PedramR commited on
Commit
ef92392
Β·
verified Β·
1 Parent(s): a7201d9

verifier report

Browse files
verifier-qwen3.5-2b-v2/report-20260929-1740/judged.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
verifier-qwen3.5-2b-v2/report-20260929-1740/progress.json ADDED
@@ -0,0 +1,28 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "stage": "judge",
3
+ "done": 360,
4
+ "total": 360,
5
+ "eta_minutes": 0.0,
6
+ "updated_at": "2026-09-29 17:37:29",
7
+ "cost_this_run_usd": 0.0,
8
+ "cost_projected_this_run_usd": 0.0,
9
+ "cost_by_model_usd": {
10
+ "xiaomi/mimo-v2.6-flash@think": 0.0026,
11
+ "deepseek/deepseek-v4.1-flash@think": 0.0192
12
+ },
13
+ "cost_total_usd": 0.0218,
14
+ "judge": {
15
+ "xiaomi/mimo-v2.6-flash@think": {
16
+ "judged": 14,
17
+ "failed": 0,
18
+ "errors": 346,
19
+ "cost_usd": 0.0026
20
+ },
21
+ "deepseek/deepseek-v4.1-flash@think": {
22
+ "judged": 6,
23
+ "failed": 3,
24
+ "errors": 0,
25
+ "cost_usd": 0.0192
26
+ }
27
+ }
28
+ }
verifier-qwen3.5-2b-v2/report-20260929-1740/report.json ADDED
@@ -0,0 +1,923 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dropped_respellings": 0,
3
+ "candidates": 8781,
4
+ "bins": {
5
+ "0.0": {
6
+ "candidates": 4951,
7
+ "sampled": 60,
8
+ "undecided": 57,
9
+ "rejected": 3,
10
+ "precision": 0.0
11
+ },
12
+ "0.3": {
13
+ "candidates": 1601,
14
+ "sampled": 60,
15
+ "undecided": 58,
16
+ "accepted": 1,
17
+ "rejected": 1,
18
+ "precision": 0.5
19
+ },
20
+ "0.5": {
21
+ "candidates": 1014,
22
+ "sampled": 60,
23
+ "undecided": 58,
24
+ "accepted": 1,
25
+ "rejected": 1,
26
+ "precision": 0.5
27
+ },
28
+ "0.7": {
29
+ "candidates": 523,
30
+ "sampled": 60,
31
+ "undecided": 59,
32
+ "accepted": 1,
33
+ "precision": 1.0
34
+ },
35
+ "0.8": {
36
+ "candidates": 414,
37
+ "sampled": 60,
38
+ "undecided": 59,
39
+ "accepted": 1,
40
+ "precision": 1.0
41
+ },
42
+ "0.9": {
43
+ "candidates": 278,
44
+ "sampled": 60,
45
+ "undecided": 58,
46
+ "accepted": 2,
47
+ "precision": 1.0
48
+ }
49
+ },
50
+ "judge_cost_usd": 0.0218,
51
+ "slices": {
52
+ "main": {
53
+ "whisper": {
54
+ "n_examples": 1978,
55
+ "wer": 0.19286111707900203,
56
+ "wer_zwnj_normalized": 0.18077200636819868,
57
+ "exact_match": 0.006066734074823054,
58
+ "hallucination_rate": 0.0,
59
+ "rows_changed": 0.0,
60
+ "fix_rate": 0.0,
61
+ "preservation_rate": 1.0,
62
+ "targeted_score": 0.5,
63
+ "fix_rate_lenient": 0.0,
64
+ "preservation_rate_lenient": 1.0,
65
+ "targeted_score_lenient": 0.5,
66
+ "fixed_lenient": 0,
67
+ "broken_lenient": 0,
68
+ "net_fixed_lenient": 0,
69
+ "broken_per_100_correct": 0.0,
70
+ "wer_clitic": 0.1839621023260069,
71
+ "fixed_clitic": 0,
72
+ "broken_clitic": 0,
73
+ "net_fixed_clitic": 0,
74
+ "broken_per_100_correct_clitic": 0.0,
75
+ "fix_rate_recoverable": 0.0,
76
+ "unrecoverable_targets": 966,
77
+ "entity_fix_rate": 0.0,
78
+ "entity_keep_rate": 1.0,
79
+ "entity_spans_whisper_wrong": 724,
80
+ "entity_spans_whisper_right": 1289,
81
+ "applied": 0,
82
+ "judged_precision": null,
83
+ "est_real_fixes": null
84
+ },
85
+ "rules only": {
86
+ "n_examples": 1978,
87
+ "wer": 0.18624502140177493,
88
+ "wer_zwnj_normalized": 0.17446440474419783,
89
+ "exact_match": 0.006572295247724975,
90
+ "hallucination_rate": 0.0,
91
+ "rows_changed": 0.6875631951466128,
92
+ "fix_rate": 0.05559820155718829,
93
+ "preservation_rate": 0.9996089846635096,
94
+ "targeted_score": 0.5276035931103489,
95
+ "fix_rate_lenient": 0.05892479464856623,
96
+ "preservation_rate_lenient": 0.9995903045402917,
97
+ "targeted_score_lenient": 0.5292575495944289,
98
+ "fixed_lenient": 3594,
99
+ "broken_lenient": 123,
100
+ "net_fixed_lenient": 3471,
101
+ "broken_per_100_correct": 0.04096954597082835,
102
+ "wer_clitic": 0.17667474784438397,
103
+ "fixed_clitic": 3571,
104
+ "broken_clitic": 137,
105
+ "net_fixed_clitic": 3434,
106
+ "broken_per_100_correct_clitic": 0.0468321630989902,
107
+ "fix_rate_recoverable": 0.059573192063571395,
108
+ "unrecoverable_targets": 966,
109
+ "entity_fix_rate": 0.009668508287292817,
110
+ "entity_keep_rate": 0.9953452288595811,
111
+ "entity_spans_whisper_wrong": 724,
112
+ "entity_spans_whisper_right": 1289,
113
+ "applied": 0,
114
+ "judged_precision": null,
115
+ "est_real_fixes": null
116
+ },
117
+ "accept all": {
118
+ "n_examples": 1978,
119
+ "wer": 0.1833225911296942,
120
+ "wer_zwnj_normalized": 0.17195238173935556,
121
+ "exact_match": 0.006572295247724975,
122
+ "hallucination_rate": 0.0,
123
+ "rows_changed": 0.9408493427704753,
124
+ "fix_rate": 0.12117556749643602,
125
+ "preservation_rate": 0.9893423255720688,
126
+ "targeted_score": 0.5552589465342525,
127
+ "fix_rate_lenient": 0.12734248192415523,
128
+ "preservation_rate_lenient": 0.9893679031919607,
129
+ "targeted_score_lenient": 0.558355192558058,
130
+ "fixed_lenient": 7767,
131
+ "broken_lenient": 3192,
132
+ "net_fixed_lenient": 4575,
133
+ "broken_per_100_correct": 1.0632096808039357,
134
+ "wer_clitic": 0.17458488667291813,
