AMKemma-3.1-Write / evaluation.json
amkkk's picture
Add AMKemma model card, verified style metrics and license
dcfb69d verified
Raw History Blame Contribute Delete
12 kB
{
"benchmark": "H2 TEST2000",
"n": 2000,
"weights": "BF16 parent + checkpoint42831 adapter; pre-quantization",
"features": {
"lexdiv": {
"human": 0.44899158253682175,
"base": 0.4158458531026242,
"expert": 0.44999183728938086,
"historical_base": 0.4127069373936656,
"historical_cone": 0.38800292698276473,
"paired_delta": 0.03414598418675671,
"paired95ci": [
0.028984627249466995,
0.038637008931309856
]
},
"lexdensity": {
"human": 0.44336953235725207,
"base": 0.45670213143953275,
"expert": 0.433369614579625,
"historical_base": 0.45355159115291144,
"historical_cone": 0.44569783553637227,
"paired_delta": -0.02333251685990781,
"paired95ci": [
-0.02622278779922698,
-0.020518426100244007
]
},
"pct_long": {
"human": 0.19021838478553185,
"base": 0.24707889950463954,
"expert": 0.1821357298909025,
"historical_base": 0.24460927426372608,
"historical_cone": 0.23024393721795905,
"paired_delta": -0.06494316961373704,
"paired95ci": [
-0.0678355478258805,
-0.061946992166030196
]
},
"entropy": {
"human": 6.918216785489374,
"base": 6.898383787069547,
"expert": 6.317382813656346,
"historical_base": 6.897676138511753,
"historical_cone": 6.757419865804984,
"paired_delta": -0.5810009734131996,
"paired95ci": [
-0.6056459439573577,
-0.5595779027138188
]
},
"burstiness": {
"human": 1.7166219880663958,
"base": 2.057251991675612,
"expert": 1.376956660025221,
"historical_base": 2.1194940380409957,
"historical_cone": 2.0435451885093436,
"paired_delta": -0.6802953316503909,
"paired95ci": [
-0.70024346636305,
-0.6588752304719451
]
},
"pct_vowels": {
"human": 0.29663441035929405,
"base": 0.3041152182182486,
"expert": 0.3033233973177885,
"historical_base": 0.30399907499760026,
"historical_cone": 0.3027035217327655,
"paired_delta": -0.0007918209004600809,
"paired95ci": [
-0.0014682615429397147,
-2.214605971942154e-05
]
},
"pct_consonants": {
"human": 0.47879499480625215,
"base": 0.4903276550737023,
"expert": 0.4850320953228257,
"historical_base": 0.4891986978956337,
"historical_cone": 0.48793290877553636,
"paired_delta": -0.005295559750876576,
"paired95ci": [
-0.006208119166421207,
-0.004418319643465491
]
},
"pct_punct": {
"human": 0.036021427481950155,
"base": 0.045666385444399014,
"expert": 0.030516165332549847,
"historical_base": 0.04683780704859862,
"historical_cone": 0.043815765011107115,
"paired_delta": -0.015150220111849163,
"paired95ci": [
-0.015941927909233808,
-0.014338510947371007
]
},
"n_words": {
"human": 733.3925,
"base": 637.5175,
"expert": 332.9965,
"historical_base": 653.679,
"historical_cone": 632.8035,
"paired_delta": -304.521,
"paired95ci": [
-312.51651250000003,
-295.99275
]
},
"n_sents": {
"human": 34.128,
"base": 30.4155,
"expert": 17.444,
"historical_base": 31.3525,
"historical_cone": 31.757,
"paired_delta": -12.9715,
"paired95ci": [
-13.55155,
-12.39845
]
},
"mean_sent_len": {
"human": 24.86668532029877,
"base": 23.64437700612384,
"expert": 28.077007539128974,
"historical_base": 23.31581804509707,
"historical_cone": 22.91044907099578,
"paired_delta": 4.432630533005134,
"paired95ci": [
2.8423998561445893,
6.2831279586983095
]
},
"parse_depth": {
"human": 10.9595,
"base": 10.3015,
"expert": 9.7005,
"historical_base": 10.2285,
"historical_cone": 10.3255,
"paired_delta": -0.601,
"paired95ci": [
-0.8065125,
-0.36242500000000005
]
},
"fog": {
"human": 12.98932660843822,
"base": 15.375829512024142,
"expert": 12.215177582054885,
"historical_base": 15.219314251852479,
"historical_cone": 14.486501705686242,
"paired_delta": -3.160651929969258,
"paired95ci": [
-3.4392152667917077,
-2.854815409701112
]
},
"linsear": {
"human": 11.590718414224664,
"base": 14.526123185002133,
"expert": 11.012325868303094,
"historical_base": 14.404184514316093,
"historical_cone": 13.631578638028637,
"paired_delta": -3.513797316699039,
"paired95ci": [
-3.7686522329083996,
-3.22759373280046
]
},
"fixed_token_count": {
"human": 858.3635,
"base": 686.666,
"expert": 370.244,
"historical_base": 701.3615,
"historical_cone": 682.722,
"paired_delta": -316.422,
"paired95ci": [
-325.86671249999995,
-306.8394875
]
},
"whitespace_words": {
"human": 628.6215,
"base": 497.659,
"expert": 284.167,
"historical_base": 506.5845,
"historical_cone": 502.7205,
"paired_delta": -213.492,
"paired95ci": [
-219.9237125,
-206.4722625
]
},
"characters": {
"human": 3690.662,
"base": 3308.795,
"expert": 1650.1825,
"historical_base": 3372.837,
"historical_cone": 3245.9915,
"paired_delta": -1658.6125,
"paired95ci": [
-1700.625425,
-1611.810025
]
},
"mean_word_len": {
