circuit.explain_detector_error_ model_errors() mixing Y and Z terms

[mpstafford]

I'm trying to use the explained Detector Error Model and have seen behaviour where the ExplainedError displays the wrong fault mechanism when the error contains Y and Z operators.

Consider the circuit below

stim.Circuit( """ R 0 1 PAULI_CHANNEL_2(i) 0 1 M 0 1 DETECTOR(0) rec[-2] DETECTOR(1) rec[-1] """

where PAULI_CHANNEL_2(i) has all zeros in the argument except for the i'th term. I would expect each detector to flip when the quibt being measured is hit by an X or Y component.

I get the DEM with

dem = circuit.detector_error_model( decompose_errors=False, approximate_disjoint_errors=False, allow_gauge_detectors=False )

and look at what is generated for each i. Indeed, when I construct the DEM I see the correct behaviour. However when I try to look at the explained DEM each time I see confusing behaviour.

For example, with i=1 (piy non-zero) I see

error(0.1000000000000000056) D1 detector(0) D0 detector(1) D1

And with i=2 (piz)

detector(0) D0 detector(1) D1

ExplainedError { dem_error_terms: D1[coords 1] CircuitErrorLocation { flipped_pauli_product: Y1 Circuit location stack trace: (after 0 TICKs) at instruction #2 (PAULI_CHANNEL_2) in the circuit at targets #1 to #2 of the instruction resolving to PAULI_CHANNEL_2(0, 0, 0.1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) 0 1 } }

I would expect the explanation for the Y error to be coupled with the non-trivial DEM. If I work through all the errors I see

Actual Error	DEM	Explained Error
IX	D1	D1 X1
IY	D1	[ ]
IZ	[ ]	D1 Y1
XI	D0	D0 X0
XX	D0 D1	D0 D1 X0*X1
XY	D0 D1	D0 X0*Z1
XZ	D0	D0 D1 X0*Y1
YI	D0	[ ]
YX	D0 D1	D1 Z0*X1
YY	D0 D1	[ ]
YZ	D0	D1 Z0*Y1
ZI	[ ]	D0 Y0
ZX	D1	D0 D1 Y0*X1
ZY	D1	D0 Y0*Z1
ZZ	[ ]	D0 D1 Y0*Y1

If you swap the Explained error rows interchanging Y - Z it would match up with what I expect.

[Strilanc]

Good catch. Fixed it. It'll be a half hour before it's on pypi probably, but after that if you pip install stim~=1.16.dev it should have the fix.