation” rather than “confirmation by observation.” A validated planet has not been independently weighed or had its atmosphere sampled. Instead, the analysis has shown that the probability of the signal being a planet far exceeds the probability of every known false-positive scenario combined. The methods paper reports standard machine-learning performance metrics, including precision and recall scores, that quantify how reliably the system distinguishes real planets from fakes. However, detailed per-planet probability tables for the full TESS sample have not yet been released publicly.
The search targeted stars observed in TESS sectors 1 through 55, covering the mission’s first four years, and focused on orbital periods between 0.5 and 16 days. That window naturally favors close-in planets that orbit their stars quickly and transit frequently, making their signals easier to detect. Several of the newly validated worlds orbit small, cool stars where even a tight orbit can place a planet in a relatively temperate zone, though the published data does not yet include a full accounting of which, if any, of the 114 might be considered potentially habitable.
What we still don’t know
Statistical validation tells astronomers that a signal is almost certainly a real planet, but it does not reveal the planet’s mass or what its atmosphere is made of. Those measurements require dedicated spectroscopic follow-up, either from ground-based instruments that detect the gravitational tug of a planet on its star or from space telescopes like the James Webb Space Telescope that can read the chemical fingerprints in a planet’s atmosphere. The published study and its preprint do not outline specific follow-up plans, leaving open the question of which of these 114 worlds will be prioritized for deeper study.
There is also the question of scale. RAVEN processed the first 55 TESS sectors, but the mission has continued observing well beyond that window. A substantial volume of newer imaging data remains untouched by this pipeline. Whether the team or other groups will extend the analysis, and how many additional planets such an extension might yield, is not addressed in the published work.
Finally, the broader astronomical community has not yet had time to cross-check a large subset of RAVEN’s validations using independent techniques. The statistical thresholds are well-defined in the methods paper, and the NASA Exoplanet Archive’s acceptance carries significant institutional weight, but some researchers may want to see traditional follow-up observations corroborate a sample of the results before treating statistical validation and observational confirmation as fully interchangeable in population studies.
A new model for mining the sky
The deeper significance of the RAVEN result is not the number 114 itself but what it demonstrates about the future of planet-hunting. Ground-based telescopes cannot keep pace with the flood of candidates generated by space surveys. Every year, the gap between detected signals and confirmed planets widens. RAVEN shows that a well-designed machine-learning pipeline can close that gap by converting vast imaging archives into curated catalogs of validated worlds, extracting discoveries from data that human analysts and traditional software had already processed without finding them.
These 114 planets were always there, encoded in the same photons that TESS beamed back to Earth years ago. What changed was the question being asked of the data, and the tool built to answer it. As TESS continues to scan the sky and its archive grows, pipelines like RAVEN may become less of a novelty and more of a necessity, reshaping how astronomers turn starlight into science.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