135
+ "fixed_clitic": 7419,
136
+ "broken_clitic": 3153,
137
+ "net_fixed_clitic": 4266,
138
+ "broken_per_100_correct_clitic": 1.0778234324898985,
139
+ "fix_rate_recoverable": 0.12885868026054942,
140
+ "unrecoverable_targets": 966,
141
+ "entity_fix_rate": 0.11187845303867404,
142
+ "entity_keep_rate": 0.9317300232738557,
143
+ "entity_spans_whisper_wrong": 724,
144
+ "entity_spans_whisper_right": 1289,
145
+ "applied": 8781,
146
+ "judged_precision": 0.2872679649242683,
147
+ "est_real_fixes": 2522.5
148
+ },
149
+ "verifier@0.3": {
150
+ "n_examples": 1978,
151
+ "wer": 0.1851102039916927,
152
+ "wer_zwnj_normalized": 0.17342048290336975,
153
+ "exact_match": 0.006572295247724975,
154
+ "hallucination_rate": 0.0,
155
+ "rows_changed": 0.8650151668351871,
156
+ "fix_rate": 0.08191687685053185,
157
+ "preservation_rate": 0.9955283887160327,
158
+ "targeted_score": 0.5387226327832823,
159
+ "fix_rate_lenient": 0.08650172970668765,
160
+ "preservation_rate_lenient": 0.995533320231961,
161
+ "targeted_score_lenient": 0.5410175249693243,
162
+ "fixed_lenient": 5276,
163
+ "broken_lenient": 1341,
164
+ "net_fixed_lenient": 3935,
165
+ "broken_per_100_correct": 0.4466679768039091,
166
+ "wer_clitic": 0.1757055782997507,
167
+ "fixed_clitic": 5170,
168
+ "broken_clitic": 1335,
169
+ "net_fixed_clitic": 3835,
170
+ "broken_per_100_correct_clitic": 0.456357209760233,
171
+ "fix_rate_recoverable": 0.08754393856098089,
172
+ "unrecoverable_targets": 966,
173
+ "entity_fix_rate": 0.07044198895027624,
174
+ "entity_keep_rate": 0.9650892164468581,
175
+ "entity_spans_whisper_wrong": 724,
176
+ "entity_spans_whisper_right": 1289,
177
+ "applied": 3830,
178
+ "judged_precision": 0.6586161879895561,
179
+ "est_real_fixes": 2522.5
180
+ },
181
+ "verifier@0.5": {
182
+ "n_examples": 1978,
183
+ "wer": 0.18571066563100805,
184
+ "wer_zwnj_normalized": 0.1739906460196004,
185
+ "exact_match": 0.006572295247724975,
186
+ "hallucination_rate": 0.0,
187
+ "rows_changed": 0.8124368048533872,
188
+ "fix_rate": 0.06968182601475725,
189
+ "preservation_rate": 0.997286286724528,
190
+ "targeted_score": 0.5334840563696426,
191
+ "fix_rate_lenient": 0.07361500500057383,
192
+ "preservation_rate_lenient": 0.9972753586500701,
193
+ "targeted_score_lenient": 0.535445181825322,
194
+ "fixed_lenient": 4490,
195
+ "broken_lenient": 818,
196
+ "net_fixed_lenient": 3672,
197
+ "broken_per_100_correct": 0.27246413499298855,
198
+ "wer_clitic": 0.1762344761338426,
199
+ "fixed_clitic": 4406,
200
+ "broken_clitic": 806,
201
+ "net_fixed_clitic": 3600,
202
+ "broken_per_100_correct_clitic": 0.27552352888894965,
203
+ "fix_rate_recoverable": 0.07449980841954454,
204
+ "unrecoverable_targets": 966,
205
+ "entity_fix_rate": 0.055248618784530384,
206
+ "entity_keep_rate": 0.9806051202482544,
207
+ "entity_spans_whisper_wrong": 724,
208
+ "entity_spans_whisper_right": 1289,
209
+ "applied": 2229,
210
+ "judged_precision": 0.7725437415881561,
211
+ "est_real_fixes": 1722.0
212
+ },
213
+ "verifier@0.7": {
214
+ "n_examples": 1978,
215
+ "wer": 0.18586491265762117,
216
+ "wer_zwnj_normalized": 0.17412010334550784,
217
+ "exact_match": 0.006572295247724975,
218
+ "hallucination_rate": 0.0,
219
+ "rows_changed": 0.7639029322548029,
220
+ "fix_rate": 0.06300816192251657,
221
+ "preservation_rate": 0.9984559907225763,
222
+ "targeted_score": 0.5307320763225465,
223
+ "fix_rate_lenient": 0.06658141098158805,
224
+ "preservation_rate_lenient": 0.9984444895960669,
225
+ "targeted_score_lenient": 0.5325129502888275,
226
+ "fixed_lenient": 4061,
227
+ "broken_lenient": 467,
228
+ "net_fixed_lenient": 3594,
229
+ "broken_per_100_correct": 0.15555104039330764,
230
+ "wer_clitic": 0.17632882007181575,
231
+ "fixed_clitic": 4011,
232
+ "broken_clitic": 473,
233
+ "net_fixed_clitic": 3538,
234
+ "broken_per_100_correct_clitic": 0.1616906069038129,
235
+ "fix_rate_recoverable": 0.06733636530228064,
236
+ "unrecoverable_targets": 966,
237
+ "entity_fix_rate": 0.03729281767955801,
238
+ "entity_keep_rate": 0.9852598913886734,
239
+ "entity_spans_whisper_wrong": 724,
240
+ "entity_spans_whisper_right": 1289,
241
+ "applied": 1215,
242
+ "judged_precision": 1.0,
243
+ "est_real_fixes": 1215.0
244
+ },
245
+ "verifier@0.8": {
246
+ "n_examples": 1978,
247
+ "wer": 0.18587868471356878,
248
+ "wer_zwnj_normalized": 0.17410908570074976,
249
+ "exact_match": 0.006572295247724975,
250
+ "hallucination_rate": 0.0,
251
+ "rows_changed": 0.7320525783619818,
252
+ "fix_rate": 0.059890652170507416,
253
+ "preservation_rate": 0.9989773445045635,
254
+ "targeted_score": 0.5294339983375355,
255
+ "fix_rate_lenient": 0.06333513681897923,
256
+ "preservation_rate_lenient": 0.9989641033498433,
257
+ "targeted_score_lenient": 0.5311496200844112,
258
+ "fixed_lenient": 3863,
259
+ "broken_lenient": 311,
260
+ "net_fixed_lenient": 3552,
261
+ "broken_per_100_correct": 0.10358966501567168,
262
+ "wer_clitic": 0.17634025570066097,
263
+ "fixed_clitic": 3828,
