gpt-5.6-sol v1 Evaluation

Generated: 2026-09-20 11:22:41 UTC

Pairs100
Distinct lemmas100
Slope0.998
R^20.974
BLEU-423.3%
chrF++51.8%
METEOR56.1%
ROUGE-L57.7%
BERTScore92.2%
COMET71.6%
BLEURT63.1%

Profile version id: 2175

First translation using this version: 2026-07-14

Total usable AI runs for this version: 100 runs across 100 lemmas

Created: 2026-07-10 11:03:49.528703+10:00

Active: True

Notes: Model timeline evaluation: GPT-5.6 Sol using prompt text from gpt-5.5 v1; approved-human-only for the 100-row Kappa corpus.

Prompt length scatter plot

Length Regression

FormulaAI words = intercept + slope * human words
Slope0.998
Intercept3.145
R^20.974
P-value1.44e-79
Pearson r0.987
Mean source words30.400
Length fallback rows0
Mean human words44.150
Mean AI words47.190
Mean absolute length residual4.288
Mean absolute percent error13.1%

Metric vs Passage Length

These plots test whether score changes with source passage length for this prompt version. The dashed line is an ordinary least-squares fit; the annotation reports n, r, R^2, p-value, and slope.

BLEU-4 vs passage length
BLEU-4: negative; r = -0.084, R^2 = 0.007, p = 0.4037.
chrF++ vs passage length
chrF++: positive; r = 0.017, R^2 = 2.75e-04, p = 0.8700.
METEOR vs passage length
METEOR: negative; r = -0.107, R^2 = 0.012, p = 0.2878.
ROUGE-L vs passage length
ROUGE-L: negative; r = -0.080, R^2 = 0.006, p = 0.4285.
BERTScore vs passage length
BERTScore: significant negative; r = -0.212, R^2 = 0.045, p = 0.0339.
COMET vs passage length
COMET: negative; r = -0.053, R^2 = 0.003, p = 0.5980.
BLEURT vs passage length
BLEURT: significant negative; r = -0.422, R^2 = 0.178, p = 1.21e-05.
Trigram precision vs passage length
Trigram precision: negative; r = -0.064, R^2 = 0.004, p = 0.5271.
Trigram recall vs passage length
Trigram recall: negative; r = -0.076, R^2 = 0.006, p = 0.4544.
Trigram F1 vs passage length
Trigram F1: negative; r = -0.069, R^2 = 0.005, p = 0.4968.
Trigram Jaccard vs passage length
Trigram Jaccard: negative; r = -0.096, R^2 = 0.009, p = 0.3418.
Metric Rows Pattern Pearson r R^2 P-value Slope Status
BLEU-4 100 negative -0.084 0.007 0.4037 -0.00039 ok
chrF++ 100 positive 0.017 2.75e-04 0.8700 0.00007 ok
METEOR 100 negative -0.107 0.012 0.2878 -0.00060 ok
ROUGE-L 100 negative -0.080 0.006 0.4285 -0.00039 ok
BERTScore 100 significant negative -0.212 0.045 0.0339 -0.00020 ok
COMET 100 negative -0.053 0.003 0.5980 -0.00014 ok
BLEURT 100 significant negative -0.422 0.178 1.21e-05 -0.00158 ok
Trigram precision 100 negative -0.064 0.004 0.5271 -0.00031 ok
Trigram recall 100 negative -0.076 0.006 0.4544 -0.00039 ok
Trigram F1 100 negative -0.069 0.005 0.4968 -0.00034 ok
Trigram Jaccard 100 negative -0.096 0.009 0.3418 -0.00032 ok

Translation Similarity Metrics

Mean BLEU-423.3%
Mean chrF++51.8%
Mean METEOR56.1%
Mean ROUGE-L57.7%
Mean BERTScore92.2%
Mean COMET71.6%
Mean BLEURT63.1%
Legacy smoothed corpus BLEU25.0%
Unigram F161.4%
Bigram F132.9%
Trigram precision18.7%
Trigram recall20.1%
Trigram F119.4%
Trigram Jaccard10.7%
4-gram F111.5%

