Identify the claim’s exact task

Quantum advantage compares a quantum method with a relevant classical method for a defined computational task. The task may be random-circuit sampling, estimating an observable, optimization, simulation, or something else. Performance on one task does not transfer automatically to another, and a sampling demonstration need not produce a commercially useful answer.

Begin by writing down the input, required output, success criterion, and allowed error. If the paper’s conclusion changes when those details change, they belong in every summary of the result. Broad phrases such as faster than a supercomputer remove the conditions that make the comparison meaningful.

Interrogate the classical baseline

A speedup is relative to a particular classical algorithm, implementation, hardware system, and resource estimate. After a quantum result appears, improved classical algorithms or better use of accelerators can narrow the gap. That does not erase the experiment; it changes the demonstrated comparison.

Check whether the baseline was executed or extrapolated, whether preprocessing and data movement were counted, and whether the classical method produced the same output quality. A fair comparison uses matching error tolerance and reports uncertainty in large extrapolations.

Trace verification and noise

Large quantum outputs can be hard to verify directly. Experiments may validate smaller instances, use statistical tests, compare selected observables, or rely on theoretical arguments about the target distribution. Each approach supports a different strength of conclusion.

Noise mitigation, postselection, and calibration can be legitimate parts of an experiment, but they consume resources and can limit scaling. Ask which runs were discarded, how uncertainty was estimated, whether the result survives reasonable noise models, and how performance changes as the instance grows.

  • Name the computational task and success threshold.
  • Match output accuracy across the quantum and classical methods.
  • Separate executed runtimes from extrapolated estimates.
  • Include sampling, calibration, error mitigation, and verification costs.
  • State whether the result demonstrates scientific, computational, or practical advantage.

Use a claim ladder

A useful reading discipline is to place the result on a ladder. First, did the device perform the intended quantum circuit? Second, did it outperform the stated classical baseline for the defined task? Third, does the task answer a useful scientific or operational question? Fourth, do total time, cost, energy, reliability, and access support practical use?

A paper can make a strong contribution on an earlier rung without satisfying the later ones. Reporting the rung precisely is more informative than accepting or rejecting quantum advantage as a single label. Follow the original methods and subsequent classical work before repeating a headline-sized estimate.