264
+ "broken_clitic": 317,
265
+ "net_fixed_clitic": 3511,
266
+ "broken_per_100_correct_clitic": 0.10836347228014521,
267
+ "fix_rate_recoverable": 0.06403784963433122,
268
+ "unrecoverable_targets": 966,
269
+ "entity_fix_rate": 0.023480662983425413,
270
+ "entity_keep_rate": 0.9899146625290923,
271
+ "entity_spans_whisper_wrong": 724,
272
+ "entity_spans_whisper_right": 1289,
273
+ "applied": 692,
274
+ "judged_precision": 1.0,
275
+ "est_real_fixes": 692.0
276
+ },
277
+ "verifier@0.9": {
278
+ "n_examples": 1978,
279
+ "wer": 0.18608251114159327,
280
+ "wer_zwnj_normalized": 0.1743074033063952,
281
+ "exact_match": 0.006572295247724975,
282
+ "hallucination_rate": 0.0,
283
+ "rows_changed": 0.7087967644084934,
284
+ "fix_rate": 0.05729011639747466,
285
+ "preservation_rate": 0.999338281738247,
286
+ "targeted_score": 0.5283141990678608,
287
+ "fix_rate_lenient": 0.0606626989982457,
288
+ "preservation_rate_lenient": 0.9993304976634035,
289
+ "targeted_score_lenient": 0.5299965983308246,
290
+ "fixed_lenient": 3700,
291
+ "broken_lenient": 201,
292
+ "net_fixed_lenient": 3499,
293
+ "broken_per_100_correct": 0.06695023365964634,
294
+ "wer_clitic": 0.17650321341170552,
295
+ "fixed_clitic": 3675,
296
+ "broken_clitic": 211,
297
+ "net_fixed_clitic": 3464,
298
+ "broken_per_100_correct_clitic": 0.07212836798457616,
299
+ "fix_rate_recoverable": 0.06133906408782715,
300
+ "unrecoverable_targets": 966,
301
+ "entity_fix_rate": 0.012430939226519336,
302
+ "entity_keep_rate": 0.9930178432893716,
303
+ "entity_spans_whisper_wrong": 724,
304
+ "entity_spans_whisper_right": 1289,
305
+ "applied": 278,
306
+ "judged_precision": 1.0,
307
+ "est_real_fixes": 278.0
308
+ },
309
+ "oracle": {
310
+ "n_examples": 1978,
311
+ "wer": 0.17497397081425903,
312
+ "wer_zwnj_normalized": 0.16355142761131952,
313
+ "exact_match": 0.007077856420626896,
314
+ "hallucination_rate": 0.0,
315
+ "rows_changed": 0.8624873609706775,
316
+ "fix_rate": 0.11846537057634765,
317
+ "preservation_rate": 0.999381727886746,
318
+ "targeted_score": 0.5589235492315467,
319
+ "fix_rate_lenient": 0.12511271785286837,
320
+ "preservation_rate_lenient": 0.9994737245314317,
321
+ "targeted_score_lenient": 0.5622932211921501,
322
+ "fixed_lenient": 7631,
323
+ "broken_lenient": 158,
324
+ "net_fixed_lenient": 7473,
325
+ "broken_per_100_correct": 0.05262754685683642,
326
+ "wer_clitic": 0.16583948951352834,
327
+ "fixed_clitic": 7318,
328
+ "broken_clitic": 171,
329
+ "net_fixed_clitic": 7147,
330
+ "broken_per_100_correct_clitic": 0.05845474372209726,
331
+ "fix_rate_recoverable": 0.1265930331350892,
332
+ "unrecoverable_targets": 966,
333
+ "entity_fix_rate": 0.11464088397790055,
334
+ "entity_keep_rate": 0.9953452288595811,
335
+ "entity_spans_whisper_wrong": 724,
336
+ "entity_spans_whisper_right": 1289,
337
+ "applied": 3443,
338
+ "judged_precision": null,
339
+ "est_real_fixes": null
340
+ }
341
+ },
342
+ "entity": {
343
+ "whisper": {
344
+ "n_examples": 1010,
345
+ "wer": 0.18660667584675655,
346
+ "wer_zwnj_normalized": 0.17409613475517982,
347
+ "exact_match": 0.0049504950495049506,
348
+ "hallucination_rate": 0.0,
349
+ "rows_changed": 0.0,
350
+ "fix_rate": 0.0,
351
+ "preservation_rate": 1.0,
352
+ "targeted_score": 0.5,
353
+ "fix_rate_lenient": 0.0,
354
+ "preservation_rate_lenient": 1.0,
355
+ "targeted_score_lenient": 0.5,
356
+ "fixed_lenient": 0,
357
+ "broken_lenient": 0,
358
+ "net_fixed_lenient": 0,
359
+ "broken_per_100_correct": 0.0,
360
+ "wer_clitic": 0.1776189974667827,
361
+ "fixed_clitic": 0,
362
+ "broken_clitic": 0,
363
+ "net_fixed_clitic": 0,
364
+ "broken_per_100_correct_clitic": 0.0,
365
+ "fix_rate_recoverable": 0.0,
366
+ "unrecoverable_targets": 471,
367
+ "entity_fix_rate": 0.0,
368
+ "entity_keep_rate": 1.0,
369
+ "entity_spans_whisper_wrong": 724,
370
+ "entity_spans_whisper_right": 1289,
371
+ "applied": 0,
372
+ "judged_precision": null,
373
+ "est_real_fixes": null
374
+ },
375
+ "rules only": {
376
+ "n_examples": 1010,
377
+ "wer": 0.18000982056126644,
378
+ "wer_zwnj_normalized": 0.16776614254758168,
379
+ "exact_match": 0.0049504950495049506,
380
+ "hallucination_rate": 0.0,
381
+ "rows_changed": 0.7148514851485148,
382
+ "fix_rate": 0.0566740770892695,
383
+ "preservation_rate": 0.9996911693742963,
384
+ "targeted_score": 0.5281826232317829,
385
+ "fix_rate_lenient": 0.06019704433497537,
386
+ "preservation_rate_lenient": 0.9996856677315474,
387
+ "targeted_score_lenient": 0.5299413560332614,
388
+ "fixed_lenient": 1833,
389
+ "broken_lenient": 49,
390
+ "net_fixed_lenient": 1784,
391
+ "broken_per_100_correct": 0.03143322684525871,
392
+ "wer_clitic": 0.1703353048452631,
393
+ "fixed_clitic": 1824,
394
+ "broken_clitic": 61,
395
+ "net_fixed_clitic": 1763,
396
+ "broken_per_100_correct_clitic": 0.04016434459690801,
397
+ "fix_rate_recoverable": 0.06077587644684613,
398
+ "unrecoverable_targets": 471,
399
+ "entity_fix_rate": 0.009668508287292817,
