S1f IBM Quadrature Mechanism Check¶
Date: 2026-05-20
Backend: ibm_kingston
Experiment ID: s1f_quadrature_mechanism_check_2026-05-20
Parent experiment: s1_dynamic_feedback_preregistration_2026-05-06
Artefacts¶
- Readiness:
data/s1_feedback_loop/s1f_xy_observable_readiness_ibm_kingston_20260520T142213Z.json - Readiness SHA-256:
d8b5e0980ab3966ec62eb29f25192cd387b4f2bc420126ea0f8f9d9fbc45d3ea - Raw counts:
data/s1_feedback_loop/s1f_xy_observable_raw_counts_ibm_kingston_20260520T142213Z.json - Raw counts SHA-256:
1be95f9996fb56e1f7e8d39921e9c46c65b1d8f4622bb46ac7d3789a592d9e87 - Analysis:
data/s1_feedback_loop/s1f_xy_observable_analysis_ibm_kingston_20260520T142213Z.json - Analysis SHA-256:
17b37753460137c796a5fb8c52784ce694d7ce4ffb63e55c92853abe8f91313e
Purpose¶
S1f is a narrow mechanism-discrimination lane for the S1 paper. It keeps only
the S1d/S1e-favoured policy, current_shallow_positive, and tests whether the
previously repeated YYI response localises to one quadrature or instead
appears across transverse or population controls.
Configuration:
n_rounds=1- correction angle
0.06 - base gain
0.4 - repetitions
5 - observables:
YYI,XXI,XYI,YXI,ZZI
Readiness¶
All arms passed readiness on ibm_kingston. Maximum transpiled depth was
237. Total estimated QPU seconds were 50.0 under the 60.0 second ceiling.
Execution¶
The 10 IBM jobs completed:
d86s75h789is73903oug, d86s779789is73903p00,
d86s791789is73903p60, d86s7atg7okc73eli47g,
d86s7d2s46sc73f7dko0, d86s7eop0eas73dldb7g,
d86s7gop0eas73dldbeg, d86s7m8p0eas73dldbp0,
d86s7odg7okc73eli530, d86s7q8p0eas73dldbtg.
Results¶
Feedback minus matched open-loop control:
| Observable | Feedback | Control | Feedback minus control |
|---|---|---|---|
XXI |
0.2187500000 |
0.2175781250 |
0.0011718750 |
XYI |
0.1617187500 |
0.1566406250 |
0.0050781250 |
YXI |
0.2242187500 |
0.2136718750 |
0.0105468750 |
YYI |
-0.0445312500 |
-0.0242187500 |
-0.0203125000 |
ZZI |
0.6523437500 |
0.6480468750 |
0.0042968750 |
Mean absolute feedback-control separation: 0.0082812500.
Interpretation¶
S1f does not support a stable quadrature-localised positive YYI mechanism in
this later calibration window. The previously repeated S1d/S1e YYI response
was large and positive, but S1f measures a small negative YYI delta while the
cross-quadrature (XYI, YXI) and population (ZZI) controls remain small.
The conservative conclusion is stronger and more honest than a promoted controller claim: the S1 direct-XY response is structured and can repeat within a calibration window, but it is not calibration-window stable enough to promote the tested feedback law as a robust controller. S1f should be the final hardware run for this paper unless a future redesigned policy is preregistered.