zero-noise extrapolation
The most common error mitigation method: measure at higher noise, then trace back to zero.
Zero-noise extrapolation runs the same program at several deliberately raised noise levels. It watches how the answer changes, then traces that trend back to zero noise. It is cheap in shots. But it assumes a particular shape for the trend. If the real trend is a different shape, the corrected answer is wrong, and more shots will not fix it.
Think of it as
Like timing a runner carrying 5, 10 and 15 kg, then guessing their speed carrying nothing.
Example
A provider reports a corrected answer of 0.62. The head of R&D asks whether it changes when a more flexible curve is fitted. If it moves to 0.55, the first answer had not settled.
Ideas that use this term