A machine-learning pipeline for real-time detection of gravitational waves from compact binary coalescences

We present Aframe, the first machine learning-based pipeline for the detection of gravitational waves to run in low-latency. The pipeline's offline performance is compared to the offline performance of traditional search pipelines during the third observing run of the LIGO-Virgo-KAGRA (LVK) collaboration, and we demonstrate state-of-the-art sensitivity for a subset of the binary black hole population. Additionally, we find that Aframe is consistently able to perform low-latency detections at a fraction of the computational cost of traditional searches. Ultimately, multi-messenger astronomy will require rapid detection of gravitational waves to maximize the amount of time available for follow-up observations, and Aframe represents a crucial step towards this goal.