devision / evaluation /accept_vs_v0.2.json
lukbit's picture
deVision v0.3 (github model-v0.3, 43cf900e42ba7a8557be1506679d79de8c45786a)
af353dc verified
Raw History Blame Contribute Delete
13.8 kB
{
"sets": {
"test_count_fresh": {
"accuracy": {
"metric": "accuracy",
"questions": 600,
"pictures": 600,
"a": 0.74,
"b": 0.7633333333333333,
"difference": 0.023333333333333317,
"ci95": [
-0.0050000000000000044,
0.05166666666666664
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 600,
"pictures": 600,
"a": 0.5763476151482702,
"b": 0.5738475317419256,
"difference": -0.0025000834063445554,
"ci95": [
-0.029590039246708333,
0.02324226594506207
],
"resamples": 2000
},
"held": true,
"proven_worse": false,
"nll_warning": false
},
"bench_countbench": {
"accuracy": {
"metric": "accuracy",
"questions": 491,
"pictures": 491,
"a": 0.1894093686354379,
"b": 0.3564154786150713,
"difference": 0.1670061099796334,
"ci95": [
0.12830957230142565,
0.20977596741344195
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 491,
"pictures": 491,
"a": 2.249204785673313,
"b": 1.6021778431421219,
"difference": -0.647026942531191,
"ci95": [
-0.7236271559411285,
-0.5697184849200367
],
"resamples": 2000
},
"nll_warning": false
},
"test_vsr": {
"accuracy": {
"metric": "accuracy",
"questions": 904,
"pictures": 552,
"a": 0.6792035398230089,
"b": 0.661504424778761,
"difference": -0.017699115044247815,
"ci95": [
-0.03871681415929196,
0.003293084522502765
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 904,
"pictures": 552,
"a": 0.6155330847335275,
"b": 0.6197663897233163,
"difference": 0.0042333049897888,
"ci95": [
-0.008103858831641997,
0.016373944210066993
],
"resamples": 2000
},
"held": false,
"proven_worse": false,
"nll_warning": false
},
"test_v7w": {
"accuracy": {
"metric": "accuracy",
"questions": 1000,
"pictures": 1000,
"a": 0.721,
"b": 0.721,
"difference": 0.0,
"ci95": [
-0.017000000000000015,
0.018000000000000016
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 1000,
"pictures": 1000,
"a": 0.6838468317414087,
"b": 0.6725771682747884,
"difference": -0.011269663466620283,
"ci95": [
-0.03041221689888618,
0.007412588911638762
],
"resamples": 2000
},
"held": true,
"proven_worse": false,
"nll_warning": false
},
"bench_pope": {
"accuracy": {
"metric": "accuracy",
"questions": 8676,
"pictures": 482,
"a": 0.851083448593822,
"b": 0.8495850622406639,
"difference": -0.001498386353158132,
"ci95": [
-0.008529276164130906,
0.005301982480405698
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 8676,
"pictures": 482,
"a": 0.3702472832864825,
"b": 0.36794475743031013,
"difference": -0.002302525856172355,
"ci95": [
-0.01004565387165457,
0.005636645091598225
],
"resamples": 2000
},
"held": true,
"proven_worse": false,
"nll_warning": false
},
"test_vqa_yesno": {
"accuracy": {
"metric": "accuracy",
"questions": 1000,
"pictures": 984,
"a": 0.71,
"b": 0.704,
"difference": -0.006000000000000005,
"ci95": [
-0.021956087824351322,
0.01106639839034207
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 1000,
"pictures": 984,
"a": 0.5768279689372598,
"b": 0.5741047062190436,
"difference": -0.002723262718216213,
"ci95": [
-0.012621285357050205,
0.0073198649324632825
],
"resamples": 2000
},
"held": false,
"proven_worse": false,
"nll_warning": false
},
"test_vqa_choice": {
"accuracy": {
"metric": "accuracy",
"questions": 1420,
"pictures": 1346,
"a": 0.897887323943662,
"b": 0.9091549295774648,
"difference": 0.011267605633802802,
"ci95": [
0.0020862308762169546,
0.022487702037948032
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 1420,
"pictures": 1346,
"a": 0.2706275980383607,
"b": 0.2705435796140065,
"difference": -8.401842435418194e-05,
"ci95": [
-0.011447355800969727,
0.011158424151348323
],
"resamples": 2000
},
"held": true,
"proven_worse": false,
"nll_warning": false