400
+ "entity_keep_rate": 0.9953452288595811,
401
+ "entity_spans_whisper_wrong": 724,
402
+ "entity_spans_whisper_right": 1289,
403
+ "applied": 0,
404
+ "judged_precision": null,
405
+ "est_real_fixes": null
406
+ },
407
+ "accept all": {
408
+ "n_examples": 1010,
409
+ "wer": 0.17671139291852136,
410
+ "wer_zwnj_normalized": 0.16492671939881084,
411
+ "exact_match": 0.0039603960396039604,
412
+ "hallucination_rate": 0.0,
413
+ "rows_changed": 0.9574257425742574,
414
+ "fix_rate": 0.12624855183642797,
415
+ "preservation_rate": 0.989673475953032,
416
+ "targeted_score": 0.55796101389473,
417
+ "fix_rate_lenient": 0.1329064039408867,
418
+ "preservation_rate_lenient": 0.9897745788589097,
419
+ "targeted_score_lenient": 0.5613404913998983,
420
+ "fixed_lenient": 4047,
421
+ "broken_lenient": 1594,
422
+ "net_fixed_lenient": 2453,
423
+ "broken_per_100_correct": 1.0225421141090283,
424
+ "wer_clitic": 0.16794066617517447,
425
+ "fixed_clitic": 3847,
426
+ "broken_clitic": 1589,
427
+ "net_fixed_clitic": 2258,
428
+ "broken_per_100_correct_clitic": 1.0462482551555217,
429
+ "fix_rate_recoverable": 0.13439407585309718,
430
+ "unrecoverable_targets": 471,
431
+ "entity_fix_rate": 0.11187845303867404,
432
+ "entity_keep_rate": 0.9317300232738557,
433
+ "entity_spans_whisper_wrong": 724,
434
+ "entity_spans_whisper_right": 1289,
435
+ "applied": 4473,
436
+ "judged_precision": null,
437
+ "est_real_fixes": null
438
+ },
439
+ "verifier@0.3": {
440
+ "n_examples": 1010,
441
+ "wer": 0.17885163480321517,
442
+ "wer_zwnj_normalized": 0.1667253765438029,
443
+ "exact_match": 0.0049504950495049506,
444
+ "hallucination_rate": 0.0,
445
+ "rows_changed": 0.8782178217821782,
446
+ "fix_rate": 0.08447881767229233,
447
+ "preservation_rate": 0.9955348238700338,
448
+ "targeted_score": 0.540006820771163,
449
+ "fix_rate_lenient": 0.08939244663382595,
450
+ "preservation_rate_lenient": 0.9955736884646472,
451
+ "targeted_score_lenient": 0.5424830675492366,
452
+ "fixed_lenient": 2722,
453
+ "broken_lenient": 690,
454
+ "net_fixed_lenient": 2032,
455
+ "broken_per_100_correct": 0.44263115353527577,
456
+ "wer_clitic": 0.16931536615244758,
457
+ "fixed_clitic": 2668,
458
+ "broken_clitic": 696,
459
+ "net_fixed_clitic": 1972,
460
+ "broken_per_100_correct_clitic": 0.45826858753193395,
461
+ "fix_rate_recoverable": 0.09039661096100604,
462
+ "unrecoverable_targets": 471,
463
+ "entity_fix_rate": 0.07044198895027624,
464
+ "entity_keep_rate": 0.9650892164468581,
465
+ "entity_spans_whisper_wrong": 724,
466
+ "entity_spans_whisper_right": 1289,
467
+ "applied": 1979,
468
+ "judged_precision": null,
469
+ "est_real_fixes": null
470
+ },
471
+ "verifier@0.5": {
472
+ "n_examples": 1010,
473
+ "wer": 0.17940137274367268,
474
+ "wer_zwnj_normalized": 0.16722174186868202,
475
+ "exact_match": 0.0049504950495049506,
476
+ "hallucination_rate": 0.0,
477
+ "rows_changed": 0.8346534653465346,
478
+ "fix_rate": 0.07135923850079845,
479
+ "preservation_rate": 0.997355637767412,
480
+ "targeted_score": 0.5343574381341052,
481
+ "fix_rate_lenient": 0.07556650246305419,
482
+ "preservation_rate_lenient": 0.9973634579115508,
483
+ "targeted_score_lenient": 0.5364649801873025,
484
+ "fixed_lenient": 2301,
485
+ "broken_lenient": 411,
486
+ "net_fixed_lenient": 1890,
487
+ "broken_per_100_correct": 0.26365420884492513,
488
+ "wer_clitic": 0.16983642178899463,
489
+ "fixed_clitic": 2263,
490
+ "broken_clitic": 411,
491
+ "net_fixed_clitic": 1852,
492
+ "broken_per_100_correct_clitic": 0.2706155021201507,
493
+ "fix_rate_recoverable": 0.07642016077921211,
494
+ "unrecoverable_targets": 471,
495
+ "entity_fix_rate": 0.055248618784530384,
496
+ "entity_keep_rate": 0.9806051202482544,
497
+ "entity_spans_whisper_wrong": 724,
498
+ "entity_spans_whisper_right": 1289,
499
+ "applied": 1143,
500
+ "judged_precision": null,
501
+ "est_real_fixes": null
502
+ },
503
+ "verifier@0.7": {
504
+ "n_examples": 1010,
505
+ "wer": 0.17956149059040788,
506
+ "wer_zwnj_normalized": 0.1673391616229545,
507
+ "exact_match": 0.0049504950495049506,
508
+ "hallucination_rate": 0.0,
509
+ "rows_changed": 0.7881188118811882,
510
+ "fix_rate": 0.06443936499984344,
511
+ "preservation_rate": 0.9985394884992762,
512
+ "targeted_score": 0.5314894267495598,
513
+ "fix_rate_lenient": 0.06824302134646962,
514
+ "preservation_rate_lenient": 0.9985438076543115,
515
+ "targeted_score_lenient": 0.5333934145003906,
516
+ "fixed_lenient": 2078,
517
+ "broken_lenient": 227,
518
+ "net_fixed_lenient": 1851,
519
+ "broken_per_100_correct": 0.1456192345688516,
520
+ "wer_clitic": 0.16994728469038764,
521
+ "fixed_clitic": 2055,
522
+ "broken_clitic": 231,
523
+ "net_fixed_clitic": 1824,
524
+ "broken_per_100_correct_clitic": 0.15209776396534014,
525
+ "fix_rate_recoverable": 0.06894826378464926,
526
+ "unrecoverable_targets": 471,
527
+ "entity_fix_rate": 0.03729281767955801,
528
+ "entity_keep_rate": 0.9852598913886734,
529
+ "entity_spans_whisper_wrong": 724,
