kiokuko 1.0.45 → 1.0.46
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/mcp/chatgpt-contract.d.ts +138 -0
- package/dist/mcp/chatgpt-contract.d.ts.map +1 -1
- package/dist/mcp/chatgpt-contract.js +2 -2
- package/dist/mcp/chatgpt-contract.js.map +1 -1
- package/dist/mcp/chatgpt-server.d.ts +4 -1
- package/dist/mcp/chatgpt-server.d.ts.map +1 -1
- package/dist/mcp/chatgpt-server.js +3 -3
- package/dist/mcp/chatgpt-server.js.map +1 -1
- package/dist/mcp/correction-review.d.ts +46 -4
- package/dist/mcp/correction-review.d.ts.map +1 -1
- package/dist/mcp/correction-review.js +24 -5
- package/dist/mcp/correction-review.js.map +1 -1
- package/dist/memory/correction-classification.d.ts +47 -0
- package/dist/memory/correction-classification.d.ts.map +1 -0
- package/dist/memory/correction-classification.js +143 -0
- package/dist/memory/correction-classification.js.map +1 -0
- package/dist/memory/correction-contract.d.ts +119 -10
- package/dist/memory/correction-contract.d.ts.map +1 -1
- package/dist/memory/correction-contract.js +35 -7
- package/dist/memory/correction-contract.js.map +1 -1
- package/docs/correction-classification.md +122 -0
- package/docs/evaluations/correction-classification-2026-10-21/assessment.json +717 -0
- package/docs/evaluations/correction-classification-2026-10-21/built-both-product-pair.log +14 -0
- package/docs/evaluations/correction-classification-2026-10-21/development.json +877 -0
- package/docs/evaluations/correction-classification-2026-10-21/dsh-build.log +9 -0
- package/docs/evaluations/correction-classification-2026-10-21/dsh-final-2-typecheck.log +0 -0
- package/docs/evaluations/correction-classification-2026-10-21/dsh-final-2-unit.log +43 -0
- package/docs/evaluations/correction-classification-2026-10-21/dsh-final-3-typecheck.log +0 -0
- package/docs/evaluations/correction-classification-2026-10-21/held-out.json +2671 -0
- package/docs/evaluations/correction-classification-2026-10-21/live-mcp-final-2.json +32 -0
- package/docs/evaluations/correction-classification-2026-10-21/live-mcp-final-2.log +1 -0
- package/docs/evaluations/correction-classification-2026-10-21/live-mcp-final.json +32 -0
- package/docs/evaluations/correction-classification-2026-10-21/main-build.log +17 -0
- package/docs/evaluations/correction-classification-2026-10-21/main-final-2-typecheck.log +0 -0
- package/docs/evaluations/correction-classification-2026-10-21/main-final-2-unit.log +64 -0
- package/docs/evaluations/correction-classification-2026-10-21/metrics-final.log +11 -0
- package/docs/evaluations/correction-classification-2026-10-21/native-final.log +57 -0
- package/docs/evaluations/correction-classification-2026-10-21/package-evidence.json +51 -0
- package/docs/evaluations/correction-classification-2026-10-21/timing-final.json +53 -0
- package/docs/evaluations/correction-classification-2026-10-21/verification.json +199 -0
- package/migrations/015_correction_classification.sql +11 -0
- package/package.json +1 -1
|
@@ -0,0 +1,717 @@
|
|
|
1
|
+
{
|
|
2
|
+
"schema": "correction-assessment-v1",
|
