verify_xtrain.py -- Sec. sec:xtrain numbers recomputed from raw IBM job results (data/). Bootstrap NB=200, seed 20261022. JOBS USED (chip run job_id submitted): ibm_pittsburgh run 91 davuanc92g1c7399i8kg 2026-10-02T13:09:49-04:00 ibm_pittsburgh run 92 davveh04oijs73e88tsg 2026-10-02T14:26:12-04:00 ibm_pittsburgh run 93 davvf7c92g1c7399k1h0 2026-10-02T14:27:41-04:00 ibm_pittsburgh run 121 db00t8k92g1c7399m6hg 2026-10-02T16:05:54-04:00 ibm_pittsburgh run 361 db16p5hb694s73drsna0 2026-10-04T11:11:18-04:00 ibm_marrakesh run 141 db023m492g1c7399o9vg 2026-10-02T17:27:52-04:00 ibm_marrakesh run 301 db15r4bid5ic73eqmhs0 2026-10-04T10:07:13-04:00 ibm_marrakesh run 302 db15s4pb694s73drrg10 2026-10-04T10:09:24-04:00 ibm_marrakesh run 303 db15skrid5ic73eqmkvg 2026-10-04T10:10:28-04:00 ibm_kingston run 141 db023hal7guc73cfnvl0 2026-10-02T17:27:33-04:00 ibm_kingston run 301 db15t52vog1s73fhmo3g 2026-10-04T10:11:32-04:00 ibm_kingston run 302 db15u31b694s73drrjvg 2026-10-04T10:13:32-04:00 ibm_kingston run 303 db161j6egvvc73bh6u50 2026-10-04T10:21:00-04:00 4 October readable qubits per job: marrakesh [136, 138, 137]; kingston [142, 146, 147]; pittsburgh [149] start-1 over start-0 excess (minus sign dropped), against ONE (the only bookkeeping for which the ratio is defined; against CURVE the start-0 excess crosses zero): marrakesh 5 us 0.901 10 us 0.843 20 us 0.596 kingston 5 us 0.940 10 us 0.745 20 us 0.620 pittsburgh 5 us 0.973 10 us 0.633 20 us 0.368 info: the paper width 0.012 on the 10-gate prediction needs a shortfall width of 0.142 deg (fit gives 0.119 deg) where quantity paper recomputed status note l.229 ON left alone 10 us level (run 121) +0.9463 +0.946341 ok l.229 ON left alone 30 us level (run 121) +0.8609 +0.860915 ok l.229 ON left alone 90 us level (run 121) +0.6846 +0.684637 ok l.229 ON left alone 300 us level (run 121) +0.3286 +0.328609 ok l.229 ON 10 us minus exponential through 300 us -0.0141 -0.014055 ok l.229 ON 30 us minus exponential through 300 us -0.0259 -0.025861 ok l.229 ON 90 us minus exponential through 300 us -0.0185 -0.018500 ok l.229 readable qubits in run 121 ("150 qubits") +150 +150.00 ok of 156 xtrain pittsburgh 3 jobs, start 0, 10 gates/300 us, vs ONE +0.0200 +0.020001 ok xtrain width +0.0016 +0.001465 ok(boot) xtrain pittsburgh 4th job (run 121), start 0, vs ONE +0.0126 +0.012633 ok xtrain width +0.0032 +0.003510 ok(boot) xtrain "about two widths apart" (diff / combined width) +2 +1.94 ok ratio, not a level xtrain marrakesh (1 job), start 0, vs ONE +0.0274 +0.027421 ok xtrain width +0.0028 +0.002671 ok(boot) xtrain kingston (1 job), start 0, vs ONE +0.0287 +0.028748 ok xtrain width +0.0028 +0.002937 ok(boot) xtrain pittsburgh start 1, vs ONE +0.0005 +0.000473 ok xtrain width +0.0013 +0.001276 ok(boot) xtrain marrakesh start 1 -0.0016 -0.001645 ok xtrain width +0.0021 +0.002252 ok(boot) xtrain kingston start 1 -0.0124 -0.012354 ok xtrain width +0.0024 +0.002379 ok(boot) xtrain marrakesh minus pittsburgh excess +0.0074 +0.007420 ok xtrain width +0.0032 +0.003046 ok(boot) xtrain pittsburgh 2 Oct back-to-back ten gates, start 0 +0.023 +0.02288 ok paper: "+-0.023" xtrain pittsburgh 2 Oct back-to-back ten gates, start 1 -0.023 -0.02322 ok paper: "+-0.023" xtrain marrakesh 2 Oct back-to-back ten gates, start 1 -0.0148 -0.014812 ok paper "+-0.0148"; start 0 reads +0.0147 xtrain marrakesh start 0 excess vs ONE at 30 us +0.0302 +0.030189 ok xtrain width +0.0020 +0.002000 ok(boot) xtrain kingston start 0 excess vs ONE at 30 us +0.0325 +0.032455 ok xtrain width +0.0019 +0.001988 ok(boot) xtrain pittsburgh start 0 excess vs ONE at 30 us +0.0171 +0.017145 ok xtrain width +0.0028 +0.003148 ok(boot) xtrain marrakesh start 0 excess vs ONE at 5 us +0.0575 +0.057491 ok xtrain kingston start 0 excess vs ONE at 5 us +0.0576 +0.057600 ok xtrain pittsburgh start 0 excess vs ONE at 5 us +0.0267 +0.026680 ok xtrain marrakesh start 0 excess vs ONE at 150 us -0.0003 -0.000340 ok xtrain kingston start 0 excess vs ONE at 150 us +0.0058 +0.005757 ok xtrain pittsburgh start 0 excess vs ONE at 150 us +0.0050 +0.005013 ok xtrain start-1 / start-0 excess at 5,10,20 us vs ONE: min +0.35 +0.3680 DIFFERS paper: "35 to 100%" xtrain start-1 / start-0 excess at 5,10,20 us vs ONE: max +1.00 +0.9732 DIFFERS paper: "35 to 100%" xtrain 4 Oct ten back-to-back, start 0: min % +1.0 +0.999 ok paper: "+1.0 to +1.2%" xtrain 4 Oct ten back-to-back, start 0: max % +1.2 +1.151 ok paper: "+1.0 to +1.2%" xtrain 4 Oct ten back-to-back, start 1: |max| % +1.1 +1.070 ok paper: "-1.0 to -1.1%" xtrain 4 Oct ten back-to-back, start 1: |min| % +1.0 +0.991 ok paper: "-1.0 to -1.1%" xtrain pittsburgh 4 Oct back-to-back ten gates, start 0 +0.010 +0.00999 ok paper: "+-0.010" xtrain pittsburgh 4 Oct back-to-back ten gates, start 1 -0.010 -0.01070 DIFFERS paper: "+-0.010" xtrain kingston start 0 vs CURVE, 20-150 us: max |.| +0.007 +0.00627 ok paper: "within +-0.007"; range -0.0063..+0.0033 xtrain kingston start 0 vs CURVE, 30-150 us: max |.| +0.007 +0.00627 ok paper: "within +-0.007" at 30 us and longer; range -0.0063..-0.0034 xtrain kingston start 1 vs CURVE, 20-150 us: max |.| +0.007 +0.00334 ok paper: "within +-0.007"; range -0.0033..+0.0027 xtrain kingston start 1 vs CURVE, 30-150 us: max |.| +0.007 +0.00267 ok paper: "within +-0.007" at 30 us and longer; range -0.0020..+0.0027 xtrain pittsburgh start 0 vs CURVE, 20-150 us: max |.| +0.007 +0.00639 ok paper: "within +-0.007"; range -0.0064..+0.0034 xtrain pittsburgh start 0 vs CURVE, 30-150 us: max |.| +0.007 +0.00639 ok paper: "within +-0.007" at 30 us and longer; range -0.0064..+0.0034 xtrain pittsburgh start 1 vs CURVE, 20-150 us: max |.| +0.007 +0.00347 ok paper: "within +-0.007"; range -0.0035..+0.0028 xtrain pittsburgh start 1 vs CURVE, 30-150 us: max |.| +0.007 +0.00283 ok paper: "within +-0.007" at 30 us and longer; range -0.0019..