Equal-idle runs of 7 October 2026, recomputed

32 jobs on 7 October 2026 compared a readout arm with an idle arm of the same scheduled length on the same untouched qubits in the same job: ibm_pittsburgh 9, ibm_fez 3, ibm_aachen 2, ibm_kingston 18 (they are among the 305 jobs of 2 to 7 October, job list). One to 25 qubits were measured; the other qubits were read for phase and coherence on two axes, 400 shots per circuit, three jobs per family (two on aachen). No IBM jobs were run for this page: the existing data were read again with the estimators of the paper (median over untouched qubits, 95 % bootstrap over qubits with 2000 resamples, the four outcomes below, a copy test, coupling distance by breadth-first search over the saved coupling edges, a guard for qubits with no coherence left). This recomputation is not pre-registered. Its results are read beside the 21 and 22 October runs and never pooled with them; those runs differ in the number of measured qubits, the neighbour held in 1, the number of readouts and the copy test.

Inclusion rule

A run enters if and only if it has a readout arm and an idle arm of equal scheduled length on the same untouched qubits in the same job, reads their phase, coherence or concurrence, and has saved job identifiers and outputs. The rule was written before the earlier runs were inventoried (INCLUSION_RULE.md, SHA-256 92ee0717…, served as .text so the bytes match the hash) and every run that passes is reported whatever it shows. Two arms do not pass and are not read: the corner-layout arms whose readout was the shorter measure_2 instruction (no idle arm of equal length), and a whole-chip arm that prepared only one untouched qubit on the equator. A single-qubit chain on kingston (one untouched qubit, one measured qubit two couplings away) cannot enter a median over qubits (cannot tell).

Files

README.md (design, adaptations, sign check) · analyze_oct07.py · est.py · sign_check.py · summarize.py · results_oct07.txt (full log) · summary_table.txt · results_oct07.json · INCLUSION_RULE.md. Run: python3 -I analyze_oct07.py > results_oct07.txt. Job identifiers are in the full log.

Results (Table II of the paper)

Kick: slope of readout-minus-idle phase against the number of readouts N (degrees per readout; negative is the direction of the 21–22 October kick), median over untouched qubits with 95 % bootstrap interval, by coupling distance d to the nearest measured qubit. Coherence: slope of readout-minus-idle coherence against N, all kept qubits (negative is a loss). Outcomes as printed: B below 0, A above 0, C consistent with 0, CT cannot tell (fewer than five qubits, or the jobs disagree). n: kept qubits, all / d≤2 / d=3–4 / d≥5. Processors: P pittsburgh, F fez, Aa aachen, K kingston. The copies are separate jobs minutes to an hour apart, a harsher test than two copies in one job.