530
+ "entity_spans_whisper_right": 1289,
531
+ "applied": 627,
532
+ "judged_precision": null,
533
+ "est_real_fixes": null
534
+ },
535
+ "verifier@0.8": {
536
+ "n_examples": 1010,
537
+ "wer": 0.17964154951377548,
538
+ "wer_zwnj_normalized": 0.16740320876164858,
539
+ "exact_match": 0.0049504950495049506,
540
+ "hallucination_rate": 0.0,
541
+ "rows_changed": 0.7534653465346535,
542
+ "fix_rate": 0.061182953940570495,
543
+ "preservation_rate": 0.9990542062087824,
544
+ "targeted_score": 0.5301185800746764,
545
+ "fix_rate_lenient": 0.06482758620689655,
546
+ "preservation_rate_lenient": 0.9990505882503881,
547
+ "targeted_score_lenient": 0.5319390872286422,
548
+ "fixed_lenient": 1974,
549
+ "broken_lenient": 148,
550
+ "net_fixed_lenient": 1826,
551
+ "broken_per_100_correct": 0.09494117496118959,
552
+ "wer_clitic": 0.17000271614108411,
553
+ "fixed_clitic": 1959,
554
+ "broken_clitic": 153,
555
+ "net_fixed_clitic": 1806,
556
+ "broken_per_100_correct_clitic": 0.10074007743158893,
557
+ "fix_rate_recoverable": 0.06547916875145936,
558
+ "unrecoverable_targets": 471,
559
+ "entity_fix_rate": 0.023480662983425413,
560
+ "entity_keep_rate": 0.9899146625290923,
561
+ "entity_spans_whisper_wrong": 724,
562
+ "entity_spans_whisper_right": 1289,
563
+ "applied": 357,
564
+ "judged_precision": null,
565
+ "est_real_fixes": null
566
+ },
567
+ "verifier@0.9": {
568
+ "n_examples": 1010,
569
+ "wer": 0.17986571449920474,
570
+ "wer_zwnj_normalized": 0.16762737374707784,
571
+ "exact_match": 0.0049504950495049506,
572
+ "hallucination_rate": 0.0,
573
+ "rows_changed": 0.7356435643564356,
574
+ "fix_rate": 0.05827097097410527,
575
+ "preservation_rate": 0.9994209425768056,
576
+ "targeted_score": 0.5288459567754554,
577
+ "fix_rate_lenient": 0.061839080459770115,
578
+ "preservation_rate_lenient": 0.9994226550171279,
579
+ "targeted_score_lenient": 0.530630867738449,
580
+ "fixed_lenient": 1883,
581
+ "broken_lenient": 90,
582
+ "net_fixed_lenient": 1793,
583
+ "broken_per_100_correct": 0.05773449828720988,
584
+ "wer_clitic": 0.17016901049317362,
585
+ "fixed_clitic": 1874,
586
+ "broken_clitic": 98,
587
+ "net_fixed_clitic": 1776,
588
+ "broken_per_100_correct_clitic": 0.06452632410650794,
589
+ "fix_rate_recoverable": 0.06244371059741819,
590
+ "unrecoverable_targets": 471,
591
+ "entity_fix_rate": 0.012430939226519336,
592
+ "entity_keep_rate": 0.9930178432893716,
593
+ "entity_spans_whisper_wrong": 724,
594
+ "entity_spans_whisper_right": 1289,
595
+ "applied": 139,
596
+ "judged_precision": null,
597
+ "est_real_fixes": null
598
+ },
599
+ "oracle": {
600
+ "n_examples": 1010,
601
+ "wer": 0.16844397476542736,
602
+ "wer_zwnj_normalized": 0.15656856779923356,
603
+ "exact_match": 0.0049504950495049506,
604
+ "hallucination_rate": 0.0,
605
+ "rows_changed": 0.897029702970297,
606
+ "fix_rate": 0.12364968531796976,
607
+ "preservation_rate": 0.9993759047772237,
608
+ "targeted_score": 0.5615127950475968,
609
+ "fix_rate_lenient": 0.13054187192118227,
610
+ "preservation_rate_lenient": 0.9994932194039233,
611
+ "targeted_score_lenient": 0.5650175456625528,
612
+ "fixed_lenient": 3975,
613
+ "broken_lenient": 79,
614
+ "net_fixed_lenient": 3896,
615
+ "broken_per_100_correct": 0.05067805960766201,
616
+ "wer_clitic": 0.1593321618820086,
617
+ "fixed_clitic": 3796,
618
+ "broken_clitic": 91,
619
+ "net_fixed_clitic": 3705,
620
+ "broken_per_100_correct_clitic": 0.059917300956043085,
621
+ "fix_rate_recoverable": 0.13195903799326195,
622
+ "unrecoverable_targets": 471,
623
+ "entity_fix_rate": 0.11464088397790055,
624
+ "entity_keep_rate": 0.9953452288595811,
625
+ "entity_spans_whisper_wrong": 724,
626
+ "entity_spans_whisper_right": 1289,
627
+ "applied": 1804,
628
+ "judged_precision": null,
629
+ "est_real_fixes": null
630
+ }
631
+ },
632
+ "typo": {
633
+ "whisper": {
634
+ "n_examples": 358,
635
+ "wer": 0.14117586990375533,
636
+ "wer_zwnj_normalized": 0.12749227745015446,
637
+ "exact_match": 0.0,
638
+ "hallucination_rate": 0.0,
639
+ "rows_changed": 0.0,
640
+ "fix_rate": 0.0,
641
+ "preservation_rate": 1.0,
642
+ "targeted_score": 0.5,
643
+ "fix_rate_lenient": 0.0,
644
+ "preservation_rate_lenient": 1.0,
645
+ "targeted_score_lenient": 0.5,
646
+ "fixed_lenient": 0,
647
+ "broken_lenient": 0,
648
+ "net_fixed_lenient": 0,
649
+ "broken_per_100_correct": 0.0,
650
+ "wer_clitic": 0.1316430880093255,
651
+ "fixed_clitic": 0,
652
+ "broken_clitic": 0,
653
+ "net_fixed_clitic": 0,
654
+ "broken_per_100_correct_clitic": 0.0,
655
+ "fix_rate_recoverable": 0.0,
656
+ "unrecoverable_targets": 68,
657
+ "entity_fix_rate": 0.0,
658
+ "entity_keep_rate": 1.0,
659
+ "entity_spans_whisper_wrong": 90,
660
+ "entity_spans_whisper_right": 215,
661
+ "applied": 0,
662
+ "judged_precision": null,
663
+ "est_real_fixes": null
664
+ },
665
+ "rules only": {
666
+ "n_examples": 358,
667
+ "wer": 0.13481912639452656,
668
+ "wer_zwnj_normalized": 0.12139082484491077,
