Gate G25 · Nexus cross-check

Three Nexus lanes, one decoder

Five representative circuits from the existing gates were re-run on every Quantinuum Nexus backend this account can reach, and compared against the local Selene baseline with the same observable decoder. Each cell carries a job id and a meter record; the tolerance is the binomial 4σ band, 4·√(0.5/shots) = 0.125 at 512 shots.

12/15 cells PASS0.00 HQC billed · 60 ceilingmode: liveH2-1LEH2-EmulatorHelios-1E-lite

Circuit × device grid

Ref	Circuit	Selene	H2-1LE	H2-Emulator	Helios-1E-lite
G16	QPDE ethylene interference cell (k=4, beta=1/2) 512 shots	0.0000	verdict verified`nexus_cross_check_verified`4/4 criteria Each criterion below is recomputed from the committed rows by quantum/verdicts.py; this chip reads the result rather than restating it. pass`all_lanes_reachable` Every declared emulator lane was reachable for this run. measured [] · threshold [] pass`all_rows_pass` Every scored cross-check row lands inside its tolerance against the Selene baseline. Rows with no target (QAOA reference lanes) are reported separately, not scored. measured [] · threshold [] pass`reference_rows_declared` Rows carrying no target are explicitly marked as reference rather than counted as passes. measured ["G22"] · threshold ["G22"] pass`no_spend` The cross-check was billed zero HQC: emulator lanes only. measured 0 · threshold 0 0.0000 PASS Δ 0.0000 / tol 0.125	verdict verified`nexus_cross_check_verified`4/4 criteria Each criterion below is recomputed from the committed rows by quantum/verdicts.py; this chip reads the result rather than restating it. pass`all_lanes_reachable` Every declared emulator lane was reachable for this run. measured [] · threshold [] pass`all_rows_pass` Every scored cross-check row lands inside its tolerance against the Selene baseline. Rows with no target (QAOA reference lanes) are reported separately, not scored. measured [] · threshold [] pass`reference_rows_declared` Rows carrying no target are explicitly marked as reference rather than counted as passes. measured ["G22"] · threshold ["G22"] pass`no_spend` The cross-check was billed zero HQC: emulator lanes only. measured 0 · threshold 0 0.0000 PASS Δ 0.0000 / tol 0.125	verdict verified`nexus_cross_check_verified`4/4 criteria Each criterion below is recomputed from the committed rows by quantum/verdicts.py; this chip reads the result rather than restating it. pass`all_lanes_reachable` Every declared emulator lane was reachable for this run. measured [] · threshold [] pass`all_rows_pass` Every scored cross-check row lands inside its tolerance against the Selene baseline. Rows with no target (QAOA reference lanes) are reported separately, not scored. measured [] · threshold [] pass`reference_rows_declared` Rows carrying no target are explicitly marked as reference rather than counted as passes. measured ["G22"] · threshold ["G22"] pass`no_spend` The cross-check was billed zero HQC: emulator lanes only. measured 0 · threshold 0 0.0020 PASS Δ 0.0020 / tol 0.125
G24a	Simon n=3 -- Nadarasa rewriter form 512 shots	1.0000	verdict verified`nexus_cross_check_verified`4/4 criteria Each criterion below is recomputed from the committed rows by quantum/verdicts.py; this chip reads the result rather than restating it. pass`all_lanes_reachable` Every declared emulator lane was reachable for this run. measured [] · threshold [] pass`all_rows_pass` Every scored cross-check row lands inside its tolerance against the Selene baseline. Rows with no target (QAOA reference lanes) are reported separately, not scored. measured [] · threshold [] pass`reference_rows_declared` Rows carrying no target are explicitly marked as reference rather than counted as passes. measured ["G22"] · threshold ["G22"] pass`no_spend` The cross-check was billed zero HQC: emulator lanes only. measured 0 · threshold 0 1.0000 PASS Δ 0.0000 / tol 0.125	verdict verified`nexus_cross_check_verified`4/4 criteria Each criterion below is recomputed from the committed rows by quantum/verdicts.py; this chip reads the result rather than restating it. pass`all_lanes_reachable` Every declared emulator lane was reachable for this run. measured [] · threshold [] pass`all_rows_pass` Every scored cross-check row lands inside its tolerance against the Selene