Files
News-ML-Decoder/eval/results/archive_20260923_0039_gold.json
T
Harwinder Sekhon 1d72e4006a Initial import: XAUUSD news sentiment models, evals, and data tooling
TF-IDF baseline (news_sentiment.py, models/xauusd_sentiment.joblib) and the
two-head DeBERTa-v3-small transformer (transformer/, news_transformer.py)
with its Google News crawler and DeepSeek labeller (data_collect/), the
golden-set and archive evals (eval/), and the 2026-09-22 and 2026-09-27
results and progress notes.

Transformer weights (model.pt, model.onnx) and the crawl database are not
tracked; rebuild with transformer/train.py and transformer/export_onnx.py.
The DeepSeek label cache in data/ is tracked.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-09-27 19:22:30 -04:00

1792 lines
38 KiB
JSON

{
"after_training": {
"all": {
"accuracy": 0.45872420262664165,
"confusion": [
[
186,
27,
106
],
[
74,
109,
192
],
[
129,
49,
194
]
],
"directional_accuracy": 0.5499276410998553,
"directional_hits": 380,
"directional_n": 691,
"hits": 489,
"macro_f1": 0.4545945057244492,
"n": 1066,
"sign_agreement": 0.6178861788617886,
"sign_hits": 380,
"sign_n": 615
},
"gold": {
"accuracy": 0.6822429906542056,
"confusion": [
[
84,
7,
19
],
[
8,
35,
16
],
[
42,
10,
100
]
],
"directional_accuracy": 0.7022900763358778,
"directional_hits": 184,
"directional_n": 262,
"hits": 219,
"macro_f1": 0.6720064440976077,
"n": 321,
"sign_agreement": 0.7510204081632653,
"sign_hits": 184,
"sign_n": 245
},
"macro": {
"accuracy": 0.3624161073825503,
"confusion": [
[
102,
20,
87
],
[
66,
74,
176
],
[
87,
39,
94
]
],
"directional_accuracy": 0.4568764568764569,
"directional_hits": 196,
"directional_n": 429,
"hits": 270,
"macro_f1": 0.36503325894343636,
"n": 745,
"sign_agreement": 0.5297297297297298,
"sign_hits": 196,
"sign_n": 370
}
},
"days": [
"2026-09-08",
"2026-09-09",
"2026-09-10",
"2026-09-11",
"2026-09-13",
"2026-09-14",
"2026-09-15",
"2026-09-16",
"2026-09-17",
"2026-09-18",
"2026-09-20",
"2026-09-21",
"2026-09-22"
],
"db": "../newsdata.db",
"extra_labels": "results/backfill_labels.jsonl",
"folds": [
{
"day": "2026-09-10",
"fit_seconds_full": 0.019556820999696356,
"fit_seconds_head": 0.010429643999941618,
"full": {
"accuracy": 0.5023255813953489,
"confusion": [
[
19,
20,
31
],
[
3,
39,
31
],
[
1,
21,
50
]
],
"directional_accuracy": 0.4859154929577465,
"directional_hits": 69,
"directional_n": 142,
"hits": 108,
"macro_f1": 0.487294777658609,
"n": 215,
"sign_agreement": 0.6831683168316832,
"sign_hits": 69,
"sign_n": 101
},
"head": {
"accuracy": 0.44651162790697674,
"confusion": [
[
21,
22,
27
],
[
1,
35,
37
],
[
9,
23,
40
]
],
"directional_accuracy": 0.4295774647887324,
"directional_hits": 61,
"directional_n": 142,
"hits": 96,
"macro_f1": 0.44263445952438385,
"n": 215,
"sign_agreement": 0.6288659793814433,
"sign_hits": 61,
"sign_n": 97
},
"keyword": {
"accuracy": 0.4558139534883721,
"confusion": [
[
23,
46,
1
],
[
3,
69,
1
],
[
11,
55,
6
]
],
"directional_accuracy": 0.20422535211267606,
"directional_hits": 29,
"directional_n": 142,
"hits": 98,
"macro_f1": 0.38260259220799203,
"n": 215,
"sign_agreement": 0.7073170731707317,
"sign_hits": 29,
"sign_n": 41
},
"n_test": 215,
