gpt-5.3-chat-latest v1 Evaluation

Generated: 2026-07-25 12:53:30 UTC

Pairs100
Distinct lemmas100
Slope0.987
R^20.962
BLEU-423.0%
chrF++52.8%
METEOR59.5%
ROUGE-L61.1%
BERTScoreN/A
COMETN/A
BLEURTN/A

Profile version id: 2334

First translation using this version: 2026-07-15

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

Created: 2026-07-15 12:28:27.643960+10:00

Active: True

Notes: Model timeline evaluation: GPT-5.3 Chat 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
Slope0.987
Intercept3.014
R^20.962
P-value3.74e-71
Pearson r0.981
Mean source words30.400
Length fallback rows0
Mean human words44.160
Mean AI words46.620
Mean absolute length residual4.012
Mean absolute percent error11.4%

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 = 0.020, R^2 = 4.10e-04, p = 0.8415.
chrF++ vs passage length
chrF++: positive; r = 0.038, R^2 = 0.001, p = 0.7038.
METEOR vs passage length
METEOR: negative; r = -0.024, R^2 = 5.58e-04, p = 0.8155.
ROUGE-L vs passage length
ROUGE-L: negative; r = -0.064, R^2 = 0.004, p = 0.5292.
Trigram precision vs passage length
Trigram precision: positive; r = 0.021, R^2 = 4.36e-04, p = 0.8366.
Trigram recall vs passage length
Trigram recall: positive; r = 0.019, R^2 = 3.80e-04, p = 0.8474.
Trigram F1 vs passage length
Trigram F1: positive; r = 0.021, R^2 = 4.34e-04, p = 0.8369.
Trigram Jaccard vs passage length
Trigram Jaccard: positive; r = 3.08e-04, R^2 = 9.47e-08, p = 0.9976.
Metric Rows Pattern Pearson r R^2 P-value Slope Status
BLEU-4 100 positive 0.020 4.10e-04 0.8415 0.00010 ok
chrF++ 100 positive 0.038 0.001 0.7038 0.00013 ok
METEOR 100 negative -0.024 5.58e-04 0.8155 -0.00012 ok
ROUGE-L 100 negative -0.064 0.004 0.5292 -0.00027 ok
BERTScore 0 not enough data N/A N/A N/A N/A needs at least two rows with metric scores
COMET 0 not enough data N/A N/A N/A N/A needs at least two rows with metric scores
BLEURT 0 not enough data N/A N/A N/A N/A needs at least two rows with metric scores
Trigram precision 100 positive 0.021 4.36e-04 0.8366 0.00010 ok
Trigram recall 100 positive 0.019 3.80e-04 0.8474 0.00009 ok
Trigram F1 100 positive 0.021 4.34e-04 0.8369 0.00010 ok
Trigram Jaccard 100 positive 3.08e-04 9.47e-08 0.9976 9.77e-07 ok

Translation Similarity Metrics

Mean BLEU-423.0%
Mean chrF++52.8%
Mean METEOR59.5%
Mean ROUGE-L61.1%
Mean BERTScoreN/A
Mean COMETN/A
Mean BLEURTN/A
Legacy smoothed corpus BLEU29.1%
Unigram F165.3%
Bigram F136.7%
Trigram precision22.1%
Trigram recall23.4%
Trigram F122.7%
Trigram Jaccard12.8%
4-gram F114.8%

