Real-time decoding of quantum error correction is essential for enabling fault-tolerant quantum computation. A practical decoder must operate with high accuracy at low latency, while remaining robust to spatial and temporal variations in hardware noise. We introduce a hardware-conditioned neural decoder framework designed to exploit the natural separation of timescales in superconducting processors, where calibration drifts occur over hours while error correction requires microsecond-scale responses. By processing calibration data through a graph-based encoder and conditioning a lightweight convolutional backbone via featurewise linear modulation (FiLM), we decouple the heavy processing of device statistics from the low-latency syndrome decoding. We evaluate this approach using the repetition code as a test bed on IBM Fez, Kingston, and Pittsburgh processors, collecting over 2.7 million experimental shots spanning distances up to $d=11$. Training one decoder per code distance, basis, and round count $(d,r)$, we demonstrate that each trained model generalizes to unseen qubit chains and new calibration data acquired days later without retraining. The FiLM-conditioned decoder achieves up to a $11.1\times$ reduction in logical error rate relative to modified minimum-weight perfect matching. Cross-conditioning ablations further show that the calibration input actively improves decoding, with performance degrading when the decoder is conditioned on incorrect backend data or the conditioning is disabled. We observe that by employing a network architecture that exploits the highly asynchronous nature of system calibration and decoding, hardware-conditioned neural decoding demonstrated promising, adaptive performance with negligible latency overhead relative to unconditioned baselines.

Close

PrevNext

Reuse & PermissionsView figure in article

• 
• 
• 
• 
• 
• 
• 
• 

• Quantum error correction
• Quantum software