"n_train": 49
},
{
"day": "2026-09-11",
"fit_seconds_full": 1.4326721040001758,
"fit_seconds_head": 2.0770400650003467,
"full": {
"accuracy": 0.4292682926829268,
"confusion": [
[
32,
20,
18
],
[
13,
23,
27
],
[
26,
13,
33
]
],
"directional_accuracy": 0.45774647887323944,
"directional_hits": 65,
"directional_n": 142,
"hits": 88,
"macro_f1": 0.42681844368953253,
"n": 205,
"sign_agreement": 0.5963302752293578,
"sign_hits": 65,
"sign_n": 109
},
"head": {
"accuracy": 0.424390243902439,
"confusion": [
[
36,
11,
23
],
[
12,
16,
35
],
[
21,
16,
35
]
],
"directional_accuracy": 0.5,
"directional_hits": 71,
"directional_n": 142,
"hits": 87,
"macro_f1": 0.4147049427353486,
"n": 205,
"sign_agreement": 0.6173913043478261,
"sign_hits": 71,
"sign_n": 115
},
"keyword": {
"accuracy": 0.40487804878048783,
"confusion": [
[
24,
43,
3
],
[
7,
54,
2
],
[
15,
52,
5
]
],
"directional_accuracy": 0.20422535211267606,
"directional_hits": 29,
"directional_n": 142,
"hits": 83,
"macro_f1": 0.3483927617415407,
"n": 205,
"sign_agreement": 0.6170212765957447,
"sign_hits": 29,
"sign_n": 47
},
"n_test": 205,
"n_train": 136
},
{
"day": "2026-09-13",
"fit_seconds_full": 1.6381026570002177,
"fit_seconds_head": 1.218333722999887,
"full": {
"accuracy": 0.4,
"confusion": [
[
1,
0,
0
],
[
0,
0,
3
],
[
0,
0,
1
]
],
"directional_accuracy": 1.0,
"directional_hits": 2,
"directional_n": 2,
"hits": 2,
"macro_f1": 0.4666666666666666,
"n": 5,
"sign_agreement": 1.0,
"sign_hits": 2,
"sign_n": 2
},
"head": {
"accuracy": 0.6,
"confusion": [
[
1,
0,
0
],
[
0,
1,
2
],
[
0,
0,
1
]
],
"directional_accuracy": 1.0,
"directional_hits": 2,
"directional_n": 2,
"hits": 3,
"macro_f1": 0.6666666666666666,
"n": 5,
"sign_agreement": 1.0,
"sign_hits": 2,
"sign_n": 2
},
"keyword": {
"accuracy": 0.8,
"confusion": [
[
1,
0,
0
],
[
0,
3,
0
],
[
0,
1,
0
]
],
"directional_accuracy": 0.5,
"directional_hits": 1,
"directional_n": 2,
"hits": 4,
"macro_f1": 0.6190476190476191,
"n": 5,
"sign_agreement": 1.0,
"sign_hits": 1,
"sign_n": 1
},
"n_test": 5,
"n_train": 227
},
{
"day": "2026-09-14",
"fit_seconds_full": 1.481112958000267,
"fit_seconds_head": 1.0056500370001231,
"full": {
"accuracy": 0.5601503759398496,
"confusion": [
[
88,
14,
42
],
[
7,
30,
40
],
[
8,
6,
31
]
],
"directional_accuracy": 0.6296296296296297,
"directional_hits": 119,
"directional_n": 189,
"hits": 149,
"macro_f1": 0.5257988718201595,
"n": 266,
"sign_agreement": 0.7041420118343196,
"sign_hits": 119,
"sign_n": 169
},
"head": {
"accuracy": 0.575187969924812,
"confusion": [
[
92,
12,
40
],
[
4,
29,
44
],
[
5,
8,
32
]
],
"directional_accuracy": 0.656084656084656,
"directional_hits": 124,
"directional_n": 189,
"hits": 153,
"macro_f1": 0.5362844654770121,
"n": 266,
"sign_agreement": 0.7337278106508875,
"sign_hits": 124,
"sign_n": 169
},
"keyword": {
"accuracy": 0.5075187969924813,
"confusion": [
[
61,
77,
6
],
[
7,
69,
1
],
[
5,
35,
5
]
],
"directional_accuracy": 0.3492063492063492,
"directional_hits": 66,
"directional_n": 189,
"hits": 135,
"macro_f1": 0.42417809966276027,
"n": 266,
"sign_agreement": 0.8571428571428571,
"sign_hits": 66,
"sign_n": 77
},
"n_test": 266,
"n_train": 227
},
{
"day": "2026-09-15",
"fit_seconds_full": 1.3284068719999595,
