Leakage

Leakage model, simulator and sampler

15/15 passed in 11929ms

Test	What it does	Result
test_arg_validation	Rates outside [0, 1] and unknown kick names are refused at construction.	✅ pass
test_kick_channels	One partner channel per axis; only the twirled arm is Pauli.	✅ pass
test_leaked_is_inert	A leaked qubit takes no gate while its neighbours do, and measuring it still adds one record bit.	✅ pass
test_matches_plain_sim	At leak=0 the leaky simulator tracks a plain one gate for gate, with weight 1.	✅ pass
test_readout_bias	readout pins P(measure 1 | leaked): 1.0 always reads 1, 0.0 always reads 0.	✅ pass
test_reset_clears_flag	A reset is the leakage-reduction unit: both MR and R clear the flag.	✅ pass
test_seep_returns_excited	Seeping back returns the qubit in |1>, where the |2> level decays to.	✅ pass
test_ideal_when_off	With no Pauli noise and no leakage every detector and observable stays 0.	✅ pass
test_leak_damages	Leakage alone damages a noiseless surface code (its X ancillas sit in |+>), and the detection rate rises with the leak rate.	✅ pass
test_needs_excitation	Leakage cannot fire where the gate pair is never both-excited.	✅ pass
test_twirl_weights_unit	The twirled arm carries unit weights; the coherent arm carries real ones.	✅ pass
test_zero_rate_identical	At leak=0 the leaky sampler reproduces the plain Python sampler bit for bit.	✅ pass
test_kick_keeps_support	Twirling the kick changes the damage, not which syndromes escape col(H).	✅ pass
test_leakage_outside_colspace	Leakage syndromes leave col(H), so no DEM at any priors can explain them.	✅ pass
test_pauli_inside_colspace	Every Pauli syndrome lies in the column space of the DEM check matrix.	✅ pass