},
"test_gqa": {
"accuracy": {
"metric": "accuracy",
"questions": 992,
"pictures": 882,
"a": 0.7842741935483871,
"b": 0.7903225806451613,
"difference": 0.006048387096774133,
"ci95": [
-0.011033099297893645,
0.02300000000000002
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 992,
"pictures": 882,
"a": 0.4089840430444238,
"b": 0.4104783294623267,
"difference": 0.0014942864179028903,
"ci95": [
-0.013564724731577937,
0.016149505199867686
],
"resamples": 2000
},
"held": true,
"proven_worse": false,
"nll_warning": false
},
"test_exist": {
"accuracy": {
"metric": "accuracy",
"questions": 1120,
"pictures": 1099,
"a": 0.9383928571428571,
"b": 0.9473214285714285,
"difference": 0.008928571428571397,
"ci95": [
-0.000889679715302516,
0.018766756032171594
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 1120,
"pictures": 1099,
"a": 0.16199041568107028,
"b": 0.16308653751641133,
"difference": 0.0010961218353410518,
"ci95": [
-0.012942116093576261,
0.01502907012673671
],
"resamples": 2000
},
"held": true,
"proven_worse": false,
"nll_warning": false
},
"test_position": {
"accuracy": {
"metric": "accuracy",
"questions": 1274,
"pictures": 1231,
"a": 0.8720565149136578,
"b": 0.8846153846153846,
"difference": 0.012558869701726816,
"ci95": [
-0.0007886435331231345,
0.026025236593059997
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 1274,
"pictures": 1231,
"a": 0.3045267949584121,
"b": 0.2882323744865599,
"difference": -0.0162944204718522,
"ci95": [
-0.03009423881355039,
-0.003500244757008719
],
"resamples": 2000
},
"held": true,
"proven_worse": false,
"nll_warning": false
},
"test_relation": {
"accuracy": {
"metric": "accuracy",
"questions": 1950,
"pictures": 1453,
"a": 0.7107692307692308,
"b": 0.7548717948717949,
"difference": 0.04410256410256408,
"ci95": [
0.028018339276617454,
0.059034907597535935
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 1950,
"pictures": 1453,
"a": 0.5401162626494179,
"b": 0.4896254970901639,
"difference": -0.050490765559254025,
"ci95": [
-0.061521469864747935,
-0.038556349244042165
],
"resamples": 2000
},
"held": true,
"proven_worse": false,
"nll_warning": false
},
"test_size": {
"accuracy": {
"metric": "accuracy",
"questions": 1100,
"pictures": 975,
"a": 0.86,
"b": 0.8718181818181818,
"difference": 0.011818181818181839,
"ci95": [
-0.0009165902841430373,
0.025711662075298514
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 1100,
"pictures": 975,
"a": 0.35032188023362587,
"b": 0.3361044083603172,
"difference": -0.014217471873308685,
"ci95": [
-0.028957271795469974,
0.0003334966656227678
],
"resamples": 2000
},
"held": true,
"proven_worse": false,
"nll_warning": false
},
"test_position_flip": {
"accuracy": {
"metric": "accuracy",
"questions": 640,
"pictures": 636,
"a": 0.85,
"b": 0.846875,
"difference": -0.0031249999999999334,
"ci95": [
-0.021943573667711602,
0.014128728414442682
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 640,
"pictures": 636,
"a": 0.37288566299069764,
"b": 0.35918046542178717,
"difference": -0.013705197568910477,
"ci95": [
-0.029400569310068958,
0.002968877802506875
],
"resamples": 2000
},
"held": false,
"proven_worse": false,
"nll_warning": false
},
"test_relation_flip": {
"accuracy": {
"metric": "accuracy",
"questions": 1492,
"pictures": 1192,
"a": 0.6689008042895442,
"b": 0.6977211796246648,
"difference": 0.028820375335120607,
"ci95": [
0.009510869565217406,
0.04778156996587024
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 1492,
"pictures": 1192,
"a": 0.585918810016869,
"b": 0.5450049355513791,
"difference": -0.040913874465489886,
"ci95": [
-0.052748318876262856,
-0.029602205956100636
],
"resamples": 2000
},
"held": true,
"proven_worse": false,
"nll_warning": false
},
"test_bind_gqa": {
"accuracy": {
"metric": "accuracy",
"questions": 1484,
