ProtocolRelayGaussian
Phase 5c API: the continuous-variable relay, where the detector is Eve's
8/8 passed in 1ms
| Test | What it does | Result |
|---|---|---|
test_corr_not_refused | Unequal arms fold to the four-mode model to 1e-11, and the same correlation unreferred is refused by the uncertainty principle. | ✅ pass |
test_corr_refers | The CorrelatedEnvironment is rescaled by 3/7 and the folded relay reproduces the four-mode model to 1e-11, where unreferred chi reads 9.6 against 12.8 and the rate rises by 0.5514 bit per use. | ✅ pass |
test_correlated_eve | A correlated environment of the helping sign drives chi below its pure-loss floor and raises the rate; one the uncertainty principle cannot support is refused. | ✅ pass |
test_engine_agreement | Every reported number is bit for bit what _core.cvmdi_noise, cvmdi_floor, cvmdi_least, cvmdi_rate and cvmdi_point return for the two arms the report names. | ✅ pass |
test_floor_is_not_least | Pure loss lands chi on cvmdi_floor, cvmdi_least sits strictly below it (exactly 4 for equal arms whatever the loss), and excess noise pushes chi above the floor. | ✅ pass |
test_guards | Asymmetric modulation, a finite-size reduction and a pulse width no formula consumes are each refused by this topology. | ✅ pass |
test_position_matters | chi is symmetric under swapping the arms and the rate is not: away from Alice costs more than away from Bob, and the rate falls monotonically with reach. | ✅ pass |
test_relay_belongs_to_eve | A relay detector has no trusted variant: eta folds into the arm as loss and v_el as noise, and an input variance through arm-then-detector lands on what the reported (tau, omega) predicts. | ✅ pass |