New decoder cuts logical errors up to 25x for correlated atom loss in Gross code
Researchers led by Jacob W. Murri at UCLA describe the Maximal-Matching-Mixture (MMM) decoder for neutral atom quantum computers, which exploits the tendency of an atom's gate partner to be lost when the first atom fails. MMM builds a weighted mixture of decoding hypergraphs based on likely loss scenarios and interprets it with standard belief propagation. Applied to the [[144,12,12]] Gross code, it lowered logical error rates by up to about 25 times compared with decoders that do not account for atom loss, and by roughly 2 times compared with earlier loss-aware methods from Baranes et al. The work is an arXiv preprint and has not yet been peer reviewed; the authors say further work will test whether the gains extend beyond this code structure.