My new study introduces Vision Transformers and Large Language Models as new tools for tracking asteroids affected by secular resonances. Secular resonances are gravitational effects that slowly alter an asteroid’s orbit over long periods, often due to consistent gravitational pulls from planets like Jupiter and Saturn. ViTs, built on the same Transformer technology as popular AI language models, quickly and accurately sort vast datasets—tasks that previously took weeks now take only seconds. This method overcomes the limits of traditional visual classification by automating and accelerating asteroid tracking.
The success of ViTs in this study shows the wider potential of Transformers beyond language, suggesting they could become a go-to tool in astronomy. With more extensive data expected from new telescopes, this approach is a big step forward for processing astronomical data efficiently and could inspire further AI applications across different scientific fields.