Reproduce “Learning to Decode the Surface Code with a Recurrent, Transformer-Based Neural Network” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Johannes Bausch, Andrew W Senior, Francisco J H Heras, Thomas Edlich, Alex Davies, Michael Newman, Cody Jones, Kevin Satzinger and 10 more
The authors train a recurrent Transformer decoder to combine noisy surface-code measurements across rounds, passing information among stabilizers and using measurement probabilities and leakage signals. On Sycamore distance-3 and distance-5 memory experiments, it achieves lower logical error per round than the previous best reported tensor-network decoder; at distance 5, rates are 2.748 ± 0.015% versus 2.915 ± 0.016%. Higher-distance results are simulated, not hardware demonstrations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.