Largest Length Residuals

Lemma ID Headword Human words AI words Residual BLEU-4 chrF++ METEOR ROUGE-L BERTScore COMET BLEURT Human Run Model
2484 Καρία 270 244 -28.503 19.1% 50.7% 43.8% 51.4% 89.6% 75.9% 48.8% reviewed/approved 10605 gpt-5.6-sol
3258 Κεκρυφάλεια 26 1 -28.083 0.0% 0.2% 0.0% 0.0% 78.4% 28.0% 6.4% reviewed/approved 10665 gpt-5.6-sol
2605 Κάσιον 56 74 14.988 14.8% 48.6% 46.0% 46.2% 90.9% 68.5% 58.3% reviewed/approved 10623 gpt-5.6-sol
2604 Καρχηδών 117 133 13.133 20.3% 53.8% 57.3% 55.6% 91.4% 75.9% 56.0% reviewed/approved 10620 gpt-5.6-sol
3288 Κορώνεια 110 126 13.117 15.1% 51.2% 43.0% 53.2% 90.9% 71.2% 56.6% reviewed/approved 10668 gpt-5.6-sol
2328 Καλλίπολις 43 58 11.957 14.9% 51.6% 56.0% 55.4% 92.4% 73.2% 67.0% reviewed/approved 10557 gpt-5.6-sol
2621 Κάστνιον 25 40 11.914 21.6% 52.5% 60.2% 49.2% 92.3% 65.9% 64.9% reviewed/approved 10635 gpt-5.6-sol
2116 Κάληρος 40 32 -11.050 16.4% 51.3% 56.4% 55.6% 93.7% 76.3% 73.4% reviewed/approved 10536 gpt-5.6-sol
2115 Καλὴ ἀκτή 64 78 11.007 17.1% 46.5% 59.5% 57.7% 89.5% 64.7% 48.6% reviewed/approved 10533 gpt-5.6-sol
2455 Κάναστρον 60 52 -11.002 13.7% 47.5% 56.1% 60.7% 91.7% 65.2% 56.9% reviewed/approved 10590 gpt-5.6-sol
2468 Καπετώλιον 138 150 9.183 16.5% 44.9% 45.5% 47.2% 86.6% 65.2% 45.4% reviewed/approved 10596 gpt-5.6-sol
7266 Κώμη 79 90 8.043 7.2% 38.7% 33.9% 37.9% 89.5% 65.7% 42.7% reviewed/approved 10752 gpt-5.6-sol
2119 Κάλλατις 65 76 8.010 16.1% 47.2% 47.2% 44.0% 90.4% 62.8% 56.7% reviewed/approved 10545 gpt-5.6-sol
2056 Καβειρία 112 107 -7.879 12.5% 46.0% 37.7% 42.0% 90.8% 65.1% 44.7% reviewed/approved 10466 gpt-5.6-sol
2603 Κάρυστος 175 185 7.271 31.6% 61.4% 67.5% 68.3% 93.2% 78.1% 57.8% reviewed/approved 10617 gpt-5.6-sol
7247 Κύρνος 50 46 -7.026 14.5% 42.7% 53.6% 52.1% 92.2% 70.1% 62.3% reviewed/approved 10695 gpt-5.6-sol
2470 Καππαδοκία 68 78 7.017 12.8% 48.5% 54.1% 61.6% 92.2% 66.8% 51.0% reviewed/approved 10599 gpt-5.6-sol
2342 Κάμιρος 54 64 6.983 27.4% 61.3% 70.9% 67.8% 93.9% 74.4% 67.4% reviewed/approved 10581 gpt-5.6-sol
2055 Καβασσός 109 118 6.114 16.7% 46.4% 39.2% 45.8% 88.2% 71.1% 43.7% reviewed/approved 10463 gpt-5.6-sol
7246 Κύρις 20 17 -6.097 25.9% 47.9% 43.7% 48.6% 90.0% 59.9% 43.2% reviewed/approved 10692 gpt-5.6-sol
7251 Κύρτος 66 75 6.012 36.9% 58.1% 68.4% 65.2% 91.3% 67.7% 53.6% reviewed/approved 10707 gpt-5.6-sol
7264 Κωλιάς 61 70 6.000 26.7% 54.5% 55.6% 59.5% 91.7% 67.3% 56.9% reviewed/approved 10746 gpt-5.6-sol
7259 Κύφος 54 63 5.983 20.9% 52.0% 61.2% 58.1% 91.8% 66.9% 51.5% reviewed/approved 10731 gpt-5.6-sol
2624 Κατάβαθμος 20 29 5.903 43.5% 73.2% 82.1% 69.4% 94.9% 81.4% 79.8% reviewed/approved 10644 gpt-5.6-sol
2078 Καινύς 22 20 -5.093 40.2% 55.6% 57.9% 57.1% 93.5% 79.3% 66.1% reviewed/approved 10497 gpt-5.6-sol
7262 Κῶβρυς 22 20 -5.093 8.2% 43.1% 29.2% 42.9% 91.6% 67.4% 61.1% reviewed/approved 10740 gpt-5.6-sol
7257 Κυτώνιον 25 23 -5.086 31.4% 55.6% 60.5% 66.7% 92.3% 72.0% 61.0% reviewed/approved 10725 gpt-5.6-sol
7261 Κύψελα 34 32 -5.064 31.4% 57.8% 59.7% 63.6% 93.9% 73.6% 67.8% reviewed/approved 10737 gpt-5.6-sol
2346 Κάναι 77 85 5.038 33.4% 60.0% 74.2% 74.1% 92.2% 75.2% 58.0% reviewed/approved 10587 gpt-5.6-sol
3496 Κοτιάειον 71 69 -4.976 14.8% 45.3% 57.1% 55.7% 90.7% 59.9% 59.1% reviewed/approved 10671 gpt-5.6-sol

Residual Predictors

Positive signed-residual terms are associated with AI translations longer than expected from the human length; negative signed-residual terms are associated with shorter AI translations. Absolute-residual models identify terms associated with larger AI-human length divergence.