669
+ "exact_match": 0.002793296089385475,
670
+ "hallucination_rate": 0.0,
671
+ "rows_changed": 0.4776536312849162,
672
+ "fix_rate": 0.06630476929042264,
673
+ "preservation_rate": 0.9998823979066828,
674
+ "targeted_score": 0.5330935835985527,
675
+ "fix_rate_lenient": 0.07141378596563859,
676
+ "preservation_rate_lenient": 0.9998827976207917,
677
+ "targeted_score_lenient": 0.5356482917932152,
678
+ "fixed_lenient": 345,
679
+ "broken_lenient": 4,
680
+ "net_fixed_lenient": 341,
681
+ "broken_per_100_correct": 0.011720237920829792,
682
+ "wer_clitic": 0.12475494092089227,
683
+ "fixed_clitic": 342,
684
+ "broken_clitic": 7,
685
+ "net_fixed_clitic": 335,
686
+ "broken_per_100_correct_clitic": 0.021002100210021003,
687
+ "fix_rate_recoverable": 0.07222338862061725,
688
+ "unrecoverable_targets": 68,
689
+ "entity_fix_rate": 0.011111111111111112,
690
+ "entity_keep_rate": 1.0,
691
+ "entity_spans_whisper_wrong": 90,
692
+ "entity_spans_whisper_right": 215,
693
+ "applied": 0,
694
+ "judged_precision": null,
695
+ "est_real_fixes": null
696
+ },
697
+ "accept all": {
698
+ "n_examples": 358,
699
+ "wer": 0.13425748640575935,
700
+ "wer_zwnj_normalized": 0.12116106303132419,
701
+ "exact_match": 0.00558659217877095,
702
+ "hallucination_rate": 0.0,
703
+ "rows_changed": 0.8519553072625698,
704
+ "fix_rate": 0.13900736719658782,
705
+ "preservation_rate": 0.9904448299179726,
706
+ "targeted_score": 0.5647260985572802,
707
+ "fix_rate_lenient": 0.14986545228731113,
708
+ "preservation_rate_lenient": 0.9904480060945238,
709
+ "targeted_score_lenient": 0.5701567291909174,
710
+ "fixed_lenient": 724,
711
+ "broken_lenient": 326,
712
+ "net_fixed_lenient": 398,
713
+ "broken_per_100_correct": 0.9551993905476281,
714
+ "wer_clitic": 0.12475494092089227,
715
+ "fixed_clitic": 688,
716
+ "broken_clitic": 322,
717
+ "net_fixed_clitic": 366,
718
+ "broken_per_100_correct_clitic": 0.9660966096609661,
719
+ "fix_rate_recoverable": 0.15179508712996012,
720
+ "unrecoverable_targets": 68,
721
+ "entity_fix_rate": 0.16666666666666666,
722
+ "entity_keep_rate": 0.9441860465116279,
723
+ "entity_spans_whisper_wrong": 90,
724
+ "entity_spans_whisper_right": 215,
725
+ "applied": 811,
726
+ "judged_precision": null,
727
+ "est_real_fixes": null
728
+ },
729
+ "verifier@0.3": {
730
+ "n_examples": 358,
731
+ "wer": 0.13443619003854893,
732
+ "wer_zwnj_normalized": 0.1209057721273391,
733
+ "exact_match": 0.002793296089385475,
734
+ "hallucination_rate": 0.0,
735
+ "rows_changed": 0.6871508379888268,
736
+ "fix_rate": 0.0934470725087243,
737
+ "preservation_rate": 0.996648340340458,
738
+ "targeted_score": 0.5450477064245911,
739
+ "fix_rate_lenient": 0.10060028979507349,
740
+ "preservation_rate_lenient": 0.9966890327873655,
741
+ "targeted_score_lenient": 0.5486446612912195,
742
+ "fixed_lenient": 486,
743
+ "broken_lenient": 113,
744
+ "net_fixed_lenient": 373,
745
+ "broken_per_100_correct": 0.33109672126344164,
746
+ "wer_clitic": 0.12456949080697292,
747
+ "fixed_clitic": 473,
748
+ "broken_clitic": 116,
749
+ "net_fixed_clitic": 357,
750
+ "broken_per_100_correct_clitic": 0.34803480348034804,
751
+ "fix_rate_recoverable": 0.10203653159773252,
752
+ "unrecoverable_targets": 68,
753
+ "entity_fix_rate": 0.1111111111111111,
754
+ "entity_keep_rate": 0.9813953488372092,
755
+ "entity_spans_whisper_wrong": 90,
756
+ "entity_spans_whisper_right": 215,
757
+ "applied": 334,
758
+ "judged_precision": null,
759
+ "est_real_fixes": null
760
+ },
761
+ "verifier@0.5": {
762
+ "n_examples": 358,
763
+ "wer": 0.1346659518521355,
764
+ "wer_zwnj_normalized": 0.12116106303132419,
765
+ "exact_match": 0.002793296089385475,
766
+ "hallucination_rate": 0.0,
767
+ "rows_changed": 0.6229050279329609,
768
+ "fix_rate": 0.0794881737107406,
769
+ "preservation_rate": 0.9983535706935583,
770
+ "targeted_score": 0.5389208722021495,
771
+ "fix_rate_lenient": 0.08548954667770647,
772
+ "preservation_rate_lenient": 0.9983298660962817,
773
+ "targeted_score_lenient": 0.5419097063869941,
774
+ "fixed_lenient": 413,
775
+ "broken_lenient": 57,
776
+ "net_fixed_lenient": 356,
777
+ "broken_per_100_correct": 0.16701339037182455,
778
+ "wer_clitic": 0.12475494092089227,
779
+ "fixed_clitic": 404,
780
+ "broken_clitic": 59,
781
+ "net_fixed_clitic": 345,
782
+ "broken_per_100_correct_clitic": 0.177017701770177,
783
+ "fix_rate_recoverable": 0.086710056686962,
784
+ "unrecoverable_targets": 68,
785
+ "entity_fix_rate": 0.08888888888888889,
786
+ "entity_keep_rate": 0.9906976744186047,
787
+ "entity_spans_whisper_wrong": 90,
788
+ "entity_spans_whisper_right": 215,
789
+ "applied": 185,
790
+ "judged_precision": null,
791
+ "est_real_fixes": null
792
+ },
793
+ "verifier@0.7": {
794
+ "n_examples": 358,
795
+ "wer": 0.13471701003293252,
796
+ "wer_zwnj_normalized": 0.12126317939291822,
797
+ "exact_match": 0.002793296089385475,
798
+ "hallucination_rate": 0.0,
799
+ "rows_changed": 0.5614525139664804,
800