baseline. Rows with no target (QAOA reference lanes) are reported separately, not scored. measured [] · threshold [] pass`reference_rows_declared` Rows carrying no target are explicitly marked as reference rather than counted as passes. measured ["G22"] · threshold ["G22"] pass`no_spend` The cross-check was billed zero HQC: emulator lanes only. measured 0 · threshold 0 1.0000 PASS Δ 0.0000 / tol 0.125	verdict verified`nexus_cross_check_verified`4/4 criteria Each criterion below is recomputed from the committed rows by quantum/verdicts.py; this chip reads the result rather than restating it. pass`all_lanes_reachable` Every declared emulator lane was reachable for this run. measured [] · threshold [] pass`all_rows_pass` Every scored cross-check row lands inside its tolerance against the Selene baseline. Rows with no target (QAOA reference lanes) are reported separately, not scored. measured [] · threshold [] pass`reference_rows_declared` Rows carrying no target are explicitly marked as reference rather than counted as passes. measured ["G22"] · threshold ["G22"] pass`no_spend` The cross-check was billed zero HQC: emulator lanes only. measured 0 · threshold 0 0.9980 PASS Δ 0.0020 / tol 0.125
G24b	Simon n=3 -- TKET level-2 compiled form 512 shots	1.0000	verdict verified`nexus_cross_check_verified`4/4 criteria Each criterion below is recomputed from the committed rows by quantum/verdicts.py; this chip reads the result rather than restating it. pass`all_lanes_reachable` Every declared emulator lane was reachable for this run. measured [] · threshold [] pass`all_rows_pass` Every scored cross-check row lands inside its tolerance against the Selene baseline. Rows with no target (QAOA reference lanes) are reported separately, not scored. measured [] · threshold [] pass`reference_rows_declared` Rows carrying no target are explicitly marked as reference rather than counted as passes. measured ["G22"] · threshold ["G22"] pass`no_spend` The cross-check was billed zero HQC: emulator lanes only. measured 0 · threshold 0 1.0000 PASS Δ 0.0000 / tol 0.125	verdict verified`nexus_cross_check_verified`4/4 criteria Each criterion below is recomputed from the committed rows by quantum/verdicts.py; this chip reads the result rather than restating it. pass`all_lanes_reachable` Every declared emulator lane was reachable for this run. measured [] · threshold [] pass`all_rows_pass` Every scored cross-check row lands inside its tolerance against the Selene baseline. Rows with no target (QAOA reference lanes) are reported separately, not scored. measured [] · threshold [] pass`reference_rows_declared` Rows carrying no target are explicitly marked as reference rather than counted as passes. measured ["G22"] · threshold ["G22"] pass`no_spend` The cross-check was billed zero HQC: emulator lanes only. measured 0 · threshold 0 1.0000 PASS Δ 0.0000 / tol 0.125	verdict verified`nexus_cross_check_verified`4/4 criteria Each criterion below is recomputed from the committed rows by quantum/verdicts.py; this chip reads the result rather than restating it. pass`all_lanes_reachable` Every declared emulator lane was reachable for this run. measured [] · threshold [] pass`all_rows_pass` Every scored cross-check row lands inside its tolerance against the Selene baseline. Rows with no target (QAOA reference lanes) are reported separately, not scored. measured [] · threshold [] pass`reference_rows_declared` Rows carrying no target are explicitly marked as reference rather than counted as passes. measured ["G22"] · threshold ["G22"] pass`no_spend` The cross-check was billed zero HQC: emulator lanes only. measured 0 · threshold 0 0.9980 PASS Δ 0.0020 / tol 0.125
