Preprint reports AI search improves quantum error-correction codes

Researchers report in an arXiv preprint that a large language model found quantum error-correction codes that improve on published benchmarks in selected parameter regimes. Their search targeted codes with fewer than 600 data qubits and check weight at most 12, with exactly certified distances for the leading results. The model modifies construction programs and their search strategies. Under the same evolution framework and step budget, the best model outperformed all random-mutation and hill-climbing runs on both optimization objectives. Optimizing circuit-level logical error rates typically produced lower error rates than optimizing the alternative code-quality metric, often despite smaller code distances.