+ "fix_rate": 0.07270259790616518,
801
+ "preservation_rate": 0.9990885837767912,
802
+ "targeted_score": 0.5358955908414782,
803
+ "fix_rate_lenient": 0.07824466984061271,
804
+ "preservation_rate_lenient": 0.9990623809663336,
805
+ "targeted_score_lenient": 0.5386535254034731,
806
+ "fixed_lenient": 378,
807
+ "broken_lenient": 32,
808
+ "net_fixed_lenient": 346,
809
+ "broken_per_100_correct": 0.09376190336663834,
810
+ "wer_clitic": 0.12480792666772639,
811
+ "fixed_clitic": 370,
812
+ "broken_clitic": 36,
813
+ "net_fixed_clitic": 334,
814
+ "broken_per_100_correct_clitic": 0.10801080108010801,
815
+ "fix_rate_recoverable": 0.07915179508712997,
816
+ "unrecoverable_targets": 68,
817
+ "entity_fix_rate": 0.044444444444444446,
818
+ "entity_keep_rate": 0.9906976744186047,
819
+ "entity_spans_whisper_wrong": 90,
820
+ "entity_spans_whisper_right": 215,
821
+ "applied": 96,
822
+ "judged_precision": null,
823
+ "est_real_fixes": null
824
+ },
825
+ "verifier@0.8": {
826
+ "n_examples": 358,
827
+ "wer": 0.1346659518521355,
828
+ "wer_zwnj_normalized": 0.1211865921217227,
829
+ "exact_match": 0.002793296089385475,
830
+ "hallucination_rate": 0.0,
831
+ "rows_changed": 0.5167597765363129,
832
+ "fix_rate": 0.069988367584335,
833
+ "preservation_rate": 0.9994707905800724,
834
+ "targeted_score": 0.5347295790822036,
835
+ "fix_rate_lenient": 0.07513972262471538,
836
+ "preservation_rate_lenient": 0.9994432886987605,
837
+ "targeted_score_lenient": 0.537291505661738,
838
+ "fixed_lenient": 363,
839
+ "broken_lenient": 19,
840
+ "net_fixed_lenient": 344,
841
+ "broken_per_100_correct": 0.055671130123941516,
842
+ "wer_clitic": 0.12475494092089227,
843
+ "fixed_clitic": 357,
844
+ "broken_clitic": 22,
845
+ "net_fixed_clitic": 335,
846
+ "broken_per_100_correct_clitic": 0.066006600660066,
847
+ "fix_rate_recoverable": 0.07600251942053328,
848
+ "unrecoverable_targets": 68,
849
+ "entity_fix_rate": 0.03333333333333333,
850
+ "entity_keep_rate": 0.9953488372093023,
851
+ "entity_spans_whisper_wrong": 90,
852
+ "entity_spans_whisper_right": 215,
853
+ "applied": 52,
854
+ "judged_precision": null,
855
+ "est_real_fixes": null
856
+ },
857
+ "verifier@0.9": {
858
+ "n_examples": 358,
859
+ "wer": 0.1346659518521355,
860
+ "wer_zwnj_normalized": 0.12123765030251972,
861
+ "exact_match": 0.002793296089385475,
862
+ "hallucination_rate": 0.0,
863
+ "rows_changed": 0.4972067039106145,
864
+ "fix_rate": 0.06746801085692129,
865
+ "preservation_rate": 0.9997941963366948,
866
+ "targeted_score": 0.533631103596808,
867
+ "fix_rate_lenient": 0.07265576485199751,
868
+ "preservation_rate_lenient": 0.9997948958363855,
869
+ "targeted_score_lenient": 0.5362253303441915,
870
+ "fixed_lenient": 351,
871
+ "broken_lenient": 7,
872
+ "net_fixed_lenient": 344,
873
+ "broken_per_100_correct": 0.02051041636145214,
874
+ "wer_clitic": 0.12456949080697292,
875
+ "fixed_clitic": 349,
876
+ "broken_clitic": 10,
877
+ "net_fixed_clitic": 339,
878
+ "broken_per_100_correct_clitic": 0.030003000300030003,
879
+ "fix_rate_recoverable": 0.07348309888725593,
880
+ "unrecoverable_targets": 68,
881
+ "entity_fix_rate": 0.011111111111111112,
882
+ "entity_keep_rate": 1.0,
883
+ "entity_spans_whisper_wrong": 90,
884
+ "entity_spans_whisper_right": 215,
885
+ "applied": 21,
886
+ "judged_precision": null,
887
+ "est_real_fixes": null
888
+ },
889
+ "oracle": {
890
+ "n_examples": 358,
891
+ "wer": 0.1254499502182737,
892
+ "wer_zwnj_normalized": 0.11240458502463557,
893
+ "exact_match": 0.00558659217877095,
894
+ "hallucination_rate": 0.0,
895
+ "rows_changed": 0.7262569832402235,
896
+ "fix_rate": 0.1355176424970919,
897
+ "preservation_rate": 0.9997647958133655,
898
+ "targeted_score": 0.5676412191552287,
899
+ "fix_rate_lenient": 0.14655350859035396,
900
+ "preservation_rate_lenient": 0.9998534970259896,
901
+ "targeted_score_lenient": 0.5732035028081718,
902
+ "fixed_lenient": 708,
903
+ "broken_lenient": 5,
904
+ "net_fixed_lenient": 703,
905
+ "broken_per_100_correct": 0.014650297401037241,
906
+ "wer_clitic": 0.11572087108567795,
907
+ "fixed_clitic": 673,
908
+ "broken_clitic": 8,
909
+ "net_fixed_clitic": 665,
910
+ "broken_per_100_correct_clitic": 0.024002400240024,
911
+ "fix_rate_recoverable": 0.14822590804115054,
912
+ "unrecoverable_targets": 68,
913
+ "entity_fix_rate": 0.16666666666666666,
914
+ "entity_keep_rate": 1.0,
915
+ "entity_spans_whisper_wrong": 90,
916
+ "entity_spans_whisper_right": 215,
917
+ "applied": 313,
918
+ "judged_precision": null,
919
+ "est_real_fixes": null
920
+ }
921
+ }
922
+ }
923
+ }
verifier-qwen3.5-2b-v2/report-20260929-1740/report.md ADDED
@@ -0,0 +1,44 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Verifier report: verifier-qwen3.5-2b-v2
2
+
3
+ 8781 candidates after dropping 0 clitic/style respellings. Judge cost $0.02. Judged precision: share of applied changes the calibrated judges call real errors (undecided excluded), estimated from a score-stratified sample.
