AI tool developed at Oxford helps astronomers find supernovae in a sky full of noise

A new AI-powered tool has reduced astronomers’ workload by 85% - filtering through thousands of data alerts to identify the few genuine signals caused by supernovae (powerful explosions from dying stars). The findings have been published today (10 Sept) in The Astrophysical Journal. 

Lead researcher Dr Héloïse Stevance (Department of Physics, University of Oxford) said: “The surprising thing is how little data it took. With just 15,000 examples and the computing power of my laptop, I could train smart algorithms to do the heavy lifting and automate what used to take a human beings hours to do each day. This demonstrates that with expert guidance, AI can transform astronomical discovery without requiring enormous data sets or computational power.” 

Finding the needles in a cosmic hay stack 

Supernovae are rare, bright explosions marking the death of massive stars; events that help scientists understand the origin of chemical elements. These explosions appear unexpectedly across the night sky and must be spotted quickly before they fade - essentially a cosmic game of spot the difference.

A team of researchers led by Oxford University and Queen’s University Belfast search for these using the Asteroid Terrestrial Impact Last Alert System (ATLAS). This system, originally built as an asteroid impact early warning system, scans the entire visible sky every 24 to 48 hours using five telescopes located around the globe. It is a NASA funded project, led by the University of Hawaii, and Oxford processes the data for high explosions beyond our galaxy. The search yields millions of potential alerts nightly, most of which are noise (either instrumental errors or known objects). 

Even after applying standard filtering and automated image analysis techniques, the researchers were left with between 200 and 400 candidate signals each day that needed to be manually sifted through. Only a handful of these would be genuinely interesting phenomena such as supernovae or extragalactic transients (the optical counterparts to gamma ray bursts).  

“This manual verification would take several hours each day,” added Dr Stevance. “Thanks to our new tool, we can free up scientists’ time for what they do best; creative problem solving and questioning the nature of our Universe. It's the astrophysical equivalent of having a bot doing your laundry so you can focus on your art!” 

The new tool, called the Virtual Research Assistant (VRA), is a collection of automated bots that mimics the human decision-making process by ranking alerts based on their likelihood of being real, extragalactic explosions. Unlike many AI-automated approaches that require vast training data and supercomputers, the VRA uses a leaner approach. Instead of data-hungry deep learning methods, the system uses smaller algorithms based on decision trees that looks for patterns in selected aspects of the data. This allows scientists to inject their expertise directly into the model and guide the algorithms to key features to look for. 

Crucially, the VRA updates its assessment each time a telescope revisits the same patch of sky. This means a signal is automatically re-checked and re-scored over several nights, with only the most promising candidates passed on to human astronomers to review. 

In its first year of use, the VRA successfully filtered over 30,000 alerts while missing fewer than 0.08% of real supernovae alerts. This ultimately reduced the number of alerts passed on to human ‘eyeballers’ to verify by around 85%, whilst retaining more than 99.9% of genuine supernovae candidates.  

Since December 2024, the VRA has been linked with the South African Lesedi Telescope so that it can automatically trigger follow-up observations for the most promising signals, even before a human has reviewed the data. This has already resulted in new supernovae being confirmed.  

Study co-author Professor Stephen Smartt (Department of Physics, University of Oxford) said: “The speed and accuracy of this tool will supercharge our team’s ability to find and study strange and rare phenomena in the cosmos – for instance, explosions from dying stars in distant galaxies, that can teach us how the chemical elements are created and how fast the Universe is expanding. We will also be able to more efficiently match optical sources to emissions in the gamma ray, x-ray and radio frequencies and possibly gravitational waves. The speed and accuracy of the models are impressive.”  

The future is bright 

This achievement comes just in time, with the upcoming launch of the Vera Rubin Observatory’s Legacy Survey of Space and Time (LSST) in early 2026. Over ten years, this will survey the entire southern hemisphere sky every few days, ultimately generating over 500 petabytes* of images and data.  

“The LSST is set to deliver over 10 million alerts each night, finding everything from moving asteroids, supernovae, matter falling onto black holes, merging neutron stars and probably new phenomena as well. Our job as astronomers will be to keep up with this avalanche of data,” added Dr Stevance. “Tools like our new Virtual Research Assistant will be invaluable in helping us to better understand how supernovae and their massive stars made all the chemical elements necessary for the world as we know it – from hydrogen to apple pies.” 

Dr Stevance is currently building Virtual Research Assistants for the UK and European LSST data brokers (Lasair, Fink) with her ambition being to use the LSST data to build bots that can pre-emptively hunt for supernovae by predicting when and where they will explode. 

Dr Stevance added: “In Astronomy new knowledge is extracted from data, and LSST will be revolutionary: in its first year alone, it will capture more data than every survey ever. I feel so privileged to live and work at such a historical moment.” 

*A petabyte is equivalent to 10^15 bytes or one million gigabytes (GB). 

Dr Stevance’s position is funded by The Eric and Wendy Schmidt AI in Science Postdoctoral Fellowship scheme. 

Notes to editors: 

For media requests or interviews, contact: Dr Héloïse Stevance: hfstevance@gmail.com, 07 707 035 548. 

The study ‘The ATLAS Virtual Research Assistant’ will be published in The Astrophysics Journal at 00:01 BST Wednesday 10 September / 19:01 ET Tuesday 9 September 2025, DOI: at 10.3847/1538-4357/adf2a1. To view a copy of the paper before this under embargo, contact: Dr Héloïse Stevance: hfstevance@gmail.com

Images related to this study that can be used to illustrate articles are available at https://drive.google.com/drive/folders/1furRHPY0mnuyCuHer-91QlQZrTx5HVVQ?usp=sharing. These images are for editorial purposes relating to this press release ONLY and MUST be credited (see file name). They MUST NOT be sold on to third parties. 

About the University of Oxford 

Oxford University has been placed number 1 in the Times Higher Education World University Rankings for the ninth year running, and ​number 3 in the QS World Rankings 2024. At the heart of this success are the twin-pillars of our ground-breaking research and innovation and our distinctive educational offer. 

Oxford is world-famous for research and teaching excellence and home to some of the most talented people from across the globe. Our work helps the lives of millions, solving real-world problems through a huge network of partnerships and collaborations. The breadth and interdisciplinary nature of our research alongside our personalised approach to teaching sparks imaginative and inventive insights and solutions. 

Through its research commercialisation arm, Oxford University Innovation, Oxford is the highest university patent filer in the UK and is ranked first in the UK for university spinouts, having created more than 300 new companies since 1988. Over a third of these companies have been created in the past five years. The university is a catalyst for prosperity in Oxfordshire and the United Kingdom, contributing around £16.9 billion to the UK economy in 2021/22, and supports more than 90,400 full time jobs. 

---

---