"fit_seconds_head": 1.4787338000001,
"full": {
"accuracy": 0.5023041474654378,
"confusion": [
[
55,
5,
30
],
[
8,
26,
34
],
[
19,
12,
28
]
],
"directional_accuracy": 0.5570469798657718,
"directional_hits": 83,
"directional_n": 149,
"hits": 109,
"macro_f1": 0.4929547597805721,
"n": 217,
"sign_agreement": 0.6287878787878788,
"sign_hits": 83,
"sign_n": 132
},
"head": {
"accuracy": 0.5069124423963134,
"confusion": [
[
52,
7,
31
],
[
4,
26,
38
],
[
14,
13,
32
]
],
"directional_accuracy": 0.5637583892617449,
"directional_hits": 84,
"directional_n": 149,
"hits": 110,
"macro_f1": 0.5020467836257311,
"n": 217,
"sign_agreement": 0.6511627906976745,
"sign_hits": 84,
"sign_n": 129
},
"keyword": {
"accuracy": 0.4009216589861751,
"confusion": [
[
24,
59,
7
],
[
4,
59,
5
],
[
11,
44,
4
]
],
"directional_accuracy": 0.18791946308724833,
"directional_hits": 28,
"directional_n": 149,
"hits": 87,
"macro_f1": 0.3306010560611167,
"n": 217,
"sign_agreement": 0.6086956521739131,
"sign_hits": 28,
"sign_n": 46
},
"n_test": 217,
"n_train": 349
},
{
"day": "2026-09-16",
"fit_seconds_full": 2.5788320259998727,
"fit_seconds_head": 2.3951005039998563,
"full": {
"accuracy": 0.403125,
"confusion": [
[
43,
25,
37
],
[
26,
39,
44
],
[
38,
21,
47
]
],
"directional_accuracy": 0.4265402843601896,
"directional_hits": 90,
"directional_n": 211,
"hits": 129,
"macro_f1": 0.40314387824599845,
"n": 320,
"sign_agreement": 0.5454545454545454,
"sign_hits": 90,
"sign_n": 165
},
"head": {
"accuracy": 0.403125,
"confusion": [
[
51,
14,
40
],
[
28,
27,
54
],
[
34,
21,
51
]
],
"directional_accuracy": 0.4834123222748815,
"directional_hits": 102,
"directional_n": 211,
"hits": 129,
"macro_f1": 0.39668462797770765,
"n": 320,
"sign_agreement": 0.5795454545454546,
"sign_hits": 102,
"sign_n": 176
},
"keyword": {
"accuracy": 0.41875,
"confusion": [
[
32,
67,
6
],
[
11,
90,
8
],
[
15,
79,
12
]
],
"directional_accuracy": 0.20853080568720378,
"directional_hits": 44,
"directional_n": 211,
"hits": 134,
"macro_f1": 0.36539844968759344,
"n": 320,
"sign_agreement": 0.676923076923077,
"sign_hits": 44,
"sign_n": 65
},
"n_test": 320,
"n_train": 433
},
{
"day": "2026-09-17",
"fit_seconds_full": 1.8287062199997308,
"fit_seconds_head": 2.3719224739998026,
"full": {
"accuracy": 0.5107142857142857,
"confusion": [
[
64,
13,
31
],
[
30,
25,
28
],
[
23,
12,
54
]
],
"directional_accuracy": 0.5989847715736041,
"directional_hits": 118,
"directional_n": 197,
"hits": 143,
"macro_f1": 0.49316073461982796,
"n": 280,
"sign_agreement": 0.686046511627907,
"sign_hits": 118,
"sign_n": 172
},
"head": {
"accuracy": 0.5464285714285714,
"confusion": [
[
70,
5,
33
],
[
13,
23,
47
],
[
23,
6,
60
]
],
"directional_accuracy": 0.6598984771573604,
"directional_hits": 130,
"directional_n": 197,
"hits": 153,
"macro_f1": 0.5237951559626457,
"n": 280,
"sign_agreement": 0.6989247311827957,
"sign_hits": 130,
"sign_n": 186
},
"keyword": {
"accuracy": 0.4107142857142857,
"confusion": [
[
37,
65,
6
],
[
13,
66,
4
],
[
11,
66,
12
]
],
"directional_accuracy": 0.24873096446700507,
"directional_hits": 49,
"directional_n": 197,
"hits": 115,
"macro_f1": 0.3751715367099983,
"n": 280,
"sign_agreement": 0.7424242424242424,
