Key Takeaways
- Astronomers at the University of Warwick validated 118 new exoplanets and identified over 2,000 high‑quality planet candidates using the AI‑driven RAVEN pipeline applied to four years of TESS data.
- The study focused on short‑period planets (orbital periods < 16 days), yielding one of the most precise measurements of how common such worlds are around Sun‑like stars.
- Notable discoveries include ultra‑short‑period planets (< 24 h orbit), planets residing in the Neptunian desert, and previously unknown tightly packed multi‑planet systems.
- RAVEN’s machine‑learning models, trained on hundreds of thousands of realistic simulations, can distinguish genuine planetary transits from false positives such as eclipsing binary stars.
- The refined dataset shows that ≈ 9‑10 % of Sun‑like stars host a close‑in planet, while Neptunian‑desert planets occur around only 0.08 % of such stars—reducing uncertainties by up to a factor of ten compared with earlier Kepler estimates.
- The team has released interactive catalogs and tools, enabling the broader community to prioritize targets for follow‑up observations with ground‑based telescopes and upcoming missions like ESA’s PLATO.
Overview of the Warwick Study and the RAVEN Pipeline
Researchers from the University of Warwick have harnessed artificial intelligence to sift through the vast trove of data collected by NASA’s Transiting Exoplanet Survey Satellite (TESS). By applying their newly developed RAVEN (Rapid Automated Validation and Extraction Network) pipeline to more than 2.2 million stars observed during TESS’s first four years, the team confirmed 118 new exoplanets and flagged over 2,000 high‑quality planet candidates, nearly half of which were previously unknown. As lead author Dr. Marina Lafarga Magro explained, “Using our newly developed RAVEN pipeline, we were able to validate 118 new planets, and over 2,000 high‑quality planet candidates, nearly 1,000 of them entirely new.” This work represents one of the best‑characterized samples of close‑in planets to date, providing a reliable foundation for statistical studies of planetary populations.
Validation of Planets and Candidate Identification
The RAVEN system does not merely flag potential transits; it subjects each signal to a rigorous vetting process that combines machine‑learning classification with statistical validation. By training on hundreds of thousands of simulated planetary transits and astrophysical false positives, the AI learns to discriminate subtle patterns that human analysts might miss. Consequently, the pipeline produced a highly reliable list of candidates, reducing the typical false‑positive rate that plagues large‑scale transit surveys. The resulting catalog not only expands the known exoplanet inventory but also supplies a clean dataset that can be used to measure occurrence rates with unprecedented precision.
Rare and Extreme Planet Types Uncovered
Among the validated worlds are several intriguing subclasses. The study identified ultra‑short‑period planets that complete an orbit in less than 24 hours, pushing the limits of planetary formation and survival theories. It also populated the so‑called Neptunian desert—a region of parameter space where planets of Neptune‑size are expected to be exceedingly rare—with a handful of objects that defy current expectations. Additionally, the researchers uncovered tightly packed multi‑planet systems, including previously unknown pairs of planets sharing the same host star. These findings highlight the diversity of planetary architectures that TESS, when paired with powerful AI tools, can reveal.
How RAVEN Improves Planet Detection
A core strength of RAVEN lies in its ability to tackle the enduring challenge of distinguishing true planetary signals from mimickers such as eclipsing binary stars. As Dr. Andreas Hadjigeorghiou, the pipeline’s lead developer, noted,
“The challenge lies in identifying if the dimming is indeed caused by a planet in orbit around the star or by something else, like eclipsing binary stars, which is what RAVEN tries to answer. Its strength stems from our carefully created dataset of hundreds of thousands of realistically simulated planets and other astrophysical events that can masquerade as planets. We trained machine learning models to identify patterns in the data that can tell us the type of event we have detected, something that AI models excel at.”
Moreover, RAVEN integrates detection, vetting, and statistical validation into a single workflow, offering an advantage over tools that address only isolated stages of the process. This end‑to‑end approach ensures consistency and objectivity when analyzing massive datasets.
Measuring How Common Planets Really Are
With a rigorously validated sample in hand, the Warwick team moved beyond individual discoveries to quantify planet occurrence rates. In a companion MNRAS study, they mapped the frequency of close‑in planets as a function of orbital period and size around Sun‑like stars. The analysis revealed that approximately 9‑10 % of such stars host a planet with an orbital period under 16 days—a figure that aligns with earlier Kepler measurements but with uncertainties reduced by up to a factor of ten.
Regarding the Neptunian desert, Dr. Kaiming Cui, first author of the population study, stated,
“For the first time, we can put a precise number on just how empty this ‘desert’ is… These measurements show that TESS can now match, and in some cases surpass, Kepler for studying planetary populations.”
The study found that Neptunian‑desert planets occur around only 0.08 % of Sun‑like stars, underscoring the extreme paucity of these worlds and providing a critical test for migration and formation models.
A New Era for Planet Discovery and Community Resources
Together, these results illustrate how artificial intelligence is reshaping exoplanet science. By marrying massive space‑based datasets with sophisticated machine‑learning techniques, researchers can not only uncover new worlds but also refine the very tools used to find them. To maximize the impact of their work, the Warwick team has released interactive catalogs and analytical tools that allow other scientists to explore the validated planet list, identify promising targets for follow‑up spectroscopy with ground‑based observatories, and prioritize candidates for upcoming missions such as ESA’s PLATO. This open‑access approach ensures that the benefits of the RAVEN pipeline extend far beyond the original study, fostering collaborative advances in our understanding of planetary systems across the galaxy.
What Is RAVEN
RAVEN is an automated system designed to convert the enormous streams of photometric data from space telescopes like TESS into trustworthy planetary discoveries. It begins by scanning light curves for the tiny, periodic dips in brightness that signal a transiting planet. Each candidate is then passed through artificial‑intelligence models trained on realistic simulations of both genuine transits and common false positives—such as eclipsing binaries or instrumental artefacts—to assess the likelihood of a planetary origin. Finally, the pipeline performs statistical validation, quantifying confidence levels and correcting for detection biases. By evaluating which planet types are more or less easily recovered, RAVEN enables researchers to derive unbiased occurrence rates. In essence, the system not only accelerates the pace of discovery but also delivers cleaner, more scientifically robust datasets that can be used to answer fundamental questions about how prevalent different kinds of planets are throughout the Milky Way.
https://www.sciencedaily.com/releases/2026/05/260502233926.htm