Greek source terms predicting signed length residuals

Features: 176; ridge alpha: 4.125; cross-validated R^2: -0.085.

Positive termCoefficientDocsMean presentMean absent
ου 2.0582 14 3.830 -0.623
της 1.7776 35 1.294 -0.697
του 1.6427 37 2.176 -1.278
ως 1.6300 52 1.244 -1.348
εν 1.6017 37 2.434 -1.430
εστι δε 1.5524 5 6.867 -0.361
εν τη 1.5437 7 5.970 -0.449
ου και 1.4907 4 8.807 -0.367
ορος 1.4683 6 6.714 -0.429
κωμη 1.4208 3 8.002 -0.247
εστι δε και 1.4056 3 10.402 -0.322
και 1.4020 66 0.701 -1.361
πολις 1.3702 68 0.280 -0.595
πολιτης 1.3594 18 3.877 -0.851
αφ ου 1.2805 5 4.902 -0.258
εξ 1.1966 3 7.384 -0.228
πολιχνιον 1.1640 3 8.335 -0.258
δευτερα 1.1580 3 8.330 -0.258
εκ 1.1553 6 5.326 -0.340
τον 1.0533 10 3.419 -0.380
Negative termCoefficientDocsMean presentMean absent
τινες -2.4303 6 -7.709 0.492
τινες δε -2.1771 3 -11.721 0.362
περι -2.1008 11 -3.329 0.411
το -2.0320 70 -0.210 0.489
φασι -1.8796 8 -3.725 0.324
πελοποννησου -1.7748 3 -6.030 0.186
τις -1.7317 3 -5.685 0.176
το εθνικον -1.6493 56 -1.532 1.950
εθνικον -1.6493 56 -1.532 1.950
θρακης -1.4859 5 -5.467 0.288
εθνος -1.3938 8 -5.860 0.510
εστι και -1.3022 20 -0.561 0.140
ακρα -1.2930 4 -3.731 0.155
το θηλυκον -1.1928 6 -5.740 0.366
ουτως -1.1819 5 -7.849 0.413
θεοπομπος -1.1485 5 -2.832 0.149
αυτην -1.0605 9 -1.661 0.164
πολις θρακης -0.8857 4 -4.084 0.170
εκαλειτο -0.8673 7 -2.731 0.206
οτι -0.7939 5 0.032 -0.002

Greek source terms predicting large absolute residuals

Features: 176; ridge alpha: 5.878; cross-validated R^2: -0.292.

Positive termCoefficientDocsMean presentMean absent
δε 1.8766 46 5.921 2.898
περι 1.5366 11 9.722 3.617
πελοποννησου 1.3986 3 14.775 3.964
τις 1.3298 3 14.429 3.975
ακρα 1.2013 4 10.311 4.037
γαρ 1.1600 9 8.889 3.833
τινες 1.1163 6 9.031 3.986
εκαλειτο 1.0984 7 10.308 3.835
τινες δε 1.0646 3 11.721 4.059
την 0.9910 13 7.562 3.799
και 0.9258 66 4.724 3.442
απο της 0.9158 6 11.637 3.819
φασι 0.8622 8 8.822 3.894
εστι δε 0.7531 5 10.019 3.987
ει 0.7529 4 13.889 3.888
ου 0.7348 14 6.103 3.993
παρα 0.7140 7 10.060 3.854
δια 0.6785 16 6.463 3.874
αφ ου 0.6736 5 8.054 4.090
εστι 0.6690 24 5.722 3.836
Negative termCoefficientDocsMean presentMean absent
δευτερω -0.9355 3 0.380 4.409
πολις -0.8264 68 4.066 4.762
εκαταιος -0.7141 14 3.128 4.477
τον -0.7020 10 3.683 4.356
στραβων -0.6347 15 3.492 4.429
οι -0.4964 11 4.284 4.289
νησος -0.4734 9 2.796 4.436
εκαταιος ασια -0.4665 5 1.490 4.436
ασια -0.4665 5 1.490 4.436
λυκιας -0.4610 3 3.102 4.325
πολιτης -0.4536 18 4.474 4.248
και το -0.4446 9 4.364 4.281
ως στραβων -0.4295 4 1.779 4.393
ποσειδωνος -0.4142 3 1.674 4.369
ηροδοτος -0.3986 9 5.646 4.154
παυσανιας -0.3824 4 3.458 4.323
φησι το -0.3809 3 1.043 4.389
προς τη -0.3710 4 2.535 4.361
τριτω -0.3670 3 1.356 4.379
λιβυης -0.3652 4 4.046 4.299

AI English terms predicting large absolute residuals

Features: 401; ridge alpha: 5.878; cross-validated R^2: -0.306.

