CvAnchors

Tier A: closed-form CV key rates vs published theory numbers

6/6 passed in 1ms

Test	What it does	Result
test_jain_2022_heterodyne	Jain 2022 (arXiv:2110.09262, 20 km LLO, heterodyne): asymptotic $K\approx 0.067$ bit/symbol, above their composable 0.027.	✅ pass
test_lodewyck_2007_holevo	Lodewyck 2007: Holevo bound ${\chi }_{BE}=0.9020$, their printed "316 kb/s" over the same 350 kHz rate.	✅ pass
test_lodewyck_2007_key_rate	Lodewyck 2007: $K=\beta I_{AB}-{\chi }_{BE}=0.0352$ bit/symbol, their printed collective-attack "12.3 kb/s". Their 2 kb/s headline is a net rate after reconciliation throughput and is not this quantity.	✅ pass
test_lodewyck_2007_mutual_information	Lodewyck 2007 (arXiv:0706.4255, 25 km, homodyne, trusted): $I_{AB}=1.0436$ bit/symbol, their "365 kb/s" over the 350 kHz effective rate. Not their printed 1.045, which that same 365 kb/s does not reproduce.	✅ pass
test_untrusted_is_more_pessimistic	Handing the detector to Eve never raises the key rate on the same hardware.	✅ pass
test_zhang_2020_worst_case	Zhang 2020 (arXiv:2001.02555, 202.81 km): asymptotic $K$ at worst-case $\xi =0.0383$ is 5.5e-5 bit/symbol, above their finite-size value.	✅ pass