4
+
5
+ | slice | method | WER | WER (clitic) | net | net (clitic) | broken/100 | broken/100 (clitic) | entity fix | applied | judged precision | est. real fixes |
6
+ |---|---|---|---|---|---|---|---|---|---|---|---|
7
+ | main | whisper | 0.1929 | 0.1840 | 0 | 0 | 0.00 | 0.00 | 0.000 | 0 | – | – |
8
+ | main | rules only | 0.1862 | 0.1767 | 3471 | 3434 | 0.04 | 0.05 | 0.010 | 0 | – | – |
9
+ | main | accept all | 0.1833 | 0.1746 | 4575 | 4266 | 1.06 | 1.08 | 0.112 | 8781 | 0.29 | 2522 |
10
+ | main | verifier@0.3 | 0.1851 | 0.1757 | 3935 | 3835 | 0.45 | 0.46 | 0.070 | 3830 | 0.66 | 2522 |
11
+ | main | verifier@0.5 | 0.1857 | 0.1762 | 3672 | 3600 | 0.27 | 0.28 | 0.055 | 2229 | 0.77 | 1722 |
12
+ | main | verifier@0.7 | 0.1859 | 0.1763 | 3594 | 3538 | 0.16 | 0.16 | 0.037 | 1215 | 1.00 | 1215 |
13
+ | main | verifier@0.8 | 0.1859 | 0.1763 | 3552 | 3511 | 0.10 | 0.11 | 0.023 | 692 | 1.00 | 692 |
14
+ | main | verifier@0.9 | 0.1861 | 0.1765 | 3499 | 3464 | 0.07 | 0.07 | 0.012 | 278 | 1.00 | 278 |
15
+ | main | oracle | 0.1750 | 0.1658 | 7473 | 7147 | 0.05 | 0.06 | 0.115 | 3443 | – | – |
16
+ | entity | whisper | 0.1866 | 0.1776 | 0 | 0 | 0.00 | 0.00 | 0.000 | 0 | – | – |
17
+ | entity | rules only | 0.1800 | 0.1703 | 1784 | 1763 | 0.03 | 0.04 | 0.010 | 0 | – | – |
18
+ | entity | accept all | 0.1767 | 0.1679 | 2453 | 2258 | 1.02 | 1.05 | 0.112 | 4473 | – | – |
19
+ | entity | verifier@0.3 | 0.1789 | 0.1693 | 2032 | 1972 | 0.44 | 0.46 | 0.070 | 1979 | – | – |
20
+ | entity | verifier@0.5 | 0.1794 | 0.1698 | 1890 | 1852 | 0.26 | 0.27 | 0.055 | 1143 | – | – |
21
+ | entity | verifier@0.7 | 0.1796 | 0.1699 | 1851 | 1824 | 0.15 | 0.15 | 0.037 | 627 | – | – |
22
+ | entity | verifier@0.8 | 0.1796 | 0.1700 | 1826 | 1806 | 0.09 | 0.10 | 0.023 | 357 | – | – |
23
+ | entity | verifier@0.9 | 0.1799 | 0.1702 | 1793 | 1776 | 0.06 | 0.06 | 0.012 | 139 | – | – |
24
+ | entity | oracle | 0.1684 | 0.1593 | 3896 | 3705 | 0.05 | 0.06 | 0.115 | 1804 | – | – |
25
+ | typo | whisper | 0.1412 | 0.1316 | 0 | 0 | 0.00 | 0.00 | 0.000 | 0 | – | – |
26
+ | typo | rules only | 0.1348 | 0.1248 | 341 | 335 | 0.01 | 0.02 | 0.011 | 0 | – | – |
27
+ | typo | accept all | 0.1343 | 0.1248 | 398 | 366 | 0.96 | 0.97 | 0.167 | 811 | – | – |
28
+ | typo | verifier@0.3 | 0.1344 | 0.1246 | 373 | 357 | 0.33 | 0.35 | 0.111 | 334 | – | – |
29
+ | typo | verifier@0.5 | 0.1347 | 0.1248 | 356 | 345 | 0.17 | 0.18 | 0.089 | 185 | – | – |
30
+ | typo | verifier@0.7 | 0.1347 | 0.1248 | 346 | 334 | 0.09 | 0.11 | 0.044 | 96 | – | – |
31
+ | typo | verifier@0.8 | 0.1347 | 0.1248 | 344 | 335 | 0.06 | 0.07 | 0.033 | 52 | – | – |
32
+ | typo | verifier@0.9 | 0.1347 | 0.1246 | 344 | 339 | 0.02 | 0.03 | 0.011 | 21 | – | – |
33
+ | typo | oracle | 0.1254 | 0.1157 | 703 | 665 | 0.01 | 0.02 | 0.167 | 313 | – | – |
34
+
35
+ Score bins (main set):
36
+
37
+ | bin | candidates | sampled | real error | not an error | undecided | precision |
38
+ |---|---|---|---|---|---|---|
39
+ | >= 0.0 | 4951 | 60 | 0 | 3 | 57 | 0.00 |
40
+ | >= 0.3 | 1601 | 60 | 1 | 1 | 58 | 0.50 |
41
+ | >= 0.5 | 1014 | 60 | 1 | 1 | 58 | 0.50 |
42
+ | >= 0.7 | 523 | 60 | 1 | 0 | 59 | 1.00 |
43
+ | >= 0.8 | 414 | 60 | 1 | 0 | 59 | 1.00 |
44
+ | >= 0.9 | 278 | 60 | 2 | 0 | 58 | 1.00 |