A new test checks the software behind quantum error correction
DEMVal compares error models instead of relying on sampled outputs.
Reading tools for text size, copying, printing and your reading list

In , a decoder reads a pattern of checks and uses an error model to decide which correction to make. RIKEN researchers developed DEMVal to check the model that software gives the decoder. If the model assigns the wrong probabilities to errors, the decoder can choose a less reliable correction.[1]
Their November 16 preprint reports six defects across tested versions of NVIDIA CUDA-Q QEC, Riverlane Deltakit and Microsoft QDK-EC’s deq component. The underlying issues were publicly reported in June: this is a new research account, not a fresh disclosure of vulnerabilities.[1][2]
Instead of estimating agreement from repeated simulated outcomes, DEMVal compares each generated error model with a reconstructed reference. The authors tested 551 memory-experiment configurations and report that developers confirmed the six defects and accepted fixes.[1]
The study concerns software models, not failures measured on quantum processors. Its conclusions depend on the reference construction and tested configurations; it does not certify every compiler or error model.[1]
Keep your reading trail
Keep up with these topics
Explore the subjects covered in this story.