"pictures": 967,
"a": 0.6785714285714286,
"b": 0.6765498652291105,
"difference": -0.0020215633423180668,
"ci95": [
-0.01720578114246385,
0.013468013468013518
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 1484,
"pictures": 967,
"a": 0.56543531676673,
"b": 0.5592760342365289,
"difference": -0.006159282530201038,
"ci95": [
-0.015578151742373092,
0.0028278953758936742
],
"resamples": 2000
},
"held": true,
"proven_worse": false,
"nll_warning": false
},
"bench_lv_vqav2_yesno": {
"accuracy": {
"metric": "accuracy",
"questions": 4887,
"pictures": 2069,
"a": 0.724780028647432,
"b": 0.7249846531614488,
"difference": 0.0002046245140168157,
"ci95": [
-0.0075005067909993395,
0.008213984835720334
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 4887,
"pictures": 2069,
"a": 0.551391599768774,
"b": 0.5517217482239184,
"difference": 0.0003301484551443412,
"ci95": [
-0.004196828299816491,
0.005027468860727358
],
"resamples": 2000
},
"held": true,
"proven_worse": false,
"nll_warning": false
},
"bench_lv_aokvqa": {
"accuracy": {
"metric": "accuracy",
"questions": 1138,
"pictures": 1138,
"a": 0.6256590509666081,
"b": 0.6124780316344464,
"difference": -0.013181019332161759,
"ci95": [
-0.02899824253075567,
0.002636203866432374
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 1138,
"pictures": 1138,
"a": 0.9295007201330518,
"b": 0.9357389801118688,
"difference": 0.006238259978816996,
"ci95": [
-0.011132842065626924,
0.023444144517427712
],
"resamples": 2000
},
"held": false,
"proven_worse": false,
"nll_warning": false
},
"bench_lv_scienceqa": {
"accuracy": {
"metric": "accuracy",
"questions": 2097,
"pictures": 2097,
"a": 0.7663328564616119,
"b": 0.7892226990939437,
"difference": 0.022889842632331847,
"ci95": [
0.010968049594659024,
0.03528850739151168
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 2097,
"pictures": 2097,
"a": 0.505658483494164,
"b": 0.4709893216879221,
"difference": -0.034669161806241855,
"ci95": [
-0.04485873743892388,
-0.024565908726882524
],
"resamples": 2000
},
"held": true,
"proven_worse": false,
"nll_warning": false
},
"bench_lv_pope": {
"accuracy": {
"metric": "accuracy",
"questions": 9000,
"pictures": 500,
"a": 0.8502222222222222,
"b": 0.8481111111111111,
"difference": -0.0021111111111110636,
"ci95": [
-0.008888888888888946,
0.004555555555555535
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 9000,
"pictures": 500,
"a": 0.37292244276041664,
"b": 0.3721321698679672,
"difference": -0.0007902728924494284,
"ci95": [
-0.008944474453509432,
0.007254876618201789
],
"resamples": 2000
},
"held": true,
"proven_worse": false,
"nll_warning": false
},
"bench_whatsup": {
"accuracy": {
"metric": "accuracy",
"questions": 820,
"pictures": 820,
"a": 0.2841463414634146,
"b": 0.2865853658536585,
"difference": 0.0024390243902439046,
"ci95": [
-0.01341463414634142,
0.018292682926829285
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 820,
"pictures": 820,
"a": 1.4723660060388828,
"b": 1.4909160791754408,
"difference": 0.018550073136557987,
"ci95": [
0.005068862787158679,
0.031356396984290935
],
"resamples": 2000
},
"nll_warning": false
},
"dev_mix": {
"accuracy": {
"metric": "accuracy",
"questions": 5178,
"pictures": 3087,
"a": 0.753765932792584,
"b": 0.7553109308613364,
"difference": 0.0015449980687524123,
"ci95": [
-0.0056006179992273974,
0.009378663540445475
],
"resamples": 2000
},
"nll": {
"metric": "nll",
"questions": 5178,
"pictures": 3087,
"a": 0.5427544545511686,
"b": 0.5446964962906559,
"difference": 0.0019420417394873102,
"ci95": [
-0.0049800214856609015,
0.00870632376742897
],
"resamples": 2000
},
"nll_warning": false
}
},
"primary": true,
"protection": false,
"not_held": [
"test_vsr",
"test_vqa_yesno",
"test_position_flip",
"bench_lv_aokvqa"
],
"proven_worse": [],
"warnings": [],
"candidate": false
}