gpt-4.1-2025-04-14 v1 Evaluation

Generated: 2026-09-20 11:19:57 UTC

Pairs100
Distinct lemmas100
Slope0.999
R^20.977
BLEU-419.6%
chrF++51.2%
METEOR57.0%
ROUGE-L58.7%
BERTScore92.3%
COMET72.7%
BLEURT65.2%

Profile version id: 2416

First translation using this version: 2026-07-16

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

Created: 2026-07-16 06:54:03.500756+10:00

Active: True

Notes: Model timeline evaluation: GPT-4.1 (2025-04-14) 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
Slope0.999
Intercept4.633
R^20.977
P-value6.28e-82
Pearson r0.988
Mean source words30.400
Length fallback rows0
Mean human words44.150
Mean AI words48.730
Mean absolute length residual4.561
Mean absolute percent error15.6%

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: positive; r = 0.063, R^2 = 0.004, p = 0.5363.
chrF++ vs passage length
chrF++: positive; r = 0.118, R^2 = 0.014, p = 0.2436.
METEOR vs passage length
METEOR: negative; r = -0.124, R^2 = 0.015, p = 0.2207.
ROUGE-L vs passage length
ROUGE-L: negative; r = -0.133, R^2 = 0.018, p = 0.1856.
BERTScore vs passage length
BERTScore: significant negative; r = -0.265, R^2 = 0.070, p = 0.0077.
COMET vs passage length
COMET: negative; r = -0.043, R^2 = 0.002, p = 0.6744.
BLEURT vs passage length
BLEURT: significant negative; r = -0.638, R^2 = 0.407, p = 9.53e-13.
Trigram precision vs passage length
Trigram precision: positive; r = 0.057, R^2 = 0.003, p = 0.5764.
Trigram recall vs passage length
Trigram recall: positive; r = 0.023, R^2 = 5.52e-04, p = 0.8165.
Trigram F1 vs passage length
Trigram F1: positive; r = 0.042, R^2 = 0.002, p = 0.6795.
Trigram Jaccard vs passage length
Trigram Jaccard: positive; r = 0.009, R^2 = 7.75e-05, p = 0.9307.
Metric Rows Pattern Pearson r R^2 P-value Slope Status
BLEU-4 100 positive 0.063 0.004 0.5363 0.00025 ok
chrF++ 100 positive 0.118 0.014 0.2436 0.00037 ok
METEOR 100 negative -0.124 0.015 0.2207 -0.00059 ok
ROUGE-L 100 negative -0.133 0.018 0.1856 -0.00053 ok
BERTScore 100 significant negative -0.265 0.070 0.0077 -0.00023 ok
COMET 100 negative -0.043 0.002 0.6744 -0.00007 ok
BLEURT 100 significant negative -0.638 0.407 9.53e-13 -0.00182 ok
Trigram precision 100 positive 0.057 0.003 0.5764 0.00025 ok
Trigram recall 100 positive 0.023 5.52e-04 0.8165 0.00011 ok
Trigram F1 100 positive 0.042 0.002 0.6795 0.00019 ok
Trigram Jaccard 100 positive 0.009 7.75e-05 0.9307 0.00003 ok

Translation Similarity Metrics

Mean BLEU-419.6%
Mean chrF++51.2%
Mean METEOR57.0%
Mean ROUGE-L58.7%
Mean BERTScore92.3%
Mean COMET72.7%
Mean BLEURT65.2%
Legacy smoothed corpus BLEU24.4%
Unigram F162.5%
Bigram F132.6%
Trigram precision17.7%
Trigram recall19.7%
Trigram F118.7%
Trigram Jaccard10.3%
4-gram F111.3%

