Twirled Readout Error eXtinction

Twirled Readout Error eXtinction (TREX) is a readout error mitigation technique that applies random X gates before measurement (readout twirling) and uses calibration data to correct the resulting expectation values. Unlike REM, TREX does not require an explicit inverse confusion matrix. Instead, it estimates readout error correction factors directly from calibration circuits. In this sense, TREX is “model-free”: it does not need a pre-characterized noise model (e.g., a confusion matrix) as input. The calibration circuits are used to estimate correction factors on-the-fly, not to build a reusable noise model.

For more discussion of the theory of TREX, see the section What is the theory behind TREX?.

Below you can find sections of the documentation that address the following questions:

• How do I use TREX?
• When should I use TREX?
• What additional options are available when using TREX?
• What happens when I use TREX?
• What is the theory behind TREX?