"sign_hits": 49,
"sign_n": 66
},
"n_test": 280,
"n_train": 547
},
{
"day": "2026-09-18",
"fit_seconds_full": 3.1369762259996605,
"fit_seconds_head": 2.6367428789999394,
"full": {
"accuracy": 0.45454545454545453,
"confusion": [
[
32,
3,
40
],
[
27,
21,
55
],
[
46,
15,
102
]
],
"directional_accuracy": 0.5630252100840336,
"directional_hits": 134,
"directional_n": 238,
"hits": 155,
"macro_f1": 0.4059989567031821,
"n": 341,
"sign_agreement": 0.6090909090909091,
"sign_hits": 134,
"sign_n": 220
},
"head": {
"accuracy": 0.4780058651026393,
"confusion": [
[
39,
0,
36
],
[
18,
21,
64
],
[
46,
14,
103
]
],
"directional_accuracy": 0.5966386554621849,
"directional_hits": 142,
"directional_n": 238,
"hits": 163,
"macro_f1": 0.4351305344442042,
"n": 341,
"sign_agreement": 0.6339285714285714,
"sign_hits": 142,
"sign_n": 224
},
"keyword": {
"accuracy": 0.41055718475073316,
"confusion": [
[
28,
34,
13
],
[
5,
89,
9
],
[
20,
120,
23
]
],
"directional_accuracy": 0.21428571428571427,
"directional_hits": 51,
"directional_n": 238,
"hits": 140,
"macro_f1": 0.3910349044019564,
"n": 341,
"sign_agreement": 0.6071428571428571,
"sign_hits": 51,
"sign_n": 84
},
"n_test": 341,
"n_train": 646
},
{
"day": "2026-09-20",
"fit_seconds_full": 1.9606132069998239,
"fit_seconds_head": 2.9577391190000526,
"full": {
"accuracy": 0.5384615384615384,
"confusion": [
[
1,
3,
3
],
[
0,
7,
1
],
[
2,
3,
6
]
],
"directional_accuracy": 0.3888888888888889,
"directional_hits": 7,
"directional_n": 18,
"hits": 14,
"macro_f1": 0.4793650793650794,
"n": 26,
"sign_agreement": 0.5833333333333334,
"sign_hits": 7,
"sign_n": 12
},
"head": {
"accuracy": 0.6923076923076923,
"confusion": [
[
2,
2,
3
],
[
0,
6,
2
],
[
1,
0,
10
]
],
"directional_accuracy": 0.6666666666666666,
"directional_hits": 12,
"directional_n": 18,
"hits": 18,
"macro_f1": 0.6397435897435897,
"n": 26,
"sign_agreement": 0.75,
"sign_hits": 12,
"sign_n": 16
},
"keyword": {
"accuracy": 0.46153846153846156,
"confusion": [
[
0,
7,
0
],
[
0,
8,
0
],
[
0,
7,
4
]
],
"directional_accuracy": 0.2222222222222222,
"directional_hits": 4,
"directional_n": 18,
"hits": 12,
"macro_f1": 0.35555555555555557,
"n": 26,
"sign_agreement": 1.0,
"sign_hits": 4,
"sign_n": 4
},
"n_test": 26,
"n_train": 750
},
{
"day": "2026-09-21",
"fit_seconds_full": 2.89376103099994,
"fit_seconds_head": 3.109174907000124,
"full": {
"accuracy": 0.5313351498637602,
"confusion": [
[
87,
7,
38
],
[
18,
45,
80
],
[
19,
10,
63
]
],
"directional_accuracy": 0.6696428571428571,
"directional_hits": 150,
"directional_n": 224,
"hits": 195,
"macro_f1": 0.5267501172607879,
"n": 367,
"sign_agreement": 0.7246376811594203,
"sign_hits": 150,
"sign_n": 207
},
"head": {
"accuracy": 0.5694822888283378,
"confusion": [
[
80,
5,
47
],
[
16,
53,
74
],
[
10,
6,
76
]
],
"directional_accuracy": 0.6964285714285714,
"directional_hits": 156,
"directional_n": 224,
"hits": 209,
"macro_f1": 0.570099253114147,
"n": 367,
"sign_agreement": 0.7323943661971831,
"sign_hits": 156,
"sign_n": 213
},
"keyword": {
"accuracy": 0.4986376021798365,
"confusion": [
[
53,
72,
7
],
[
4,
125,
14
],
[
5,
82,
5
]
],
"directional_accuracy": 0.25892857142857145,
"directional_hits": 58,
