Ming Wang, Ang Li, Frank Mueller (Jul 02 2025).
Abstract: In this work, we propose a lightweight decoder based solely on belief-propagation (BP), augmented with a speculative post-processing strategy inspired by classical Chase decoding. Our method identifies unreliable bits via BP oscillation statistics, generates a set of modified test patterns, and decodes them in parallel using low-iteration BP. We demonstrate that our approach can achieve logical error rates comparable to or even better than BP-OSD, but has lower latency over its parallelization for a variety of bivariate bicycle codes, which significantly reduces decoding complexity.