Good news for people who like planets and dislike manual data sifting: an artificial intelligence called Raven helped confirm the existence of more than one hundred planets outside our Solar System by combing through data from NASA’s TESS space telescope.

What the team did

Researchers at the University of Warwick fed Raven the light measurements collected by TESS during its first four years. In total, Raven analyzed more than 2.2 million individual observations, concentrating on signals from planets that orbit close to their stars — typically completing an orbit in under 16 days.

Key numbers you can repeat at parties

  • 118 exoplanets were validated with Raven’s help.
  • Among those validated, 31 had not been identified before.
  • Raven also flagged over 2,000 candidate exoplanets, about 1,000 of which appear to be entirely new candidates.

What makes these finds interesting

The analysis highlights several categories of interest:

  • Planets that orbit their star in less than 24 hours — yes, they are extremely close in.
  • Worlds sitting in the so-called "Neptune desert", a region where Neptune-sized planets are expected to be rare.
  • Systems where multiple planets circle very close to their star, giving compact, tightly packed planetary systems.

From the dataset the team also estimated that about 10% of Sun-like stars host at least one close-in planet. By contrast, Neptune-like planets appear very uncommon around Sun-like stars, showing up around roughly 0.08% of them.

How Raven tells planets from impostors

Raven is a machine learning system trained to spot patterns in light curves that match specific types of events. In practice it helps determine whether a dip in a star’s brightness is likely caused by a planet passing in front of the star or by something else, such as stellar variability or instrumental effects.

The goal was not just to make a long list of possible planets. The team built Raven to be a reliable tool that can produce a sample suitable for studying the prevalence of different planet types around stars similar to the Sun.

Who published this and what’s next

The results appear across three papers in Monthly Notices of the Royal Astronomical Society, led by researchers at the University of Warwick. The work was guided by scientists who developed Raven and by those who trained its machine learning models to recognize different signal types.

The validated planets and the large candidate list give astronomers a richer set of targets to follow up with other telescopes and instruments. In short: Raven narrowed the haystack and handed astronomers a lot more needles.

Image credit listed with the original report.