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 270 254 -20.310 18.0% 48.4% 40.6% 48.5% 87.9% 75.3% 43.6% reviewed/approved 13281 gpt-4.1-2025-04-14
7264 Κωλιάς 61 82 16.440 9.1% 41.5% 44.5% 39.2% 90.4% 64.9% 54.2% reviewed/approved 13422 gpt-4.1-2025-04-14
2468 Καπετώλιον 138 157 14.532 9.9% 40.0% 34.7% 41.4% 85.5% 68.7% 47.4% reviewed/approved 13272 gpt-4.1-2025-04-14
3288 Κορώνεια 110 129 14.499 17.4% 50.5% 50.5% 54.2% 90.5% 72.3% 55.4% reviewed/approved 13344 gpt-4.1-2025-04-14
2329 Κάλπη 57 76 14.435 20.4% 49.6% 51.6% 52.2% 89.7% 69.2% 62.1% reviewed/approved 13236 gpt-4.1-2025-04-14
7266 Κώμη 79 96 12.462 10.0% 43.1% 38.9% 36.6% 88.6% 68.0% 48.4% reviewed/approved 13428 gpt-4.1-2025-04-14
2115 Καλὴ ἀκτή 64 81 12.444 13.9% 40.7% 51.5% 49.7% 87.5% 61.9% 45.8% reviewed/approved 13209 gpt-4.1-2025-04-14
2055 Καβασσός 109 102 -11.502 18.4% 51.3% 45.0% 46.4% 92.4% 74.6% 47.6% reviewed/approved 13137 gpt-4.1-2025-04-14
2477 Καρδαμύλη 43 59 11.419 20.1% 58.8% 73.0% 56.9% 91.7% 74.8% 71.9% reviewed/approved 13278 gpt-4.1-2025-04-14
2609 Κάσπειρος 135 129 -10.471 20.1% 55.0% 49.1% 56.8% 88.3% 71.5% 58.2% reviewed/approved 13305 gpt-4.1-2025-04-14
2455 Κάναστρον 60 55 -9.561 14.9% 47.7% 52.6% 52.2% 91.5% 67.8% 58.7% reviewed/approved 13266 gpt-4.1-2025-04-14
2604 Καρχηδών 117 131 9.507 24.9% 57.4% 49.5% 56.8% 91.3% 76.2% 53.5% reviewed/approved 13296 gpt-4.1-2025-04-14
2465 Κάνωπος 83 96 8.466 32.5% 65.6% 80.3% 74.9% 93.5% 74.5% 63.8% reviewed/approved 13269 gpt-4.1-2025-04-14
2346 Κάναι 77 90 8.459 31.2% 57.3% 73.2% 73.1% 92.1% 74.7% 60.0% reviewed/approved 13263 gpt-4.1-2025-04-14
2605 Κάσιον 56 69 8.434 12.0% 52.0% 66.7% 57.6% 92.5% 72.1% 61.3% reviewed/approved 13299 gpt-4.1-2025-04-14
2342 Κάμιρος 54 67 8.432 29.2% 63.4% 76.9% 69.4% 93.2% 76.1% 68.0% reviewed/approved 13257 gpt-4.1-2025-04-14
7262 Κῶβρυς 22 35 8.394 6.7% 41.4% 54.8% 49.1% 90.3% 72.0% 66.3% reviewed/approved 13416 gpt-4.1-2025-04-14
3496 Κοτιάειον 71 68 -7.548 17.3% 46.2% 56.8% 54.7% 92.6% 71.6% 65.3% reviewed/approved 13347 gpt-4.1-2025-04-14
2597 Καρπασία 112 109 -7.499 19.9% 57.5% 56.5% 67.0% 93.6% 72.0% 59.2% reviewed/approved 13284 gpt-4.1-2025-04-14
2328 Καλλίπολις 43 55 7.419 5.2% 46.0% 43.4% 53.1% 90.3% 71.3% 59.0% reviewed/approved 13233 gpt-4.1-2025-04-14
7250 Κυρταία 40 52 7.415 16.6% 41.7% 41.7% 45.7% 89.1% 69.4% 64.5% reviewed/approved 13380 gpt-4.1-2025-04-14
2085 Καλάσιρις 13 25 7.383 6.0% 43.0% 33.8% 36.8% 90.9% 66.5% 60.1% reviewed/approved 13197 gpt-4.1-2025-04-14
2630 Κάτρη 23 21 -6.605 9.4% 42.6% 47.6% 59.1% 94.3% 78.5% 72.4% reviewed/approved 13338 gpt-4.1-2025-04-14
2087 Καλαύρεια 30 28 -6.597 21.1% 54.5% 59.1% 65.5% 94.1% 76.4% 71.0% reviewed/approved 13203 gpt-4.1-2025-04-14
2079 Καισάρεια 39 37 -6.586 38.5% 58.5% 67.8% 68.4% 92.5% 73.5% 64.9% reviewed/approved 13179 gpt-4.1-2025-04-14
2623 Κασώριον 20 31 6.391 9.7% 48.2% 49.7% 43.1% 90.2% 63.4% 60.3% reviewed/approved 13317 gpt-4.1-2025-04-14
2624 Κατάβαθμος 20 19 -5.609 35.9% 66.6% 64.2% 71.8% 94.4% 76.5% 69.0% reviewed/approved 13320 gpt-4.1-2025-04-14
2330 Καλύβη 24 23 -5.604 19.9% 46.3% 58.6% 68.1% 92.8% 71.8% 70.3% reviewed/approved 13239 gpt-4.1-2025-04-14
3530 Κριώα 26 25 -5.602 15.4% 37.6% 28.5% 47.1% 89.6% 62.8% 64.1% reviewed/approved 13353 gpt-4.1-2025-04-14
7247 Κύρνος 50 49 -5.573 20.2% 49.6% 62.7% 58.6% 92.7% 71.8% 68.7% reviewed/approved 13371 gpt-4.1-2025-04-14

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: 8.377; cross-validated R^2: -0.384.

