When black holes or neutron stars merge, they disturb the very fabric of spacetime, producing gravitational waves — ripples predicted by Einstein in his general theory of relativity. Detectors like LIGO pick up this 'cosmic chime'; its frequency and rising pitch (the so-called 'chirp') carry information about the masses of the colliding objects. Yet in this cosmic orchestra, there is a zone of silence: objects with masses between 2 and 5 solar masses are almost never detected. This lower mass gap lies between the most massive neutron stars and the lightest black holes, and its origin remains a mystery.

To peer into this 'silent' region, the neural network GWSkyNet-MassGap was created. It analyzes the 'chirp' of gravitational waves and determines whether a neutron star or an object from the mass gap was involved — much like an experienced listener distinguishes instruments in a symphony orchestra. The algorithm shows high accuracy for loud signals, but may err in estimating mass distributions if the signal is weak. Additional information comes from the electromagnetic channel: the collision of neutron stars is accompanied by a kilonova — a flash in which heavy elements, including gold, are synthesized; mergers of black holes, however, remain 'dark' to telescopes.

Combing through catalogs of gravitational-wave events, GWSkyNet-MassGap identifies rare candidates in the gap. The mass gap may hide surprises: perhaps these are 'ghosts' of quark matter — objects that are no longer neutron stars but not yet black holes, and whose nature challenges our understanding of matter under extreme conditions.