Real-Time Decoding, Algorithmic GPU Decoders, and AI Inference Enhancements in NVIDIA CUDA-Q QEC

Was the experiment a circuit level noise simulation or something simpler, like code capacity where DEMs were not used?