Positive termCoefficientDocsMean presentMean absent
και 2.0976 66 0.631 -1.226
κωμη 1.1951 3 12.108 -0.374
εστι δε 1.0213 5 9.074 -0.478
δια του 1.0128 10 3.769 -0.419
εκ 0.9234 6 5.603 -0.358
εστι δε και 0.8479 3 11.808 -0.365
γαρ 0.8236 9 2.182 -0.216
θηλυκως 0.7037 5 5.042 -0.265
και θηλυκως 0.7037 5 5.042 -0.265
εν τη 0.6924 7 3.996 -0.301
παρα 0.6622 7 1.917 -0.144
δυο 0.6441 4 7.451 -0.310
εις 0.6362 7 3.661 -0.276
του δε 0.6362 3 5.760 -0.178
λεγεται 0.6292 9 2.592 -0.256
αιγυπτου 0.6210 3 5.754 -0.178
του 0.5965 37 1.609 -0.945
αι 0.5768 3 8.454 -0.261
την 0.5521 13 1.882 -0.281
χωρας 0.5500 4 3.648 -0.152
Negative termCoefficientDocsMean presentMean absent
και του -0.7332 4 -6.075 0.253
περι -0.7218 11 -1.724 0.213
καππαδοκιας -0.6788 3 -5.898 0.182
το -0.6643 70 0.314 -0.732
το εθνικον -0.6525 56 -0.290 0.369
εθνικον -0.6525 56 -0.290 0.369
προς τη -0.6042 4 -3.854 0.161
εκαλειτο -0.5924 7 -4.925 0.371
ελλανικος -0.5757 4 -9.972 0.415
εθνος -0.5726 8 -3.428 0.298
τα -0.5462 12 -1.007 0.137
αλλ -0.5325 3 -5.528 0.171
το θηλυκον -0.5003 6 -3.680 0.235
μετα -0.4980 3 -10.767 0.333
μεταξυ -0.4962 5 -1.566 0.082
τους -0.4653 4 -2.579 0.107
τινες δε -0.4560 3 -4.254 0.132
εκαταιος -0.4417 14 -1.078 0.176
μια -0.4374 3 -3.254 0.101
νησος -0.4258 9 -1.352 0.134

Greek source terms predicting large absolute residuals

Features: 176; ridge alpha: 2.031; cross-validated R^2: 1.44e-04.

Positive termCoefficientDocsMean presentMean absent
γαρ 4.3238 9 11.578 3.867
δε 2.7421 46 5.879 3.437
την 2.6947 13 7.727 4.087
και 2.5328 66 5.236 3.250
εστι 2.4730 24 6.800 3.853
κωμη 1.9785 3 12.108 4.327
τα 1.8046 12 9.080 3.944
εστι δε 1.5567 5 10.073 4.270
τε 1.5290 5 9.070 4.323
περι 1.4720 11 7.048 4.253
ως 1.4556 52 5.219 3.848
ακρα 1.3737 4 9.250 4.365
εν 1.3474 37 5.687 3.899
δια 1.3426 16 6.932 4.109
εστι δε και 1.3386 3 11.808 4.336
ηρωδιανος 1.3299 3 13.816 4.274
τα τοπικα 1.2457 3 11.181 4.356
τοπικα 1.2457 3 11.181 4.356
εστι και 1.1993 20 6.367 4.109
ομηρος 1.1966 3 10.451 4.378
Negative termCoefficientDocsMean presentMean absent
στραβων -1.8616 15 3.235 4.794
παυσανιας -1.4972 4 1.781 4.676
τους -1.3184 4 2.787 4.634
ης -1.3042 3 1.510 4.655
ουδετερως -1.1965 3 1.568 4.653
λιβυης -1.0910 4 4.734 4.553
πολις προς -1.0667 3 1.865 4.644
παιδος -1.0595 4 3.171 4.618
πολυβιος -1.0388 5 2.074 4.691
ην -1.0112 13 3.967 4.649
εκαταιος -0.9930 14 3.890 4.670
αυτην -0.9738 9 3.445 4.671
φησι -0.8933 12 4.828 4.524
το εθνικον -0.8713 56 4.534 4.594
εθνικον -0.8713 56 4.534 4.594
αλλη -0.8466 4 4.266 4.573
ιβηριας -0.8301 3 3.895 4.581
νησος -0.8232 9 3.735 4.642
οτι -0.7889 5 3.698 4.606
φησι το -0.7775 3 0.853 4.675

AI English terms predicting large absolute residuals

Features: 438; ridge alpha: 1.425; cross-validated R^2: -0.039.