"directional_n": 224,
"hits": 183,
"macro_f1": 0.4078515256335195,
"n": 367,
"sign_agreement": 0.8285714285714286,
"sign_hits": 58,
"sign_n": 70
},
"n_test": 367,
"n_train": 756
},
{
"day": "2026-09-22",
"fit_seconds_full": 2.1956956489998447,
"fit_seconds_head": 2.8106658059996334,
"full": {
"accuracy": 0.4513888888888889,
"confusion": [
[
57,
3,
33
],
[
24,
34,
54
],
[
35,
9,
39
]
],
"directional_accuracy": 0.5454545454545454,
"directional_hits": 96,
"directional_n": 176,
"hits": 130,
"macro_f1": 0.44968001130559426,
"n": 288,
"sign_agreement": 0.5853658536585366,
"sign_hits": 96,
"sign_n": 164
},
"head": {
"accuracy": 0.4548611111111111,
"confusion": [
[
49,
3,
41
],
[
15,
34,
63
],
[
31,
4,
48
]
],
"directional_accuracy": 0.5511363636363636,
"directional_hits": 97,
"directional_n": 176,
"hits": 131,
"macro_f1": 0.45807722616233254,
"n": 288,
"sign_agreement": 0.5739644970414202,
"sign_hits": 97,
"sign_n": 169
},
"keyword": {
"accuracy": 0.4895833333333333,
"confusion": [
[
37,
49,
7
],
[
4,
99,
9
],
[
17,
61,
5
]
],
"directional_accuracy": 0.23863636363636365,
"directional_hits": 42,
"directional_n": 176,
"hits": 141,
"macro_f1": 0.40101416707531706,
"n": 288,
"sign_agreement": 0.6363636363636364,
"sign_hits": 42,
"sign_n": 66
},
"n_test": 288,
"n_train": 885
}
],
"generated_at": "20260923_0039_gold",
"in_sample": {
"headline_description": {
"accuracy": 0.5608646188850968,
"confusion": [
[
624,
73,
219
],
[
182,
335,
379
],
[
222,
83,
520
]
],
"directional_accuracy": 0.6570936243538197,
"directional_hits": 1144,
"directional_n": 1741,
"hits": 1479,
"macro_f1": 0.5534290089278285,
"mae": 0.2846768854759196,
"n": 2637,
"sign_agreement": 0.7217665615141956,
"sign_hits": 1144,
"sign_n": 1585,
"spearman": 0.5336683695481723
},
"headline_only": {
"accuracy": 0.5733788395904437,
"confusion": [
[
652,
43,
221
],
[
152,
304,
440
],
[
210,
59,
556
]
],
"directional_accuracy": 0.6938541068351523,
"directional_hits": 1208,
"directional_n": 1741,
"hits": 1512,
"macro_f1": 0.5623952358379628,
"mae": 0.30869570231323473,
"n": 2637,
"sign_agreement": 0.7370347773032336,
"sign_hits": 1208,
"sign_n": 1639,
"spearman": 0.554711637369288
}
},
"label_distribution": {
"bearish": 916,
"bullish": 825,
"neutral": 896
},
"model": {
"label_distribution": {
"bearish": 250,
"bullish": 221,
"neutral": 135
},
"model_path": "/home/harwinder/Development/ML-NEWS/models/xauusd_sentiment.joblib",
"model_type": "domain_augmented_tfidf_logistic_regression",
"neutral_threshold": 0.05,
"sklearn_version": "1.9.1",
"trained_at": "2026-09-17T18:54:46.714699+00:00",
"training_examples": 606
},
"pooled_oof": {
"full": {
"accuracy": 0.48300395256916995,
"confusion": [
[
479,
113,
303
],
[
156,
289,
397
],
[
217,
122,
454
]
],
"directional_accuracy": 0.5527251184834123,
"directional_hits": 933,
"directional_n": 1688,
"hits": 1222,
"macro_f1": 0.47928678930733953,
"mae": 0.31561411067193673,
"n": 2530,
"sign_agreement": 0.6421197522367516,
"sign_hits": 933,
"sign_n": 1453,
"spearman": 0.3274765580255068
},
"full_gold": {
"accuracy": 0.6780219780219781,
"confusion": [
[
239,
29,
85
],
[
17,
132,
39
],
[
68,
55,
246
]
],
"directional_accuracy": 0.6717451523545707,