+0.0028 xtrain marrakesh start 0 vs CURVE, 20-150 us: least negative -0.012 -0.01196 ok paper: "-0.012 to -0.018" xtrain marrakesh start 0 vs CURVE, 20-150 us: most negative -0.018 -0.01817 ok paper: "-0.012 to -0.018" xtrain marrakesh start 0 vs CURVE, 30-150 only: least negative -0.012 -0.01376 info info: the -0.012 end needs the 20 us (interpolated) point xtrain marrakesh start 1 vs CURVE, 20-150 us: most negative -0.016 -0.01564 ok paper: "by up to -0.016"; max positive -0.0018 abstract marrakesh over-correction 20-150 us, start 0 least/most (%) - - info 1.20 to 1.82 %; start 1 up to 1.56 % (paper: 1 to 2%) xtrain marrakesh start 0 vs CURVE at 5 us +0.0191 +0.019065 ok xtrain marrakesh start 0 vs CURVE at 10 us +0.0040 +0.003962 ok xtrain marrakesh start 1 vs CURVE at 5, 10 us (paper: opposite sign) - -0.019605 ok 5 us -0.0196, 10 us -0.0104 xtrain kingston start 0 vs CURVE at 5 us +0.0337 +0.033724 ok xtrain kingston start 0 vs CURVE at 10 us +0.0195 +0.019494 ok xtrain kingston start 1 vs CURVE at 5, 10 us (paper: opposite sign) - -0.032888 ok 5 us -0.0329, 10 us -0.0129 xtrain pittsburgh start 0 vs CURVE at 5 us +0.0108 +0.010757 ok xtrain pittsburgh start 0 vs CURVE at 10 us +0.0073 +0.007311 ok xtrain pittsburgh start 1 vs CURVE at 5, 10 us (paper: opposite sign) - -0.014557 ok 5 us -0.0146, 10 us -0.0068 gate ten back-to-back gates, start 0 (%) +2.29 +2.2882 ok gate ten back-to-back gates, start 1 (%) -2.32 -2.3223 ok gate X shortfall (deg) +1.47 +1.4727 ok gate width (deg) +0.12 +0.1191 ok(boot) gate 4b (% per gate) +0.261 +0.26080 ok gate width of 4b (%) +0.104 +0.10426 ok(boot) gate upper edge 4(b + 2 se) (%) +0.47 +0.4693 ok gate equator start, 2 gates, measured loss +0.0262 +0.026234 ok gate equator start, 10 gates, measured loss +0.114 +0.11375 ok gate shortfall-alone prediction, 2 gates +0.026 +0.02569 ok gate width from the shortfall width +0.002 +0.00208 ok(boot) gate shortfall-alone prediction, 10 gates +0.127 +0.12711 ok gate width from the shortfall width +0.012 +0.01006 DIFFERS 83 rows, 4 marked DIFFERS. NOTES 1. Rows marked DIFFERS are listed first-hand above; none was adjusted. The bookkeeping conventions are exactly those of readclose.py / readtime.py / readnoname.py. 2. "start-1 / start-0 ... 35 to 100%": recomputed range is 37 to 97% (pittsburgh 20 us and 5 us are the ends); the paper rounds outward. 3. Ten-gate prediction width: 0.012 in the paper; the fitted shortfall width (0.119 deg) gives 0.0101. 0.012 needs a shortfall width of about 0.142 deg; the notebook (section 149, 1.48 deg) printed 0.128 +- 0.012 and 0.0258 +- 0.0024, i.e. a width of about 0.14 deg, not the 0.12 deg now in the paper. 4. Pittsburgh 4 October: runs 362 and 363 (2 more jobs, same circuits, 15:33 and 15:34 EDT) exist in the record; the paper uses run 361 only ("one on pittsburgh"), as here. 5. Not numeric, not recomputed: "the population in the band where a |2> level at mid-line would read is flat across arms" and "the bookkeeping removes most of the excess". 6. Widths are bootstrap values and differ from the reference scripts by their own resampling noise (up to ~12% here); values (means) are not resampled and match exactly.