Layout, arm	n	kick, all	kick, d≤2	kick, d=3–4	kick, d≥5	coherence, all
P corner, 12 measured	76/4/5/67	+0.000 C [−0.082,+0.061]	CT	+0.036 C [−1.911,+0.380]	+0.003 C [−0.082,+0.065]	−0.000 C [−0.001,+0.000]
F corner, 12 measured	73/4/5/64	−0.036 C [−0.087,+0.063]	CT	−0.242 C [−0.423,+0.068]	−0.009 C [−0.055,+0.085]	−0.001 B [−0.002,−0.000]
Aa corner, 12 measured	75/4/5/66	+0.008 C [−0.127,+0.150]	CT	−0.171 C [−1.836,+0.340]	+0.030 C [−0.094,+0.158]	CT
K 25, measured in \|0⟩	57/25/17/15	−0.323 B [−0.622,−0.112]	−0.734 B [−12.857,−0.263]	−0.384 B [−1.096,−0.025]	−0.096 C [−0.267,+0.128]	−0.000 C [−0.001,+0.001]
K 25, in \|1⟩	56/25/16/15	−0.257 B [−0.692,−0.100]	−0.692 B [−17.486,−0.251]	−0.650 B [−1.450,−0.097]	−0.066 C [−0.124,+0.159]	−0.001 B [−0.001,−0.000]
K 25, on the equator	56/25/16/15	−0.302 B [−0.527,−0.109]	−0.846 B [−14.930,−0.307]	−0.146 C [−0.722,+0.174]	−0.106 C [−0.287,+0.091]	−0.000 C [−0.001,+0.000]
K 25 (2nd run), \|0⟩	58/25/18/15	−0.338 B [−0.873,−0.052]	−1.024 B [−12.692,−0.161]	−0.682 C [−1.280,+0.445]	−0.024 C [−0.157,+0.204]	−0.002 C [−0.003,+0.000]
K 25 (2nd run), \|1⟩	58/25/18/15	−0.466 B [−0.712,−0.076]	−1.133 B [−17.225,−0.082]	−0.302 C [−1.532,+0.414]	CT	−0.001 C [−0.003,+0.000]
K 25 (2nd run), equator	58/25/18/15	−0.577 B [−1.005,−0.298]	−1.378 B [−13.451,−0.346]	−0.357 C [−1.005,+0.047]	−0.305 C [−0.680,+0.001]	−0.003 B [−0.004,−0.001]
K 25, 3 µs gaps, \|0⟩	57/25/17/15	−0.572 B [−1.000,−0.114]	−1.270 B [−12.614,−0.602]	−0.124 C [−1.283,+0.281]	+0.329 C [−0.572,+0.647]	+0.001 C [−0.001,+0.002]
K 25, 3 µs gaps, \|1⟩	57/25/17/15	−0.434 B [−0.708,−0.186]	−0.722 B [−12.588,−0.351]	−0.510 B [−1.335,−0.117]	+0.152 C [−0.448,+0.435]	CT
K 25, 3 µs gaps, equator	57/25/17/15	−0.437 B [−0.718,−0.011]	−1.192 C [−12.673,+0.041]	−0.509 C [−1.149,+0.055]	+0.039 C [−0.437,+0.449]	+0.000 C [−0.001,+0.002]
K 25, role rotation 0	58/25/18/15	−0.428 B [−0.758,−0.290]	−1.116 B [−18.259,−0.411]	−0.416 C [−1.013,+0.143]	+0.004 C [−0.360,+0.314]	+0.000 C [−0.002,+0.002]
K 25, role rotation 1	58/25/18/15	−0.290 B [−0.966,−0.075]	−1.211 B [−13.749,−0.033]	−0.563 B [−1.536,−0.128]	+0.039 C [−0.311,+0.253]	CT
K 25, role rotation 2	58/25/18/15	−0.518 B [−0.928,−0.319]	−1.124 B [−16.048,−0.303]	−0.767 B [−1.229,−0.202]	−0.173 C [−0.417,+0.001]	−0.001 C [−0.004,+0.001]
K 25, index order	58/25/18/15	−0.386 B [−0.894,−0.144]	−1.066 B [−13.476,−0.489]	−0.520 B [−1.699,−0.083]	+0.239 C [−0.041,+0.524]	−0.001 C [−0.003,+0.001]
K 25, reversed order	58/25/18/15	−0.450 C [−0.945,+0.064]	−1.192 C [−12.986,+0.019]	−0.702 C [−1.537,+0.068]	+0.175 A [+0.036,+0.311]	−0.000 C [−0.003,+0.001]
K whole chip, set A (20)	64/20/32/12	−0.688 B [−1.182,−0.345]	−2.840 B [−5.900,−0.515]	−0.490 B [−0.789,−0.062]	−0.455 C [−1.063,+0.133]	−0.000 C [−0.004,+0.002]
K whole chip, set B (19)	64/18/30/16	−0.811 B [−1.128,−0.214]	−3.129 B [−5.545,−0.449]	−0.707 B [−1.091,−0.214]	+0.024 C [−1.128,+0.229]	−0.001 C [−0.003,+0.002]
K whole chip, set C (1)	64/3/8/53	−0.138 C [−0.309,+0.115]	CT	−0.134 C [−0.590,+0.431]	−0.145 C [−0.345,+0.130]	+0.000 C [−0.002,+0.001]
P sparse set 0 (5)	84/15/30/39	−0.169 C [−0.288,+0.015]	−0.330 B [−0.813,−0.169]	−0.087 C [−0.267,+0.274]	−0.099 C [−0.372,+0.118]	−0.001 C [−0.002,+0.000]
P sparse set 1 (7)	84/15/34/35	−0.235 C [−0.358,+0.017]	−0.410 C [−0.623,+0.111]	−0.142 C [−0.492,+0.147]	−0.236 C [−0.373,+0.041]	−0.001 C [−0.003,+0.001]
P 25, measured in \|0⟩	58/26/16/16	−0.224 B [−0.400,−0.051]	−0.433 B [−2.950,−0.166]	−0.223 C [−0.608,+0.029]	+0.280 C [−0.134,+0.447]	CT
P 25, in \|1⟩	58/26/16/16	−0.356 B [−0.751,−0.137]	−0.796 B [−3.230,−0.225]	−0.570 B [−0.937,−0.106]	−0.093 C [−0.255,+0.270]	CT
P 25, on the equator	59/27/16/16	−0.334 B [−0.647,−0.053]	−0.826 B [−2.865,−0.054]	−0.558 C [−0.821,+0.111]	+0.009 C [−0.368,+0.364]	−0.001 C [−0.004,+0.001]

