> Results tables of the second study, instrument v1.2 (revised 18 October 2026 after an independent review; on 24 October 2026 the instrument's rhymed-verse rule was applied at the quoted Schumann song, P072); numbers only. dose00–dose24 = first runs; r2, r3 = repeat runs of doses 0 and 20; C0 = the first study's control brief on the same model; H = Sophie Duvernoy's Chapter 25, scored by the same instrument; A, B, C, C-let, D-let, D-aim = the first study's six machine versions (Claude Opus 4.7, June 2026). Categories are the instrument's options; `other` collects renderings matching none of them.

# 1. New instrument (240 points)

Checks (agreement between procedures, not accuracy): her chapter takes its designated option at 240/240 in both batches; hidden audits, coder vs revised script 84/88 and 71/72 (vs the original script 82/88 and 71/72); coder vs builder on her chapter's fragments 60/62 and 60/62; texts coded in both batches: H 75/76, dose20 75/78; fragment samples 62 and 62, 50 in common.

## 1.1 Using an option with a documented earlier rendering (155 points with an example in ch. 1-24), by the point's latest example

| text | ch. 24 | ch. 22-23 | ch. 15-21 | ch. 5-14 | ch. 1-4 | all | excluding weak analogies |
|---|--:|--:|--:|--:|--:|--:|--:|
| dose00 | 18/29 | 23/36 | 32/54 | 15/32 | 2/4 | 90/155 | 84 |
| dose00r2 | 19/29 | 22/36 | 31/54 | 15/32 | 2/4 | 89/155 | 83 |
| dose00r3 | 20/29 | 22/36 | 33/54 | 15/32 | 2/4 | 92/155 | 86 |
| dose01 | 23/29 | 22/36 | 38/54 | 16/32 | 2/4 | 101/155 | 95 |
| dose03 | 23/29 | 25/36 | 38/54 | 16/32 | 2/4 | 104/155 | 97 |
| dose10 | 22/29 | 24/36 | 41/54 | 16/32 | 2/4 | 105/155 | 98 |
| dose20 | 22/29 | 27/36 | 43/54 | 24/32 | 2/4 | 118/155 | 112 |
| dose20r2 | 22/29 | 26/36 | 44/54 | 25/32 | 2/4 | 119/155 | 112 |
| dose20r3 | 22/29 | 24/36 | 43/54 | 26/32 | 2/4 | 117/155 | 110 |
| dose24 | 23/29 | 22/36 | 43/54 | 24/32 | 3/4 | 115/155 | 107 |
| C0 | 20/29 | 21/36 | 35/54 | 15/32 | 3/4 | 94/155 | 88 |
| A | 20/29 | 20/36 | 33/54 | 13/32 | 2/4 | 88/155 | 80 |
| B | 19/29 | 22/36 | 31/54 | 14/32 | 2/4 | 88/155 | 80 |
| C | 20/29 | 23/36 | 34/54 | 16/32 | 3/4 | 96/155 | 86 |
| C-let | 19/29 | 19/36 | 34/54 | 14/32 | 3/4 | 89/155 | 81 |
| D-let | 18/29 | 19/36 | 33/54 | 15/32 | 2/4 | 87/155 | 79 |
| D-aim | 17/29 | 18/36 | 33/54 | 17/32 | 3/4 | 88/155 | 78 |
| H | 22/29 | 27/36 | 42/54 | 22/32 | 2/4 | 115/155 | 104 |

## 1.1a Gains over dose 0 (credited where none of the three dose-0 runs is), and whether the option is attested inside the dose's reading

| run | gains | attested inside its reading | only outside | losses |
|---|--:|--:|--:|--:|
| dose01 | 9 | 3 | 6 | 0 |
| dose03 | 11 | 4 | 7 | 0 |
| dose10 | 12 | 11 | 1 | 1 |
| dose20 | 24 | 24 | 0 | 1 |
| dose20r2 | 24 | 24 | 0 | 1 |
| dose20r3 | 23 | 23 | 0 | 1 |
| dose24 | 23 | 23 | 0 | 2 |