G22	NHS discharge QAOA, 2 patients x 2 placements 512 shots	0.1895	verdict verified`nexus_cross_check_verified`4/4 criteria Each criterion below is recomputed from the committed rows by quantum/verdicts.py; this chip reads the result rather than restating it. pass`all_lanes_reachable` Every declared emulator lane was reachable for this run. measured [] · threshold [] pass`all_rows_pass` Every scored cross-check row lands inside its tolerance against the Selene baseline. Rows with no target (QAOA reference lanes) are reported separately, not scored. measured [] · threshold [] pass`reference_rows_declared` Rows carrying no target are explicitly marked as reference rather than counted as passes. measured ["G22"] · threshold ["G22"] pass`no_spend` The cross-check was billed zero HQC: emulator lanes only. measured 0 · threshold 0 0.1582 reference	verdict verified`nexus_cross_check_verified`4/4 criteria Each criterion below is recomputed from the committed rows by quantum/verdicts.py; this chip reads the result rather than restating it. pass`all_lanes_reachable` Every declared emulator lane was reachable for this run. measured [] · threshold [] pass`all_rows_pass` Every scored cross-check row lands inside its tolerance against the Selene baseline. Rows with no target (QAOA reference lanes) are reported separately, not scored. measured [] · threshold [] pass`reference_rows_declared` Rows carrying no target are explicitly marked as reference rather than counted as passes. measured ["G22"] · threshold ["G22"] pass`no_spend` The cross-check was billed zero HQC: emulator lanes only. measured 0 · threshold 0 0.1855 reference	verdict verified`nexus_cross_check_verified`4/4 criteria Each criterion below is recomputed from the committed rows by quantum/verdicts.py; this chip reads the result rather than restating it. pass`all_lanes_reachable` Every declared emulator lane was reachable for this run. measured [] · threshold [] pass`all_rows_pass` Every scored cross-check row lands inside its tolerance against the Selene baseline. Rows with no target (QAOA reference lanes) are reported separately, not scored. measured [] · threshold [] pass`reference_rows_declared` Rows carrying no target are explicitly marked as reference rather than counted as passes. measured ["G22"] · threshold ["G22"] pass`no_spend` The cross-check was billed zero HQC: emulator lanes only. measured 0 · threshold 0 0.1934 reference
G20	Floquet chain n=6, 2 cycles, eps=0 512 shots	1.0000	verdict verified`nexus_cross_check_verified`4/4 criteria Each criterion below is recomputed from the committed rows by quantum/verdicts.py; this chip reads the result rather than restating it. pass`all_lanes_reachable` Every declared emulator lane was reachable for this run. measured [] · threshold [] pass`all_rows_pass` Every scored cross-check row lands inside its tolerance against the Selene baseline. Rows with no target (QAOA reference lanes) are reported separately, not scored. measured [] · threshold [] pass`reference_rows_declared` Rows carrying no target are explicitly marked as reference rather than counted as passes. measured ["G22"] · threshold ["G22"] pass`no_spend` The cross-check was billed zero HQC: emulator lanes only. measured 0 · threshold 0 1.0000 PASS Δ 0.0000 / tol 0.125	verdict verified`nexus_cross_check_verified`4/4 criteria Each criterion below is recomputed from the committed rows by quantum/verdicts.py; this chip reads the result rather than restating it. pass`all_lanes_reachable` Every declared emulator lane was reachable for this run. measured [] · threshold [] pass`all_rows_pass` Every scored cross-check row lands inside its tolerance against the Selene baseline. Rows with no target (QAOA reference lanes) are reported separately, not scored. measured [] · threshold [] pass`reference_rows_declared` Rows carrying no target are explicitly marked as reference rather than counted as passes. measured ["G22"] · threshold ["G22"] pass`no_spend` The cross-check was billed zero HQC: emulator lanes only. measured 0 · threshold 0 0.9980 PASS Δ 0.0020 / tol 0.125	verdict verified`nexus_cross_check_verified`4/4 criteria Each criterion below is recomputed from the committed rows by quantum/verdicts.py; this chip reads the result rather than restating it. pass`all_lanes_reachable` Every declared emulator lane was reachable for this run. measured [] · threshold [] pass`all_rows_pass` Every scored cross-check row lands inside its tolerance against the Selene baseline. Rows with no target (QAOA reference lanes) are reported separately, not scored. measured [] · threshold [] pass`reference_rows_declared` Rows carrying no target are explicitly marked as reference rather than counted as passes. measured ["G22"] · threshold ["G22"] pass`no_spend` The cross-check was billed zero HQC: emulator lanes only. measured 0 · threshold 0 0.9948 PASS Δ 0.0052 / tol 0.125