Largest Length Residuals

Lemma ID Headword Human words AI words Residual BLEU-4 chrF++ METEOR ROUGE-L BERTScore COMET BLEURT Human Run Model
2465 Κάνωπος 83 147 62.027 23.3% 60.5% 78.8% 59.1% N/A N/A N/A reviewed/approved 11167 gpt-5.3-chat-latest
2484 Καρία 270 251 -18.626 18.7% 50.1% 52.7% 57.6% N/A N/A N/A reviewed/approved 11182 gpt-5.3-chat-latest
7266 Κώμη 79 97 15.977 3.5% 34.2% 32.1% 33.0% N/A N/A N/A reviewed/approved 11320 gpt-5.3-chat-latest
2056 Καβειρία 112 101 -12.609 16.6% 52.6% 54.2% 61.0% N/A N/A N/A reviewed/approved 11037 gpt-5.3-chat-latest
7267 Κωνώπη 64 78 11.789 36.4% 51.6% 46.9% 54.9% N/A N/A N/A reviewed/approved 11324 gpt-5.3-chat-latest
2116 Κάληρος 40 33 -9.512 21.6% 53.1% 56.2% 54.8% N/A N/A N/A reviewed/approved 11111 gpt-5.3-chat-latest
2603 Κάρυστος 175 185 9.182 27.8% 57.1% 66.4% 62.8% N/A N/A N/A reviewed/approved 11193 gpt-5.3-chat-latest
2328 Καλλίπολις 43 54 8.525 8.4% 48.4% 44.7% 47.4% N/A N/A N/A reviewed/approved 11131 gpt-5.3-chat-latest
2609 Κάσπειρος 135 128 -8.320 35.4% 62.3% 61.7% 66.2% N/A N/A N/A reviewed/approved 11204 gpt-5.3-chat-latest
7254 Κύτα 84 94 8.040 27.9% 56.7% 56.0% 58.4% N/A N/A N/A reviewed/approved 11286 gpt-5.3-chat-latest
2057 Καβελλιών 34 44 7.412 16.1% 47.5% 47.0% 43.6% N/A N/A N/A reviewed/approved 11038 gpt-5.3-chat-latest
2626 Κατάνη 95 90 -6.822 33.0% 56.3% 63.4% 60.5% N/A N/A N/A reviewed/approved 11222 gpt-5.3-chat-latest
2342 Κάμιρος 54 63 6.664 23.4% 60.8% 66.9% 66.7% N/A N/A N/A reviewed/approved 11157 gpt-5.3-chat-latest
2055 Καβασσός 109 104 -6.646 21.6% 51.5% 52.0% 49.8% N/A N/A N/A reviewed/approved 11034 gpt-5.3-chat-latest
7255 Κυτέριον 22 31 6.262 14.6% 46.3% 48.7% 52.8% N/A N/A N/A reviewed/approved 11289 gpt-5.3-chat-latest
7253 Κυρτώνιος 18 27 6.212 4.8% 35.7% 36.4% 35.6% N/A N/A N/A reviewed/approved 11282 gpt-5.3-chat-latest
7245 Κύρης 13 22 6.149 8.7% 34.9% 50.5% 45.7% N/A N/A N/A reviewed/approved 11259 gpt-5.3-chat-latest
2624 Κατάβαθμος 20 17 -5.763 29.6% 58.3% 54.9% 64.9% N/A N/A N/A reviewed/approved 11213 gpt-5.3-chat-latest
7261 Κύψελα 34 31 -5.588 43.8% 65.0% 73.0% 76.9% N/A N/A N/A reviewed/approved 11306 gpt-5.3-chat-latest
2081 Καλάθη 49 46 -5.399 25.4% 55.6% 70.0% 73.7% N/A N/A N/A reviewed/approved 11082 gpt-5.3-chat-latest
2075 Καινίνη 15 13 -4.826 46.0% 63.3% 73.6% 78.6% N/A N/A N/A reviewed/approved 11064 gpt-5.3-chat-latest
2080 Καλαβρία 18 16 -4.788 21.9% 52.2% 39.5% 47.1% N/A N/A N/A reviewed/approved 11079 gpt-5.3-chat-latest
2335 Κάλυτις 26 24 -4.688 5.9% 46.6% 39.2% 56.0% N/A N/A N/A reviewed/approved 11153 gpt-5.3-chat-latest
2079 Καισάρεια 39 37 -4.525 30.7% 56.4% 67.7% 68.4% N/A N/A N/A reviewed/approved 11075 gpt-5.3-chat-latest
3288 Κορώνεια 110 116 4.366 16.2% 51.3% 49.6% 52.9% N/A N/A N/A reviewed/approved 11239 gpt-5.3-chat-latest
2623 Κασώριον 20 27 4.237 28.1% 55.8% 59.7% 55.3% N/A N/A N/A reviewed/approved 11212 gpt-5.3-chat-latest
2599 Καρπήσιοι 11 10 -3.876 22.3% 55.5% 54.0% 57.1% N/A N/A N/A reviewed/approved 11187 gpt-5.3-chat-latest
2061 Καδούσιοι 14 13 -3.839 47.8% 61.8% 77.2% 74.1% N/A N/A N/A reviewed/approved 11050 gpt-5.3-chat-latest
2115 Καλὴ ἀκτή 64 70 3.789 15.1% 41.5% 55.4% 52.2% N/A N/A N/A reviewed/approved 11108 gpt-5.3-chat-latest
2620 Κάσταξ 20 19 -3.763 8.2% 43.5% 50.8% 56.4% N/A N/A N/A reviewed/approved 11205 gpt-5.3-chat-latest

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: 412.463; cross-validated R^2: -0.223.