## 1.2 Using her Chapter 25 option, by the instrument's classification of her Chapter 25 (unvalidated AI judgment)

| text | all | repeats all examples | some | none | no earlier example | other (any) |
|---|--:|--:|--:|--:|--:|--:|
| dose00 | 108/240 | 52/78 | 19/52 | 7/25 | 30/85 | 7 |
| dose00r2 | 98/240 | 49/78 | 18/52 | 6/25 | 25/85 | 7 |
| dose00r3 | 106/240 | 50/78 | 20/52 | 6/25 | 30/85 | 8 |
| dose01 | 115/240 | 58/78 | 23/52 | 6/25 | 28/85 | 10 |
| dose03 | 122/240 | 61/78 | 24/52 | 5/25 | 32/85 | 8 |
| dose10 | 115/240 | 63/78 | 19/52 | 6/25 | 27/85 | 12 |
| dose20 | 121/240 | 64/78 | 26/52 | 5/25 | 26/85 | 5 |
| dose20r2 | 118/240 | 65/78 | 18/52 | 4/25 | 31/85 | 5 |
| dose20r3 | 124/240 | 65/78 | 22/52 | 4/25 | 33/85 | 7 |
| dose24 | 114/240 | 64/78 | 21/52 | 4/25 | 25/85 | 6 |
| C0 | 103/240 | 53/78 | 18/52 | 7/25 | 25/85 | 10 |
| A | 99/240 | 50/78 | 15/52 | 8/25 | 26/85 | 12 |
| B | 100/240 | 48/78 | 19/52 | 8/25 | 25/85 | 11 |
| C | 104/240 | 53/78 | 16/52 | 7/25 | 28/85 | 9 |
| C-let | 94/240 | 50/78 | 17/52 | 5/25 | 22/85 | 12 |
| D-let | 90/240 | 48/78 | 14/52 | 3/25 | 25/85 | 11 |
| D-aim | 95/240 | 46/78 | 17/52 | 6/25 | 26/85 | 9 |
| H | 240/240 | 78/78 | 52/52 | 25/25 | 85/85 | 0 |

## 1.3 Structure (blind coder; compared with her chapter as the same coder read it, within each batch)

| text | fragments as she handles them | completes where she completes | keeps where she keeps | same, on the 50 fragments common to both samples | tense as hers | refrain |
|---|--:|--:|--:|--:|--:|--:|
| dose00 | 29/62 | 11/42 | 20/20 | 17/50 | 2/6 | 0.684 |
| dose00r2 | 29/62 | 11/42 | 20/20 | 17/50 | 2/6 | 0.689 |
| dose00r3 | 30/62 | 12/42 | 20/20 | 18/50 | 2/6 | 0.689 |
| dose01 | 32/62 | 14/42 | 20/20 | 20/50 | 2/6 | 0.7 |
| dose03 | 31/62 | 13/42 | 20/20 | 19/50 | 2/6 | 0.7 |
| dose10 | 31/62 | 12/42 | 20/20 | 19/50 | 2/6 | 0.688 |
| dose20 | 28/62 | 9/42 | 20/20 | 16/50 | 2/6 | 0.696 |
| dose20r2 | 30/62 | 11/42 | 20/20 | 18/50 | 2/6 | 0.692 |
| dose20r3 | 29/62 | 11/42 | 20/20 | 17/50 | 2/6 | 0.692 |
| dose24 | 32/62 | 13/42 | 20/20 | 20/50 | 2/6 | 0.686 |
| C0 | 29/62 | 11/42 | 20/20 | 17/50 | 2/6 | 0.696 |
| A | 27/62 | 9/42 | 19/20 | 15/50 | 2/6 | 0.696 |
| B | 29/62 | 11/42 | 19/20 | 18/50 | 1/6 | 0.689 |
| C | 29/62 | 10/42 | 20/20 | 17/50 | 1/6 | 0.689 |
| C-let | 32/62 | 14/42 | 20/20 | 20/50 | 2/6 | 0.679 |
| D-let | 32/62 | 13/42 | 20/20 | 20/50 | 3/6 | 0.688 |
| D-aim | 31/62 | 14/42 | 19/20 | 20/50 | 2/6 | 0.697 |
| H | 62/62 | 42/42 | 20/20 | 50/50 | 6/6 | 0.732 |

## 1.4 Same instrument category, point by point (240 points)

| set | pairs | mean | min | max | shared `other` per pair |
|---|--:|--:|--:|--:|---|
| dose 0, three runs (identical input) | 3 | 91.8% | 91.2% | 92.5% | [5, 5, 6] |
| dose 20, three runs (identical input) | 3 | 87.9% | 85.8% | 90.0% | [4, 4, 5] |
| June six (Opus 4.7, six briefs) | 15 | 79.4% | 71.2% | 87.5% | [10, 9, 10, 9, 7, 7, 9, 8, 5, 8, 7, 6, 11, 6, 6] |
| October first runs (doses 0-24) | 15 | 83.4% | 75.8% | 90.8% | [5, 5, 6, 2, 3, 7, 10, 4, 5, 7, 3, 2, 5, 6, 4] |
| C0 with each dose-0 run | 3 | — | 85.4% | 88.3% | [7, 6, 6] |
| C with C0 | 1 | — | 78.8% | 78.8% | [5] |

dose 0, three runs: 211 points unanimous (5 on `other`), 29 split (26 decided by the script in all three); her option where unanimous 94; her option at the splits: {'minority': 10, 'majority': 10, 'neither': 9}.

