QASM-Eval: A Dataset to Train and Evaluate LLMs on OpenQASM-3 Beyond Quantum Circuits
A preprint introduces QASM-Eval, a dataset intended to train and evaluate large language models on OpenQASM-3 code. The abstract frames the dataset around NISQ-era constraints, arguing that useful quantum programs need hardware-facing features beyond ordinary gate sequences, such as mid-circuit measurement and classical feedback for quantum error correction, and precise timing for dynamical decoupling. The abstract text is cut off after those examples.
AI analysis — not reported by the source
What this could mean
- 0–2 yearsPlausible
Within two years, an LLM fine-tuned on QASM-Eval could reliably generate OpenQASM-3 snippets that include mid-circuit measurement and timing controls, making it easier for experimental teams to prototype QEC and dynamical decoupling routines on available hardware.
Existing LLMs already translate natural-language specifications into code in other domains; a targeted dataset with evaluation metrics gives the training signal needed for OpenQASM-3's less common constructs. The main preconditions are sufficient dataset coverage of those constructs and adoption by a quantum software team.
This is a brief. The day’s lead story carries the full analysis.