Check reasoning on cross talk in 2 qubit system is corrent

[fieldofnodes]

If I have a two-qubit system and hence the set of preparation states is defined as {|00>,|10>,|01>,|11>} and I measure in the same fashion {|00>,|10>,|01>,|11>}, then to measure readout error for crosstalk I would want to the associate probabilities of a A := Prep: |10> and measured: |10> B := Prep: |10> and measured: |11> C := Prep: |01> and measured: |01> D := Prep: |01> and measured: |11> Then I compute |A-B| and |C-D| where |##| is the absolute value.

Is cross-talk assumed to be symmetric? Meaning, should I expect ||A-B| - |C-D|| < epsilon, for epsilon > 0?

Or could there be cross-talk in a directional manner? Something like qubit 1 affects qubit 2, but qubit 2 does not affect qubit 1?

Finally, it would be nice to have a function, f, such that y = f(x), x is the calibration matrix and y is a rea-valued number that represents the degree of cross-talk and the function is the process to convert a calibration matrix into a cross-talk number.

I do not recall if this is described in the paper you shared with me.

Is my reasoning correct?

Cheers, Jon

[fieldofnodes]

There is no reason for cross-talk to be symmetric, in fact the Filip et. al. paper showed that it is asymmetric in the sense that it is possible for qubit A to induce an effect on the readout of qubit B while qubit B does not affect the measurement of qubit A at all. As far as I know, nobody did any quantification of the cross-talk, so it would be an extremely good result if we can somehow give a quantitative estimation, but it seems hard and probably there is no common function for all the qubits. Caution: I may be wrong and not fully aware of the recent development.