"human": 4.7124710255587585,
"base": 5.29901914381053,
"expert": 4.605062292174038,
"historical_base": 5.295809816341491,
"historical_cone": 5.1088054261547,
"paired_delta": -0.6939568516364925,
"paired95ci": [
-0.7154329793514712,
-0.6720580533154238
]
},
"longword_rate_ge10": {
"human": 0.05786167468063766,
"base": 0.09969084815111688,
"expert": 0.05266852065506798,
"historical_base": 0.09940085953093548,
"historical_cone": 0.08867284858456216,
"paired_delta": -0.047022327496048885,
"paired95ci": [
-0.04905133038812932,
-0.044914202959057216
]
},
"ttr": {
"human": 0.4646555560960963,
"base": 0.476706651231559,
"expert": 0.46731385829588046,
"historical_base": 0.47515714736115605,
"historical_cone": 0.4408826569302387,
"paired_delta": -0.009392792935678546,
"paired95ci": [
-0.01456295441215851,
-0.004579432244846972
]
},
"page_burst": {
"human": 1003.1127753576015,
"base": 124.19555369839392,
"expert": 336.17720182456014,
"historical_base": 129.7658371986992,
"historical_cone": 177.07001540720202,
"paired_delta": 211.98164812616616,
"paired95ci": [
74.20152785262952,
396.0984267448633
]
},
"repeat_start_run3": {
"human": 0.2375,
"base": 0.2285,
"expert": 0.253,
"historical_base": 0.3285,
"historical_cone": 0.2765,
"paired_delta": 0.0245,
"paired95ci": [
0.0009875000000000007,
0.049012499999999994
]
},
"empty": {
"human": 0.0,
"base": 0.0,
"expert": 0.0,
"historical_base": 0.0,
"historical_cone": 0.0,
"paired_delta": 0.0,
"paired95ci": [
0.0,
0.0
]
},
"cap_hit": {
"human": 0.0,
"base": 0.169,
"expert": 0.0125,
"historical_base": 0.0,
"historical_cone": 0.0,
"paired_delta": -0.1565,
"paired95ci": [
-0.1725,
-0.1405
]
},
"rep_loop": {
"human": 0.0,
"base": 0.0,
"expert": 0.001,
"historical_base": 0.0,
"historical_cone": 0.0,
"paired_delta": 0.001,
"paired95ci": [
0.0,
0.0025
]
},
"unexpected_character_rate": {
"human": 7.950872827889747e-07,
"base": 1.5070067758599604e-06,
"expert": 6.345177664974619e-07,
"historical_base": 1.2987026658161848e-06,
"historical_cone": 0.0,
"paired_delta": -8.724890093624985e-07,
"paired95ci": [
-2.676294633538666e-06,
8.939417595515629e-07
]
},
"non_english_character_rate": {
"human": 0.0003932134307056947,
"base": 0.00013167691455761094,
"expert": 0.00025468870873991364,
"historical_base": 0.00014189700091632934,
"historical_cone": 0.00013879863283128906,
"paired_delta": 0.00012301179418230273,
"paired95ci": [
3.5832596639529664e-05,
0.00020752526917641972
]
},
"training_word_rarity_nats": {
"human": 7.8035010384940415,
"base": 8.064658653593,
"expert": 7.504689380269631,
"historical_base": 8.059660641155142,
"historical_cone": 7.877223583893059,
"paired_delta": -0.5599692733233704,
"paired95ci": [
-0.5868189555013194,
-0.5312468054922478
]
},
"training_word_oov_rate": {
"human": 0.010148877612250551,
"base": 0.0049545350394098525,
"expert": 0.007205710940527181,
"historical_base": 0.004951255660104235,
"historical_cone": 0.004713799595342232,
"paired_delta": 0.0022511759011173285,
"paired95ci": [
0.001137831458371458,
0.0036573017120947646
]
},
"option_wrapper": {
"human": 0.0,
"base": 0.2,
"expert": 0.0505,
"historical_base": 0.2,
"historical_cone": 0.2005,
"paired_delta": -0.1495,
"paired95ci": [
-0.1655,
-0.133
]
},
"assistant_preface": {
"human": 0.0,
"base": 0.0,
"expert": 0.053,
"historical_base": 0.0,
"historical_cone": 0.0,
"paired_delta": 0.053,
"paired95ci": [
0.0434875,
0.0635
]
}
},
"distances": {
"L2-1": {
"base": 0.02906163662772162,
"expert": 0.022144033488984215,
"historical_base": 0.028940514092103572,
"historical_cone": 0.029127328372322736,
"human_floor": 0.008073846371447806,
"recovery": 0.32960130886907196
},
"L2-2": {
"base": 0.01241105042702878,
"expert": 0.014798941629477258,
"historical_base": 0.012706657466151293,
"historical_cone": 0.013251067221917458,
"human_floor": 0.005002273829992486,
"recovery": -0.32230573714473953
},
"L2-3": {
"base": 0.006615217004254781,
"expert": 0.004305352452957473,
"historical_base": 0.006978098397409446,
"historical_cone": 0.00723905748621644,
"human_floor": 0.003945395560136238,
"recovery": 0.865175667978022
},
"JSD-nats": {
"base": 0.1054327105606854,
"expert": 0.05729762336622043,
"historical_base": 0.10695706211294842,
"historical_cone": 0.09503636149711513,
"human_floor": 0.08331861271188122,
"recovery": 2.176669720988319
}
},
"feature_definition": "existing metrics.Footprint Shan14/style; additional diagnostics explicitly named, not quality judgments",
"uncertainty": "corpus distances point estimates; paired feature95% bootstrap CIs (1000 draws, seed20261021)",
"historical_comparison": "historical_base and historical_cone use a separate archived runtime stratum"
}