"directional_hits": 485,
"directional_n": 722,
"hits": 617,
"macro_f1": 0.6750953410681971,
"mae": 0.31894142857142854,
"n": 910,
"sign_agreement": 0.7601880877742947,
"sign_hits": 485,
"sign_n": 638,
"spearman": 0.5523303019577694
},
"full_macro": {
"accuracy": 0.3734567901234568,
"confusion": [
[
240,
84,
218
],
[
139,
157,
358
],
[
149,
67,
208
]
],
"directional_accuracy": 0.463768115942029,
"directional_hits": 448,
"directional_n": 966,
"hits": 605,
"macro_f1": 0.3731241060572001,
"mae": 0.31374506172839506,
"n": 1620,
"sign_agreement": 0.5496932515337424,
"sign_hits": 448,
"sign_n": 815,
"spearman": 0.13211416842558318
},
"full_since_2026-09-18_all": {
"accuracy": 0.48336594911937375,
"confusion": [
[
177,
16,
114
],
[
69,
107,
190
],
[
102,
37,
210
]
],
"directional_accuracy": 0.5899390243902439,
"directional_hits": 387,
"directional_n": 656,
"hits": 494,
"macro_f1": 0.4779921711293548,
"n": 1022,
"sign_agreement": 0.6417910447761194,
"sign_hits": 387,
"sign_n": 603
},
"full_since_2026-09-18_gold": {
"accuracy": 0.7380191693290735,
"confusion": [
[
77,
5,
27
],
[
5,
40,
14
],
[
24,
7,
114
]
],
"directional_accuracy": 0.7519685039370079,
"directional_hits": 191,
"directional_n": 254,
"hits": 231,
"macro_f1": 0.7323332634960543,
"n": 313,
"sign_agreement": 0.7892561983471075,
"sign_hits": 191,
"sign_n": 242
},
"full_since_2026-09-18_macro": {
"accuracy": 0.3709449929478138,
"confusion": [
[
100,
11,
87
],
[
64,
67,
176
],
[
78,
30,
96
]
],
"directional_accuracy": 0.48756218905472637,
"directional_hits": 196,
"directional_n": 402,
"hits": 263,
"macro_f1": 0.37282240539746586,
"n": 709,
"sign_agreement": 0.5429362880886427,
"sign_hits": 196,
"sign_n": 361
},
"full_train_ge_300": {
"accuracy": 0.4758020663404024,
"confusion": [
[
339,
59,
212
],
[
133,
197,
296
],
[
182,
82,
339
]
],
"directional_accuracy": 0.5589447650453421,
"directional_hits": 678,
"directional_n": 1213,
"hits": 875,
"macro_f1": 0.4708974349686336,
"mae": 0.31355356171832516,
"n": 1839,
"sign_agreement": 0.6324626865671642,
"sign_hits": 678,
"sign_n": 1072,
"spearman": 0.3119104790099809
},
"head": {
"accuracy": 0.49486166007905136,
"confusion": [
[
493,
81,
321
],
[
111,
271,
460
],
[
194,
111,
488
]
],
"directional_accuracy": 0.5811611374407583,
"directional_hits": 981,
"directional_n": 1688,
"hits": 1252,
"macro_f1": 0.49035021582717747,
"mae": 0.3531845454545455,
"n": 2530,
"sign_agreement": 0.6557486631016043,
"sign_hits": 981,
"sign_n": 1496,
"spearman": 0.3493073941083059
},
"keyword": {
"accuracy": 0.4474308300395257,
"confusion": [
[
320,
519,
56
],
[
58,
731,
53
],
[
110,
602,
81
]
],
"directional_accuracy": 0.23755924170616113,
"directional_hits": 401,
"directional_n": 1688,
"hits": 1132,
"macro_f1": 0.39008373087437453,
"n": 2530,
"sign_agreement": 0.7072310405643739,
"sign_hits": 401,
"sign_n": 567
}
},
"reprobe": {},
"routing": [
{
"confidence_min": 0.0,
"hybrid_agreement": 0.48300395256916995,
"kept": 2530,
"kept_accuracy": 0.48300395256916995,
"kept_share": 1.0,
"llm_calls_saved_share": 1.0,
"margin_min": 0.0
},
{
"confidence_min": 0.0,
"hybrid_agreement": 0.6284584980237155,
"kept": 1917,
"kept_accuracy": 0.5096504955659885,