Positive termCoefficientDocsMean presentMean absent
και 0.0722 66 0.101 -0.196
δια του 0.0433 10 6.159 -0.684
και το 0.0427 9 7.645 -0.756
απο 0.0330 27 2.457 -0.909
και νησος 0.0287 3 21.327 -0.660
δια 0.0270 16 3.924 -0.748
και το κτητικον 0.0263 4 15.905 -0.663
λιβυης 0.0242 4 15.108 -0.630
πολις εν 0.0228 5 11.409 -0.600
εδει 0.0227 3 17.669 -0.546
τοπος 0.0225 4 16.299 -0.679
ποσειδωνος 0.0219 3 18.714 -0.579
κτητικον 0.0179 7 4.803 -0.362
το κτητικον 0.0179 7 4.803 -0.362
νησος 0.0175 9 6.263 -0.619
κωμη 0.0172 3 5.764 -0.178
εστι 0.0167 24 1.469 -0.464
του 0.0165 37 1.583 -0.930
απο του 0.0156 9 4.135 -0.409
εν 0.0136 37 2.112 -1.240
Negative termCoefficientDocsMean presentMean absent
δε -0.0195 46 0.454 -0.386
εθνος -0.0195 8 -4.920 0.428
εκαταιος -0.0188 14 -2.228 0.363
εκαλειτο -0.0183 7 -4.012 0.302
πλησιον -0.0182 8 -3.017 0.262
και του -0.0161 4 -5.412 0.225
μεταξυ -0.0157 5 -3.116 0.164
εκαταιος ασια -0.0149 5 -2.905 0.153
ασια -0.0149 5 -2.905 0.153
οτι -0.0137 5 -1.878 0.099
το θηλυκον -0.0133 6 -5.426 0.346
χωρα -0.0121 6 -4.694 0.300
ποταμου -0.0120 3 -3.171 0.098
μετα -0.0116 3 -11.256 0.348
ης -0.0114 3 -5.274 0.163
θηλυκον και -0.0112 3 -6.624 0.205
παυσανιας -0.0094 4 -4.225 0.176
λυκιας -0.0093 3 -3.383 0.105
καππαδοκιας -0.0091 3 -3.628 0.112
τινες -0.0091 6 -2.141 0.137

Greek source terms predicting large absolute residuals

Features: 176; ridge alpha: 24.245; cross-validated R^2: -0.239.

Positive termCoefficientDocsMean presentMean absent
και 0.9793 66 4.663 2.749
δια του 0.6055 10 10.916 3.245
δια 0.5703 16 8.069 3.239
ουτως 0.5455 5 21.061 3.115
εστι 0.5179 24 7.056 3.051
και το 0.4992 9 9.295 3.489
απο 0.4562 27 6.847 2.963
εδει 0.4080 3 23.683 3.403
τοπος 0.4068 4 19.305 3.375
πολις εν 0.3949 5 16.067 3.377
ποσειδωνος 0.3910 3 22.638 3.436
και νησος 0.3651 3 21.327 3.476
θηλυκον 0.3644 12 11.123 3.042
και το κτητικον 0.3554 4 17.149 3.464
κτητικον 0.3342 7 14.438 3.227
το κτητικον 0.3342 7 14.438 3.227
λιβυης 0.2822 4 16.133 3.507
εστι και 0.2779 20 7.501 3.140
του 0.2418 37 5.514 3.129
εν 0.2127 37 5.630 3.062
Negative termCoefficientDocsMean presentMean absent
ως -0.2746 52 3.594 4.465
ηροδοτος -0.2053 9 4.569 3.957
πολιτης -0.1833 18 3.039 4.225
δευτερω -0.1751 3 1.670 4.084
πολις και -0.1745 6 1.199 4.191
ην -0.1720 13 2.366 4.258
το εθνικον -0.1580 56 4.435 3.473
εθνικον -0.1580 56 4.435 3.473
και πολις -0.1467 5 1.930 4.121
δεκατω -0.1444 3 0.718 4.114
εκ -0.1437 6 1.627 4.164
πολις -0.1418 68 4.627 2.704
φησιν -0.1170 10 4.010 4.012
του δε -0.1133 3 0.936 4.107
της -0.1112 35 4.160 3.932
μια -0.1026 3 1.676 4.084
τοπικα -0.0989 3 2.528 4.058
τα τοπικα -0.0989 3 2.528 4.058
ος -0.0923 6 2.529 4.106
ακρα -0.0911 4 1.344 4.123

AI English terms predicting large absolute residuals

Features: 403; ridge alpha: 8.377; cross-validated R^2: -0.215.