dose 20, three runs: 198 points unanimous (4 on `other`), 42 split (39 decided by the script in all three); her option where unanimous 107; her option at the splits: {'minority': 12, 'majority': 15, 'neither': 15}.

# 2. Style and surface (model-free; unchanged from v1.1)

| text | words | Delta to her Ch. 25 | Delta to her ch. 1-24 | Delta to the machines | em dashes /1k | contractions /1k | spelling forms (British/American) | dialogue marks | chrF to H | chrF, quotation marks unified |
|---|--:|--:|--:|--:|--:|--:|---|---|--:|--:|
| dose00 | 4369 | 0.907 | 1.556 | 0.456 | 0.0 | 17.84 | 7/0 | double | 70.58 | 70.6 |
| dose00r2 | 4345 | 0.894 | 1.571 | 0.441 | 0.0 | 16.76 | 7/0 | single | 67.71 | 70.26 |
| dose00r3 | 4310 | 0.908 | 1.549 | 0.432 | 4.84 | 18.43 | 0/4 | double | 70.15 | 70.17 |
| dose01 | 4339 | 0.849 | 1.528 | 0.471 | 6.41 | 20.15 | 0/5 | double | 71.74 | 71.76 |
| dose03 | 4329 | 0.824 | 1.528 | 0.481 | 4.37 | 22.08 | 0/6 | double | 72.18 | 72.19 |
| dose10 | 4283 | 0.85 | 1.558 | 0.446 | 9.02 | 19.66 | 0/4 | double | 71.42 | 71.44 |
| dose20 | 4291 | 0.791 | 1.53 | 0.449 | 7.86 | 22.67 | 0/7 | double | 72.28 | 72.3 |
| dose20r2 | 4313 | 0.847 | 1.586 | 0.481 | 8.28 | 20.24 | 0/8 | double | 71.59 | 71.62 |
| dose20r3 | 4310 | 0.8 | 1.546 | 0.456 | 8.74 | 21.63 | 0/6 | double | 71.85 | 71.87 |
| dose24 | 4317 | 0.843 | 1.561 | 0.503 | 6.67 | 23.47 | 0/7 | double | 72.12 | 72.14 |
| C0 | 4359 | 0.941 | 1.581 | 0.482 | 0.0 | 16.74 | 6/0 | double | 70.62 | 70.66 |
| A | 4235 | 0.978 | 1.622 | 0.532 | 12.01 | 16.72 | 0/3 | double | 68.21 | 68.24 |
| B | 4399 | 1.009 | 1.622 | 0.556 | 10.65 | 13.37 | 3/0 | double | 68.57 | 68.59 |
| C | 4352 | 1.001 | 1.595 | 0.5 | 12.81 | 17.62 | 1/3 | double | 69.64 | 69.66 |
| C-let | 4246 | 0.903 | 1.557 | 0.578 | 11.68 | 20.56 | 1/3 | double | 68.85 | 68.86 |
| D-let | 4274 | 0.977 | 1.603 | 0.596 | 8.81 | 14.37 | 2/0 | double | 68.3 | 68.32 |
| D-aim | 4331 | 1.065 | 1.619 | 0.77 | 18.41 | 9.66 | 5/0 | double | 67.33 | 67.34 |
| H | 4229 | — | 1.494 | 0.905 | 3.77 | 26.14 | 0/9 | double | — | — |

# 3. The first study's 45 points (blind coding, two batches; calibration anchors 45/45 in each; unchanged)

| text | same coded choice as her Chapter 25 (of 45) |
|---|--:|
| dose00 | 19 |
| dose00r2 | 16 |
| dose00r3 | 17 |
| dose01 | 19 |
| dose03 | 20 |
| dose10 | 21 |
| dose20 | 22 |
| dose20r2 | 20 |
| dose20r3 | 23 |
| dose24 | 20 |
| C0 | 21 |
| A | 13 |
| B | 12 |
| C | 17 |
| C-let | 18 |
| D-let | 11 |
| D-aim | 12 |