Imagine a simultaneous interpreter in a room where two people are speaking: they must keep up with both meaning and intonation, but if the speech is too fast, only a paraphrase remains. That's roughly how scientists feel as the stream of gravitational waves from black hole and neutron star mergers grows, while precise signal calculations lag behind. Detectors LIGO and Virgo have already caught hundreds of such events, but future facilities, Einstein Telescope and Cosmic Explorer, will increase the rate by orders of magnitude. Each event must be processed: compared with theoretical templates and parameters estimated, which takes hours to days of computation.

The research group proposed a two-stage neural network that acts like that interpreter: first quickly produces a draft, then refines details. The first stage, a conditional autoencoder, compresses information about masses and spins into a compact code and predicts the amplitude and phase of the wave from it. The second stage, a calibration convolutional network, corrects residual errors, especially in the merger region where the signal is most complex. Then the wave is "groomed": edges are smoothed with sinusoidal windows and embedded in an eight-second zero-padded segment to avoid Fourier transform artifacts.

Then comes the most interesting part. The authors didn't just measure accuracy; they ran the surrogate through a full Bayesian pipeline: generated artificial signals, added noise, and recovered parameters. It turned out that at a signal-to-noise ratio (SNR) of about 10, the bias in mass estimation becomes comparable to the statistical uncertainty. For weak signals, such a surrogate is usable without refinement; for loud ones, either 10⁻⁴ accuracy or reweighting is needed.

This work differs from neighboring projects in that it tests the surrogate in action, not just on waveform accuracy. It sets a practical criterion for future catalogs: handle weak events quickly, strong ones carefully. Then the interpreter, even if making mistakes in details, won't lose the thread of the conversation.