In an extraordinary leap forward for exoplanetary science, astronomers at the University of Warwick have harnessed the power of artificial intelligence to validate more than 100 new exoplanets, including 31 that were previously undiscovered. This significant achievement stems from the application of an innovative AI pipeline named RAVEN, specifically designed to analyze the vast, intricate datasets collected by NASA’s Transiting Exoplanet Survey Satellite (TESS). TESS’s mission to survey the sky for subtle dips in starlight—which indicate planets transiting their host stars—has generated a mountain of data that until now posed considerable challenges for processing and interpretation.
The RAVEN framework is groundbreaking because it automates the entire cascade from initial detection of potential planetary signals to their vetting and statistical validation, a task demanding both precision and computational sophistication. By applying RAVEN to observations of over 2.2 million stars gathered during TESS’s first four years, the Warwick team focused on planets with exceptionally short orbital periods, specifically those completing an orbit in less than 16 days. This niche class is critically important for understanding planetary system architectures and the dynamic processes near host stars.
What sets RAVEN apart from previous exoplanet detection tools is its integrated approach. Traditional methods often isolate stages of verification, yet RAVEN simultaneously sifts through vast datasets, utilizes machine learning to discard false positives caused by eclipsing binary stars or instrumental noise, and produces statistically robust planetary validations. This fusion markedly reduces human biases and errors, ensuring that the planets confirmed are truly celestial bodies rather than artifacts. RAVEN’s development was fortified by an extensive training regime on hundreds of thousands of simulated planetary and non-planetary signals, enabling it to identify subtle patterns indiscernible to other methods.
The resulting advances have expanded our census of nearby planetary systems with unprecedented detail and reliability. Among the newly validated planets, researchers unearthed ultra-short-period exoplanets orbiting their stars in under 24 hours—worlds so close that they defy previous theoretical predictions. These extreme environments offer fertile grounds for studying planetary formation and atmospheric evolution under intense stellar radiation and tidal forces. Additionally, the team has confirmed planets residing within the so-called ‘Neptunian desert,’ a parameter space long predicted to be barren due to atmospheric erosion and other destructive processes, yet now known to harbor rare, enigmatic survivors.
Moreover, the Warwick scientists uncovered new multi-planet systems with particularly tight orbits, revealing complex celestial dances that stretch our understanding of system stability in close stellar proximity. The ascertainment of these planets was pivotal in mapping planetary population demographics. Through their companion study published in Monthly Notices of the Royal Astronomical Society, the team delivered one of the most precise statistical frameworks to date, estimating that 9–10% of Sun-like stars host at least one close-in planet. This finding aligns closely with prior results from NASA’s famed Kepler mission but importantly slashes the margin of uncertainty by factors of up to ten.
Dr. Marina Lafarga Magro, the lead author of the validation study, emphasized the transformative impact of these findings, noting that the enlarged and rigorously vetted planetary sample will serve as a cornerstone for future explorations of close-in exoplanetary systems. “RAVEN is not just faster; it is smarter,” she attested, underscoring how the pipeline’s comprehensive validation approach redefines standards within the exoplanet field. The automated nature of RAVEN also provides a viable pathway to tackle the imminent data deluge from forthcoming missions and observatories.
Complementing the discovery pipeline, RAVEN handles the essential challenge of bias quantification. Different planetary sizes and orbital configurations possess varying detectabilities, a factor often obscuring true population statistics. By actively assessing its own detection sensitivity and classification accuracy, RAVEN compensates for these observational biases, leading to cleaner, more representative samples used for further astrophysical inquiry. This self-awareness enables astronomers to ask not only “what planets are there?” but “how representative is our observed sample?”—a subtle but profound dimension to planetary demographics.
This capability has been pivotal for measuring the occurrence rate of planets within the Neptunian desert—a rarely occupied orbital niche where mid-sized planets are theoretically stripped of their gaseous envelopes by intense stellar irradiation and stellar winds. The Warwick team’s estimations indicate an occurrence of merely 0.08% for Sun-like stars, delivering a precise quantification of this sparsity for the first time. Such constraints provide valuable insights into planetary atmospheric retention thresholds and evolutionary pathways, thereby sharpening theoretical models.
