S1c IBM Shallow Direct-XY Result¶

Date: 2026-05-20

Backend: ibm_kingston

Status: completed

Artefacts:

data/s1_feedback_loop/s1c_xy_observable_readiness_ibm_kingston_20260520T131646Z.json
data/s1_feedback_loop/s1c_xy_observable_readiness_ibm_kingston_20260520T132455Z.json
data/s1_feedback_loop/s1c_xy_observable_raw_counts_ibm_kingston_20260520T132455Z.json
data/s1_feedback_loop/s1c_xy_observable_analysis_ibm_kingston_20260520T132455Z.json

Purpose¶

S1c is a continuation inside the same S1 paper. It is not a separate paper. It tests whether the S1/S1b weakness is partly caused by dynamic-circuit depth and feedback-policy strength.

Configuration¶

Field	Value
Dynamic rounds	1
Correction angle	0.06
Base gain	0.4
Observables	`XXI`, `YYI`, `IXX`, `IYY`
Shots per circuit	1024
Repetitions per arm/observable	3
Estimated QPU seconds	24.0
Max QPU seconds	120.0

Readiness Summary¶

The S1c feedback arms transpile with maximum depths 236-237, compared with the S1b three-round direct-XY feedback depths 719-720. The matched open-loop arms transpile with maximum depth 233. This makes S1c the clean next hardware run if the paper needs to separate policy/depth overhead from the underlying dynamic-feedback idea.

Submission Command¶

PYTHONDONTWRITEBYTECODE=1 .venv-linux/bin/python scripts/submit_s1b_ibm_xy_observable_pair.py \
  --backend ibm_kingston \
  --lane s1c \
  --experiment-id s1c_shallow_gain_tuned_xy_extension_2026-05-20 \
  --n-rounds 1 \
  --correction-angle 0.06 \
  --base-gain 0.4 \
  --repetitions 3 \
  --max-depth 900 \
  --timeout-s 7200 \
  --submit \
  --confirm-budget

Boundary¶

S1c is a same-paper extension. It was submitted after the readiness artefact was prepared. The original live runner completed the IBM jobs, then failed locally during packaging because the generic count extractor selected the monitor register instead of the final readout register for one-round circuits. The jobs were recovered from IBM without resubmission after fixing the extractor to prefer the final readout register.

Jobs¶

Observable	Feedback job	Matched open-loop job
`XXI`	`d86rca1789is73902fc0`	`d86rcgh789is73902fm0`
`YYI`	`d86rcc2s46sc73f7cf50`	`d86rcias46sc73f7cfh0`
`IXX`	`d86rcdgp0eas73dlc2j0`	`d86rcjp789is73902fsg`
`IYY`	`d86rcf0p0eas73dlc2lg`	`d86rclgp0eas73dlc2sg`

Result¶

Observable	Feedback minus control
`IXX`	-0.0364583333
`IYY`	-0.0319010417
`XXI`	-0.0221354167
`YYI`	-0.0292968750

Mean absolute feedback-control separation: 0.0299479167.

S1c does not rescue the feedback policy. Under the shallow one-round, lower-gain configuration, all four direct XY channels move negative for the feedback arm relative to the matched open-loop arm. This strengthens the same paper's conservative conclusion: the current monitored-feedback policy is not a robust controller on this backend/window.