What the table says

On kingston the 14 arms with 25 measured qubits (six from the same three jobs, all on overlapping untouched qubits) read the kick over all untouched qubits below zero in 13 of them (0.26 to 0.58° per readout), with the measured qubits in 0, in 1 or on the equator, back to back or with 3 µs gaps, in all three role rotations and in index order; within two couplings it is below zero in 12 of 14 arms (0.69 to 1.38°, intervals heavy-tailed on the low side). The whole-chip measured sets of 20 and 19 qubits give −0.69 and −0.81° overall and −2.84 and −3.13° within two couplings.
At five or more couplings no arm is below zero; the reversed-order arm reads above zero, +0.175 [+0.036, +0.311]° (15 qubits), the only interval above zero among the 25 arms at five or more couplings; with about 100 distance-bin readings in this table it is not distinguished from the false-reading floor. A single measured qubit (whole-chip preparation, set C) gives −0.138 [−0.309, +0.115], consistent with zero.
On pittsburgh the 25-qubit layout is below zero overall in all three states (−0.22 to −0.36°) and within two couplings (−0.43 to −0.83°); the sparse sets of 5 and 7 measured qubits are consistent with zero overall and below zero within two couplings for the first set only (−0.330 [−0.813, −0.169]).
The corner layout with 12 measured qubits, where 64 to 67 of the 73 to 76 kept untouched qubits are five or more couplings away, is consistent with zero overall on pittsburgh (+0.000 [−0.082, +0.061]), fez (−0.036 [−0.087, +0.063]) and aachen (+0.008 [−0.127, +0.150]), with only four qubits within two couplings (cannot tell).
Pooling the plain 0 arms per processor (untouched qubits repeat across families, so n counts repeated measures) gives pittsburgh −0.136 [−0.199, −0.063] overall and −0.094 [−0.147, −0.026] at five or more couplings; the second interval treats repeated measures of the same qubits as independent and is too narrow, and no single pittsburgh family is below zero at five or more couplings; and kingston −0.433 [−0.633, −0.221] overall and −0.065 [−0.284, +0.085] at five or more. The per-family readings are the readings of record.
With the measured qubits in 1, each against its own idle in 1, the kick on kingston and pittsburgh is below zero overall in every arm, with sizes in the range of the 0 arms.
Coherence: the slope over all untouched qubits is below zero in 3 of 25 arms (fez corner, −0.0008 per readout; kingston 25 in 1; kingston second run on the equator), cannot tell in 5 and consistent with zero in 17; at N = 4 it is below zero on aachen (−0.010 [−0.015, −0.003]), in the kingston equator arm (−0.007) and in the pittsburgh 1 arm (−0.009). The aachen N = 4 reading sits beside a slope whose copies shifted. These runs resolve a coherence cost in only 3 of 25 arms; at 4 readouts 20 of the 22 arms with a reading have negative medians (arms share qubits, so these are not independent).
On kingston, the processor in both, the 7 October overall medians (−0.26 to −0.81°; 19 to 25 measured, up to 9 readouts) and the 21–22 October medians (2.41 to 3.24° in size; 39 measured, every untouched qubit within two couplings) are of the same sign and are not compared as one quantity.

Bones · Wayfinder · Sage · Shuttle · Nova
Heuremen LLC
“We found it.”