Identifying Logical Operations from Syndrome Histories for qLDPC Codes

Abstract

Syndrome histories are normally used to detect and correct errors in fault-tolerant quantum computation, but they can also reveal information about the logical operations executed during computation. We study logical-operation identification using syndrome histories alone or together with other classical measurement outcomes. We develop an exact inference method and a scalable approximation that account for unknown circuit faults and ideal measurement randomness. We evaluate automorphisms, logical Pauli measurements, T injections, and logical circuits combining these instructions on \([[18,4,2]]\) and \([[72,12,6]]\) bivariate bicycle (BB) codes under multiple noise models and error rates. A single logical instruction can be identified with over 94\% accuracy under circuit-level noise with error rates between \(10^ {-5} \) and \(10^{-3}\). We also achieve over 99\% accuracy for automorphism-based logical circuits and 78.32\% for logical circuits combining all three instructions. We further analyze identification ambiguities arising from insufficient or indistinguishable measurement information, fault timing, and approximation inference. Together, these results extend the role of syndromes beyond error decoding to identifying executed logical operations and circuits during fault-tolerant quantum computation. They also suggest that access to syndrome histories and related classical measurement outcomes may unintentionally reveal information about the executed computation.