Positive termCoefficientDocsMean presentMean absent
be 1.3611 8 13.719 3.168
written 1.2538 4 17.098 3.467
written with 1.2538 4 17.098 3.467
with a 1.2434 5 16.461 3.357
and the 1.0598 17 9.133 2.963
and 0.9757 56 4.961 2.804
the possessive 0.9320 3 31.087 3.174
there 0.8816 31 6.298 2.985
is a city 0.8783 4 20.391 3.329
with 0.8718 16 7.977 3.257
adjective 0.8699 3 29.564 3.222
is a 0.8615 5 17.041 3.326
possessive 0.8459 4 23.495 3.200
usage 0.8261 5 18.043 3.273
named from 0.8186 4 21.211 3.295
have 0.8165 6 16.504 3.214
is 0.8028 88 4.147 3.022
a place in 0.8022 3 25.658 3.342
feminine 0.7577 16 10.448 2.786
to be 0.7566 7 14.785 3.201
Negative termCoefficientDocsMean presentMean absent
the ethnic -0.4814 55 3.105 5.120
name is -0.4676 15 1.754 4.410
ethnic name is -0.4631 13 1.704 4.357
ethnic -0.4300 58 3.314 4.975
in the -0.4221 36 3.202 4.467
the ethnic name -0.4123 15 2.058 4.357
ethnic name -0.4123 15 2.058 4.357
at -0.3797 7 0.942 4.243
are -0.3580 21 3.585 4.125
the ethnic is -0.3212 37 3.284 4.439
an iota -0.3128 4 1.627 4.111
which -0.3038 16 3.856 4.042
is called -0.2863 17 3.480 4.121
also called -0.2794 12 4.291 3.974
forms are -0.2750 3 1.066 4.103
ethnic is -0.2653 38 3.530 4.307
settlement -0.2640 3 0.617 4.117
at the -0.2504 3 0.372 4.124
strabo book -0.2449 6 2.092 4.134
book of his -0.2439 10 2.619 4.167

Human English terms predicting large absolute residuals

Features: 336; ridge alpha: 8.377; cross-validated R^2: -0.205.

Positive termCoefficientDocsMean presentMean absent
written 1.8953 3 26.735 3.309
for 1.3333 11 12.883 2.915
there 1.2597 34 6.548 2.705
and the 1.0759 15 8.314 3.253
with 1.0632 16 8.087 3.236
a place 1.0043 3 25.658 3.342
the 0.9873 90 4.169 2.598
is written 0.9686 3 26.735 3.309
is a city 0.9055 3 23.125 3.421
feminine is 0.8999 12 12.138 2.904
poseidon 0.8903 3 22.638 3.436
the feminine is 0.8888 11 12.511 2.961
it should 0.8347 3 23.683 3.403
should 0.8347 3 23.683 3.403
is also 0.8217 31 6.624 2.838
there is also 0.7984 22 7.151 3.127
sanctuary 0.7889 3 21.811 3.461
and 0.7855 52 4.972 2.972
feminine 0.7817 16 10.448 2.786
there is 0.7781 24 6.742 3.150
Negative termCoefficientDocsMean presentMean absent
at -0.4357 6 0.830 4.215
of the -0.4152 33 3.583 4.223
was -0.3964 20 3.421 4.159
also called -0.3817 6 1.448 4.175
city in -0.3696 41 3.742 4.200
a city in -0.3571 40 3.768 4.174
citizen -0.3250 18 3.039 4.225
a citizen -0.3250 18 3.039 4.225
the ethnonym is -0.3062 52 4.045 3.976
ethnonym is -0.3062 52 4.045 3.976
called -0.2866 21 3.846 4.056
near -0.2857 16 3.321 4.143
herodotos -0.2821 9 4.569 3.957
in book -0.2630 16 3.926 4.028
his -0.2487 34 3.720 4.162
of his -0.2443 18 3.807 4.057
citizen is a -0.2358 15 3.216 4.152
citizen is -0.2358 15 3.216 4.152
a citizen is -0.2358 15 3.216 4.152
herodotos book -0.2351 5 3.293 4.050

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.