G22 is a reference row rather than a pass/fail one: the QAOA energy on a two-patient discharge instance is a distribution statistic, not a single analytic target, so it is compared against the emulator value rather than a closed form.

No hardware leg — and why

No H-series hardware backend is reachable for this account; every lane below is an emulator. The --device H2-1 submission path is implemented and guarded by max_cost, but unexecuted.

The device census this account returns is: aer_simulator, Aer, aer_simulator_statevector, AerState, aer_simulator_unitary, AerUnitary, sv1, Braket, Helios-1E-lite, H1-1LE, Quantinuum, H2-1LE, H1-Emulator, H2-Emulator, Qulacs, Selene, SelenePlus. Neither H2-1, H2-1E nor H1-1 appears, so no lane here ran on a QPU.

Helios-1E-lite now runs. Its earlier "You do not have access to this machine (code: 14)" was our bug, not an entitlement gap: HeliosConfig() defaults to system_name="Helios-1" — the hardware system — so every submission asked for a QPU. Carrying the device name into system_name, and supplying the required HeliosEmulatorConfig(n_qubits=…), makes the emulator accept the job. Its results decode as a QsysResult with Guppy-style tagged entries, not a pytket distribution.

The --device H2-1 submission path is implemented and guarded by max_cost; it fires unchanged the moment hardware access is granted on the account.

Per-job meters

Device	Mode	Qubits	Shots	Seed	Est. HQC	Job id
H2-1LE	live	2	512	25	5.512	8248d834…
H2-1LE	live	6	512	25	5.512	4ff92f6e…
H2-1LE	live	6	512	25	5.3072	faa322bb…
H2-1LE	live	4	512	25	5.8192	c2a3e361…
H2-1LE	live	6	512	25	7.048	811e2229…
H2-Emulator	live	2	512	25	5.512	c55c7ccc…
H2-Emulator	live	6	512	25	5.512	cbd39908…
H2-Emulator	live	6	512	25	5.3072	362f7c6a…
H2-Emulator	live	4	512	25	5.8192	827e6ce5…
H2-Emulator	live	6	512	25	7.048	b3dbc461…
Helios-1E-lite	live	2	512	25	5.512	bb107bd4…
Helios-1E-lite	live	6	512	25	5.3072	698cb9c8…
Helios-1E-lite	live	6	512	25	5.3072	476b6381…
Helios-1E-lite	live	4	512	25	5.8192	06186044…
Helios-1E-lite	live	6	512	25	7.048	35467bbc…

Billed cost is 0.0 HQC on every lane: these are emulators, free for this account. The estimate is still recorded so the same table reads honestly the day a hardware job appears in it.

What the third lane taught us

Program form is per-backend. The *LE local emulators and Helios reject pytket circuits outright and take an uploaded HUGR package (a compiled Guppy program) instead; the H2-Emulator takes a pytket circuit through a compile job. The driver now picks the form from the device name.
Compile jobs only accept circuits. A HUGR program is executed as uploaded — and some backends refuse compile jobs entirely with "Compilation jobs are not supported for this backend", which is a routing fact, not a circuit bug.
Do not score runtime ops against a native gate set. QAlloc, MeasureFree and target-extension ops are not gates; scoring them produced a false "device does not accept" rejection on a circuit the device was happy to run.
A failed lane must not kill the sweep. Each cell captures its own error, keeps its job ids, and the run continues — otherwise one inaccessible machine costs you the two lanes that worked.

Reproduce with python -m quantum.gate25.run --devices H2-1LE,H2-Emulator --execute; job ids live under quantum/gate25/_cache_jobs/<device>/ and are re-attached, never resubmitted, by python -m quantum.resume --gate G25. Field notes are on the Nexus page.