Stim DetectorErrorModel Support

ILPQEC parses Stim DetectorErrorModels into an ILP by extracting the error(p) mechanisms and building parity-check and observable matrices.

Supported Instructions

• error(p) ... lines are parsed. Tags in error[...] are ignored.
• detector and logical_observable metadata lines are ignored.
• shift_detectors offsets are applied.
• repeat blocks are flattened by default (flatten_dem=True).
• detector_separator is unsupported and raises an error.

The ^ Separator

Stim uses ^ to describe alternative components of a correlated mechanism. ILPQEC currently treats ^ as whitespace, which discards the alternative structure. If your DEM contains ^, prefer decompose_errors=True when building the DEM so each error is a single unambiguous mechanism.

Flattening and Size

Flattening expands repeat blocks and applies shift_detectors. This is convenient for parsing but can increase the DEM size significantly. To fail fast instead, pass flatten_dem=False when creating the decoder.

Recommendations

• Use decompose_errors=True when constructing DEMs from Stim circuits.
• Use flatten_dem=True unless the DEM is very large, and then pre-flatten only when needed.

Equivalent Distance Analysis

ILPQEC also exposes dem_distance(...) for exact minimum-weight logical-fault search directly on a DEM.

import stim
from ilpqec import dem_distance

circuit = stim.Circuit.generated(
    "surface_code:rotated_memory_x",
    distance=3,
    rounds=3,
    after_clifford_depolarization=0.01,
)

dem = circuit.detector_error_model(decompose_errors=True)

result = dem_distance(dem, solver="highs")
targeted = dem_distance(dem, target_observables=[1], solver="highs")

print(result.distance)
print(targeted.observable_mask)

This objective ignores the error(p) probabilities stored in the DEM and minimizes the number of selected DEM mechanisms instead.