Positive termCoefficientDocsMean presentMean absent
for 4.6389 15 8.724 3.826
is 2.5005 90 4.759 2.770
usage 2.3877 4 13.412 4.192
iota 2.1654 6 11.318 4.129
among 1.9105 6 10.613 4.174
promontory 1.8634 6 9.534 4.243
there is 1.8007 26 6.549 3.862
diphthong 1.7875 3 14.114 4.265
and 1.7848 54 5.758 3.155
or 1.7700 18 7.667 3.879
people 1.6849 8 8.372 4.229
from 1.6740 41 6.174 3.439
a promontory 1.6582 3 13.500 4.284
on 1.6397 16 7.183 4.061
with 1.6353 18 7.489 3.918
there is also 1.6227 22 6.145 4.114
thus 1.6040 4 13.922 4.171
also 1.5771 43 5.234 4.052
place 1.5642 10 7.401 4.245
derived 1.5474 5 12.036 4.167
Negative termCoefficientDocsMean presentMean absent
name -1.5554 20 4.451 4.588
strabo -1.2439 15 3.235 4.794
pausanias -1.2039 4 1.781 4.676
after -1.1335 13 4.255 4.606
a city -1.1158 66 4.453 4.770
strabo in the -1.1066 5 1.496 4.722
strabo in -1.0688 7 2.480 4.717
also a city -1.0118 7 5.196 4.513
of the -0.9750 44 4.597 4.532
ethnic designation is -0.9641 30 4.285 4.679
designation is -0.9641 30 4.285 4.679
formed -0.9054 6 2.505 4.692
hecataeus mentions -0.9053 5 1.728 4.710
they -0.9019 8 3.499 4.653
arcadia -0.8860 3 1.591 4.652
named after -0.8717 10 3.204 4.711
city -0.8717 69 4.484 4.730
hecataeus -0.8658 14 3.890 4.670
the ethnicon is -0.8611 9 2.689 4.746
the ethnicon -0.8611 9 2.689 4.746

Human English terms predicting large absolute residuals

Features: 335; ridge alpha: 2.031; cross-validated R^2: -0.043.

Positive termCoefficientDocsMean presentMean absent
with 2.7853 16 7.921 3.921
and 2.2735 52 5.847 3.167
is 2.2701 87 4.827 2.781
there is 1.9450 24 6.839 3.841
is also 1.8186 31 6.269 3.793
there is also 1.7752 22 6.779 3.935
on 1.6892 16 7.125 4.072
there 1.4636 34 6.384 3.621
is also a 1.4404 14 8.496 3.920
as 1.4326 60 5.213 3.581
or 1.4247 18 8.338 3.731
also a 1.4025 15 7.957 3.961
among 1.4018 4 11.651 4.265
the 1.3809 90 4.682 3.468
is from 1.2867 12 6.919 4.239
place 1.2803 6 7.839 4.351
long 1.2581 3 15.768 4.214
form 1.2579 11 8.767 4.041
among the 1.2294 4 11.651 4.265
with the 1.1794 5 11.652 4.187
Negative termCoefficientDocsMean presentMean absent
strabo -1.2381 15 3.235 4.794
was from -1.2103 4 1.031 4.708
was -0.9813 20 4.608 4.549
it is also -0.9052 3 0.751 4.678
also called -0.8538 6 3.265 4.643
the ethnonym -0.8455 56 4.531 4.598
pausanias -0.8095 4 1.781 4.676
of the -0.7739 33 5.325 4.184
strabo in -0.7620 3 0.848 4.675
ethnonym -0.7410 57 4.654 4.437
pausanias book -0.7265 3 1.507 4.655
is the -0.7193 3 1.194 4.665
which -0.6977 20 4.345 4.615
mount -0.6948 4 3.280 4.614
strabo book the -0.6557 3 0.518 4.686
they -0.6535 12 4.601 4.555
book the ethnonym -0.6392 7 2.807 4.693
the ethnonym is -0.6387 52 4.690 4.420
ethnonym is -0.6387 52 4.690 4.420
an island -0.6382 9 3.413 4.674

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.