"kept_share": 0.7577075098814229,
"llm_calls_saved_share": 0.7577075098814229,
"margin_min": 0.1
},
{
"confidence_min": 0.0,
"hybrid_agreement": 0.7399209486166007,
"kept": 1430,
"kept_accuracy": 0.5398601398601398,
"kept_share": 0.5652173913043478,
"llm_calls_saved_share": 0.5652173913043478,
"margin_min": 0.2
},
{
"confidence_min": 0.0,
"hybrid_agreement": 0.833201581027668,
"kept": 1015,
"kept_accuracy": 0.5842364532019705,
"kept_share": 0.40118577075098816,
"llm_calls_saved_share": 0.40118577075098816,
"margin_min": 0.3
},
{
"confidence_min": 0.0,
"hybrid_agreement": 0.8964426877470356,
"kept": 697,
"kept_accuracy": 0.624103299856528,
"kept_share": 0.2754940711462451,
"llm_calls_saved_share": 0.2754940711462451,
"margin_min": 0.4
},
{
"confidence_min": 0.5,
"hybrid_agreement": 0.7154150197628458,
"kept": 1562,
"kept_accuracy": 0.5390524967989757,
"kept_share": 0.6173913043478261,
"llm_calls_saved_share": 0.6173913043478261,
"margin_min": 0.0
},
{
"confidence_min": 0.5,
"hybrid_agreement": 0.7189723320158102,
"kept": 1543,
"kept_accuracy": 0.5392093324692158,
"kept_share": 0.6098814229249012,
"llm_calls_saved_share": 0.6098814229249012,
"margin_min": 0.1
},
{
"confidence_min": 0.5,
"hybrid_agreement": 0.7537549407114624,
"kept": 1378,
"kept_accuracy": 0.5478955007256894,
"kept_share": 0.5446640316205533,
"llm_calls_saved_share": 0.5446640316205533,
"margin_min": 0.2
},
{
"confidence_min": 0.5,
"hybrid_agreement": 0.833201581027668,
"kept": 1015,
"kept_accuracy": 0.5842364532019705,
"kept_share": 0.40118577075098816,
"llm_calls_saved_share": 0.40118577075098816,
"margin_min": 0.3
},
{
"confidence_min": 0.5,
"hybrid_agreement": 0.8964426877470356,
"kept": 697,
"kept_accuracy": 0.624103299856528,
"kept_share": 0.2754940711462451,
"llm_calls_saved_share": 0.2754940711462451,
"margin_min": 0.4
},
{
"confidence_min": 0.6,
"hybrid_agreement": 0.8656126482213439,
"kept": 876,
"kept_accuracy": 0.6118721461187214,
"kept_share": 0.34624505928853755,
"llm_calls_saved_share": 0.34624505928853755,
"margin_min": 0.0
},
{
"confidence_min": 0.6,
"hybrid_agreement": 0.8656126482213439,
"kept": 876,
"kept_accuracy": 0.6118721461187214,
"kept_share": 0.34624505928853755,
"llm_calls_saved_share": 0.34624505928853755,
"margin_min": 0.1
},
{
"confidence_min": 0.6,
"hybrid_agreement": 0.8656126482213439,
"kept": 876,
"kept_accuracy": 0.6118721461187214,
"kept_share": 0.34624505928853755,
"llm_calls_saved_share": 0.34624505928853755,
"margin_min": 0.2
},
{
"confidence_min": 0.6,
"hybrid_agreement": 0.8703557312252964,
"kept": 850,
"kept_accuracy": 0.6141176470588235,
"kept_share": 0.3359683794466403,
"llm_calls_saved_share": 0.3359683794466403,
"margin_min": 0.3
},
{
"confidence_min": 0.6,
"hybrid_agreement": 0.8964426877470356,
"kept": 697,
"kept_accuracy": 0.624103299856528,
"kept_share": 0.2754940711462451,
"llm_calls_saved_share": 0.2754940711462451,
"margin_min": 0.4
},
{
"confidence_min": 0.7,
"hybrid_agreement": 0.9561264822134388,
"kept": 436,
"kept_accuracy": 0.7454128440366973,
"kept_share": 0.17233201581027668,
"llm_calls_saved_share": 0.17233201581027668,
"margin_min": 0.0
},
{
"confidence_min": 0.7,
