Evolutionary Discovery of Bivariate Bicycle Codes with LLM-Guided Search

Juan Cruz-Benito, Andrew W. Cross, David Kremer, Ismael Faro · 2026

Read Evolutionary Discovery of Bivariate Bicycle Codes with LLM-Guided Search by Juan Cruz-Benito, Andrew W. Cross, David Kremer, Ismael Faro (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-02418.txt.

Description

Cruz-Benito and co-authors use language models to hunt for new quantum error-correcting codes. The search space for quantum LDPC codes is vast. Each candidate must also pass strict checks. Their workflow runs as evolution. Language models mutate Python programs. Those programs emit bivariate bicycle codes and perturbed variants.

Five campaigns ran about 1,650 evolution steps. They screened some 200,000 candidate codes. A staged pipeline then certifies the survivors. Checks cover rank, distance, and graph duplication. Local-Clifford equivalence tests also run. At block length up to 360, the hunt found 465 distinct candidates. Of these, 97 are CSS bicycle codes. The rest are perturbed non-CSS variants. Known top codes were recovered. New finds include an indecomposable code with parameters 288, 16, and 12. The work shows guided evolution as a practical discovery tool.

Catalog record

CREATORJuan Cruz-Benito, Andrew W. Cross, David Kremer, Ismael Faro
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE106 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.02418

Browse the full library · About the Tower of Babel