RAVEN’s impact transcends pure discovery; it establishes a new paradigm where large astronomical datasets are not merely cataloged but thoroughly interrogated with artificial intelligence to reveal fundamental cosmic patterns. This synergy between big data and AI is emblematic of a broader scientific renaissance whereby computational advances unlock new realms in observational astrophysics. The pipeline’s publicly released interactive tools and catalogs enable global researchers to dive into the findings, identify promising candidates, and set observational priorities for facilities ranging from ground-based telescopes to space missions like ESA’s upcoming PLATO.
The validation of hundreds of new planets represents more than a simple expansion of known worlds; it refines astronomers’ ability to statistically assess planet formation scenarios, migration histories, and star-planet interactions at close range. Such refined knowledge is essential for understanding the evolutionary fates of planetary systems, including our own Solar System. Notably, the ultra-short-period planets discovered may harbor exotic atmospheric phenomena induced by intense magnetic interactions and extreme temperatures exceeding thousands of degrees Fahrenheit.
In integrating the latest advances in AI, statistical rigor, and rich observational datasets, the University of Warwick research team asserts a future where the rate of planetary discovery accelerates dramatically, and our grasp of planetary diversity deepens profoundly. The RAVEN project exemplifies how interdisciplinary approaches—combining astrophysics, data science, and machine learning—can surmount previously insurmountable barriers, heralding a new epoch in exoplanetary astronomy.
As TESS continues its survey and new missions come online, the foundational work embodied in RAVEN and its validated planets charts a course for both targeted follow-up observations and large-scale demographic studies. It also suggests that Earth-like planetary systems in close orbits, or with unusual atmospheric compositions, may be more accessible for discovery than once anticipated. The accompanying release of datasets and rankings positions the global astronomical community to maximize scientific return and prioritize exploratory investigations that push the frontiers of our knowledge about other worlds orbiting distant suns.
The prospect of discovering more ultra-short-period planets, clarifying the boundaries and characteristics of the Neptunian desert, and decoding the architecture of tightly clustered planetary systems collectively enriches our cosmic perspective. Each newly validated world is a vital piece within the grand puzzle of planet formation and evolution, contributing uniquely to our understanding of how common—or rare—varied planetary types truly are. This work represents a decisive milestone, unlocking previously hidden patterns within the immense cosmic survey data, and elevating humanity’s quest to understand worlds beyond our own.
Subject of Research: Not applicable
Article Title: Automatic search for transiting planets in TESS-SPOC FFIs with RAVEN: Over 100 newly validated planets and over 2000 vetted candidates
News Publication Date: 24-Mar-2026
Web References:
- https://academic.oup.com/mnras/article-lookup/doi/10.1093/mnras/stag512
- https://doi.org/10.1093/mnras/stag022
- https://arxiv.org/abs/2509.17645
References:
- Lafarga Magro, M., et al. “Automatic search for transiting planets in TESS-SPOC FFIs with RAVEN: Over 100 newly validated planets and over 2000 vetted candidates.” Monthly Notices of the Royal Astronomical Society, 2026. DOI: 10.1093/mnras/stag512
- Cui, K., et al. “Demographics of Close-In TESS Exoplanets Orbiting FGK Main-sequence Stars.” Monthly Notices of the Royal Astronomical Society, 2026. DOI: 10.1093/mnras/stag022
- Hadjigeorghiou, A., et al. “RAVEN: RAnking and Validation of ExoplaNets.” arXiv preprint arXiv:2509.17645, 2025.
Image Credits: NASA, ESA, and A. Schaller (for STScI)
Keywords
Exoplanets, Artificial Intelligence, TESS, RAVEN, Ultra-Short-Period Planets, Neptunian Desert, Planetary Validation, Machine Learning, Exoplanet Demographics, Astronomical Surveys, Data Science, Planetary System Architecture