"hybrid_agreement": 0.9561264822134388,
"kept": 436,
"kept_accuracy": 0.7454128440366973,
"kept_share": 0.17233201581027668,
"llm_calls_saved_share": 0.17233201581027668,
"margin_min": 0.1
},
{
"confidence_min": 0.7,
"hybrid_agreement": 0.9561264822134388,
"kept": 436,
"kept_accuracy": 0.7454128440366973,
"kept_share": 0.17233201581027668,
"llm_calls_saved_share": 0.17233201581027668,
"margin_min": 0.2
},
{
"confidence_min": 0.7,
"hybrid_agreement": 0.9561264822134388,
"kept": 436,
"kept_accuracy": 0.7454128440366973,
"kept_share": 0.17233201581027668,
"llm_calls_saved_share": 0.17233201581027668,
"margin_min": 0.3
},
{
"confidence_min": 0.7,
"hybrid_agreement": 0.9561264822134388,
"kept": 436,
"kept_accuracy": 0.7454128440366973,
"kept_share": 0.17233201581027668,
"llm_calls_saved_share": 0.17233201581027668,
"margin_min": 0.4
},
{
"confidence_min": 0.8,
"hybrid_agreement": 0.9905138339920949,
"kept": 216,
"kept_accuracy": 0.8888888888888888,
"kept_share": 0.08537549407114625,
"llm_calls_saved_share": 0.08537549407114625,
"margin_min": 0.0
},
{
"confidence_min": 0.8,
"hybrid_agreement": 0.9905138339920949,
"kept": 216,
"kept_accuracy": 0.8888888888888888,
"kept_share": 0.08537549407114625,
"llm_calls_saved_share": 0.08537549407114625,
"margin_min": 0.1
},
{
"confidence_min": 0.8,
"hybrid_agreement": 0.9905138339920949,
"kept": 216,
"kept_accuracy": 0.8888888888888888,
"kept_share": 0.08537549407114625,
"llm_calls_saved_share": 0.08537549407114625,
"margin_min": 0.2
},
{
"confidence_min": 0.8,
"hybrid_agreement": 0.9905138339920949,
"kept": 216,
"kept_accuracy": 0.8888888888888888,
"kept_share": 0.08537549407114625,
"llm_calls_saved_share": 0.08537549407114625,
"margin_min": 0.3
},
{
"confidence_min": 0.8,
"hybrid_agreement": 0.9905138339920949,
"kept": 216,
"kept_accuracy": 0.8888888888888888,
"kept_share": 0.08537549407114625,
"llm_calls_saved_share": 0.08537549407114625,
"margin_min": 0.4
},
{
"confidence_min": 0.9,
"hybrid_agreement": 1.0,
"kept": 84,
"kept_accuracy": 1.0,
"kept_share": 0.03320158102766799,
"llm_calls_saved_share": 0.03320158102766799,
"margin_min": 0.0
},
{
"confidence_min": 0.9,
"hybrid_agreement": 1.0,
"kept": 84,
"kept_accuracy": 1.0,
"kept_share": 0.03320158102766799,
"llm_calls_saved_share": 0.03320158102766799,
"margin_min": 0.1
},
{
"confidence_min": 0.9,
"hybrid_agreement": 1.0,
"kept": 84,
"kept_accuracy": 1.0,
"kept_share": 0.03320158102766799,
"llm_calls_saved_share": 0.03320158102766799,
"margin_min": 0.2
},
{
"confidence_min": 0.9,
"hybrid_agreement": 1.0,
"kept": 84,
"kept_accuracy": 1.0,
"kept_share": 0.03320158102766799,
"llm_calls_saved_share": 0.03320158102766799,
"margin_min": 0.3
},
{
"confidence_min": 0.9,
"hybrid_agreement": 1.0,
"kept": 84,
"kept_accuracy": 1.0,
"kept_share": 0.03320158102766799,
"llm_calls_saved_share": 0.03320158102766799,
"margin_min": 0.4
}
],
"rows": 2637,
"rows_per_day": {
"2026-09-08": 71,
"2026-09-09": 36,
"2026-09-10": 215,
"2026-09-11": 205,
"2026-09-13": 5,
"2026-09-14": 266,
"2026-09-15": 217,
"2026-09-16": 320,
"2026-09-17": 280,
"2026-09-18": 341,
"2026-09-20": 26,
"2026-09-21": 367,
"2026-09-22": 288
},
"threshold": 0.05,
"train_scope": "gold"
}