Positive termCoefficientDocsMean presentMean absent
with 1.0702 19 6.728 3.716
and 1.0270 56 5.203 3.124
for 0.9283 12 7.411 3.863
usage 0.8957 5 10.045 3.985
caria 0.8744 3 14.523 3.972
on 0.8659 26 6.164 3.629
was called 0.7816 6 11.162 3.850
was 0.7715 24 6.674 3.535
diphthong 0.7245 3 13.269 4.011
he 0.6211 7 8.378 3.981
it 0.5910 40 5.636 3.390
is a 0.5815 6 11.106 3.853
by 0.5504 16 6.555 3.857
either 0.5476 3 17.558 3.878
was named 0.5290 5 10.899 3.941
it was named 0.5268 4 11.805 3.975
zeus 0.5229 5 12.997 3.830
it was 0.5220 13 7.358 3.830
from 0.5010 36 5.698 3.496
a mountain 0.4971 3 10.647 4.092
Negative termCoefficientDocsMean presentMean absent
a city -0.7131 66 3.794 5.248
book -0.6039 39 3.866 4.559
an -0.5941 24 3.310 4.597
of the -0.5376 40 4.470 4.168
herodotus book -0.4974 5 1.649 4.427
city -0.4973 68 4.066 4.762
is called -0.4908 24 3.883 4.416
are called -0.4814 9 3.210 4.395
hecataeus -0.4797 14 3.128 4.477
inhabitants -0.4120 6 2.185 4.423
of his -0.4100 10 3.025 4.429
are -0.3944 26 4.673 4.153
settlement -0.3874 3 1.382 4.378
is called a -0.3874 21 3.825 4.412
called a -0.3874 21 3.825 4.412
strabo -0.3858 15 3.492 4.429
book of the -0.3816 8 2.396 4.453
book of -0.3727 16 4.364 4.274
near -0.3701 14 3.456 4.424
it is also -0.3692 4 1.229 4.416

Human English terms predicting large absolute residuals

Features: 335; ridge alpha: 2.894; cross-validated R^2: -0.210.

Positive termCoefficientDocsMean presentMean absent
however 2.5036 10 11.400 3.498
the 2.5023 90 4.533 2.085
in the peloponnese 1.6885 3 14.775 3.964
peloponnese 1.6885 3 14.775 3.964
the peloponnese 1.6885 3 14.775 3.964
which the 1.5405 3 12.374 4.038
on 1.5088 16 6.553 3.857
this 1.4626 15 8.759 3.500
and 1.3728 52 5.295 3.198
promontory 1.3571 3 12.068 4.048
is with 1.3343 4 13.230 3.916
mountain 1.2479 3 11.391 4.069
used to be 1.2316 6 11.513 3.827
his on 1.2287 3 15.491 3.942
an island in 1.2201 3 10.985 4.081
island in 1.2201 3 10.985 4.081
island in the 1.2201 3 10.985 4.081
say that it 1.2172 3 11.721 4.059
karia 1.2171 3 14.523 3.972
with 1.1857 16 6.628 3.843
Negative termCoefficientDocsMean presentMean absent
a city -0.9785 64 3.851 5.067
in book -0.9388 16 3.867 4.369
in book of -0.9350 14 2.165 4.634
strabo -0.8520 15 3.492 4.429
book -0.8220 39 3.866 4.559
but -0.7975 12 3.073 4.454
city -0.7930 70 4.107 4.711
hekataios -0.7739 14 3.128 4.477
is a -0.7663 31 4.711 4.099
herodotos book -0.7342 5 1.649 4.427
are -0.7240 25 5.389 3.922
near -0.6821 16 3.670 4.406
they -0.6795 12 4.806 4.218
those -0.6505 5 2.845 4.364
lykia -0.6442 3 3.102 4.325
in lykia -0.6442 3 3.102 4.325
written -0.6192 3 3.295 4.319
they are -0.6131 7 3.470 4.350
hekataios in -0.6078 12 3.134 4.446
one of -0.6020 5 2.220 4.397

Prompt Text

Show prompt text
You are an expert classical philologist and translator specializing in Byzantine Greek geographical texts.
You will receive Greek text from a lemma entry in Stephanos of Byzantium's Ethnika.
Translate the Greek text into clear, scholarly English.
Preserve technical terminology and place names appropriately.