|
3
|
+
"originalLiveReportSha256": "11ba14264c4945d2e6dd15fc254cd81b4ed2bbc53b866c8bd90ea524d92e6c54",
|
|
4
|
+
"sourceSha": "01e27576d2efe5020c19e93bdba1a6528bfcaee1",
|
|
5
|
+
"model": "jev-1.13.0",
|
|
6
|
+
"policy": "correction-meaning-v1",
|
|
7
|
+
"corpusSha256": "55350662ade98aaeb475829cfc5158828d629da2ac06ae72b93d959afb20c80d",
|
|
8
|
+
"batchContractSha256": "fc16c03dae9fac288f713acc0e1e64cdff2bc338600e8cf2b3b873441f63054e",
|
|
9
|
+
"semantic": {
|
|
10
|
+
"overall": {
|
|
11
|
+
"attempted": 120,
|
|
12
|
+
"correct": 107,
|
|
13
|
+
"confusion": {
|
|
14
|
+
"capability_compatibility": {
|
|
15
|
+
"capability_compatibility": 45,
|
|
16
|
+
"reference": 0,
|
|
17
|
+
"uncertain": 0,
|
|
18
|
+
"provider_failure": 0
|
|
19
|
+
},
|
|
20
|
+
"reference": {
|
|
21
|
+
"capability_compatibility": 7,
|
|
22
|
+
"reference": 47,
|
|
23
|
+
"uncertain": 6,
|
|
24
|
+
"provider_failure": 0
|
|
25
|
+
},
|
|
26
|
+
"uncertain": {
|
|
27
|
+
"capability_compatibility": 0,
|
|
28
|
+
"reference": 0,
|
|
29
|
+
"uncertain": 15,
|
|
30
|
+
"provider_failure": 0
|
|
31
|
+
}
|
|
32
|
+
},
|
|
33
|
+
"accuracy": 0.8916666666666667,
|
|
34
|
+
"semanticOnlyAccuracy": 0.9292929292929293,
|
|
35
|
+
"coverage": 0.825,
|
|
36
|
+
"trueSpecial": 45,
|
|
37
|
+
"falseSpecial": 7,
|
|
38
|
+
"falseNonSpecial": 0,
|
|
39
|
+
"trueNonSpecial": 68,
|
|
40
|
+
"goldSpecial": 45,
|
|
41
|
+
"goldNonSpecial": 75,
|
|
42
|
+
"specialPrecision": 0.8653846153846154,
|
|
43
|
+
"specialRecall": 1,
|
|
44
|
+
"falseSpecialRate": 0.09333333333333334,
|
|
45
|
+
"falseSpecialDenominator": "all gold non-special cases, including gold uncertain",
|
|
46
|
+
"abstentionRate": 0.175,
|
|
47
|
+
"providerFailureRate": 0,
|
|
48
|
+
"providerFailures": 0,
|
|
49
|
+
"latencyMs": {
|
|
50
|
+
"mean": 165.68923816666666,
|
|
51
|
+
"p50": 157.47154099999898,
|
|
52
|
+
"p95": 210.27695900000072
|
|
53
|
+
},
|
|
54
|
+
"inputTokens": {
|
|
55
|
+
"reportedCases": 120,
|
|
56
|
+
"total": 71438,
|
|
57
|
+
"mean": 595.3166666666667
|
|
58
|
+
},
|
|
59
|
+
"denominatorZero": "null, never zero or success"
|
|
60
|
+
},
|
|
61
|
+
"languages": {
|
|
62
|
+
"ja": {
|
|
63
|
+
"attempted": 24,
|
|
64
|
+
"correct": 22,
|
|
65
|
+
"confusion": {
|
|
66
|
+
"capability_compatibility": {
|
|
67
|
+
"capability_compatibility": 9,
|
|
68
|
+
"reference": 0,
|
|
69
|
+
"uncertain": 0,
|
|
70
|
+
"provider_failure": 0
|
|
71
|
+
},
|
|
72
|
+
"reference": {
|
|
73
|
+
"capability_compatibility": 1,
|
|
74
|
+
"reference": 10,
|
|
75
|
+
"uncertain": 1,
|
|
76
|
+
"provider_failure": 0
|
|
77
|
+
},
|
|
78
|
+
"uncertain": {
|
|
79
|
+
"capability_compatibility": 0,
|
|
80
|
+
"reference": 0,
|
|
81
|
+
"uncertain": 3,
|
|
82
|
+
"provider_failure": 0
|
|
83
|
+
}
|
|
84
|
+
},
|
|
85
|
+
"accuracy": 0.9166666666666666,
|
|
86
|
+
"semanticOnlyAccuracy": 0.95,
|
|
87
|
+
"coverage": 0.8333333333333334,
|
|
88
|
+
"trueSpecial": 9,
|
|
89
|
+
"falseSpecial": 1,
|
|
90
|
+
"falseNonSpecial": 0,
|
|
91
|
+
"trueNonSpecial": 14,
|
|
92
|
+
"goldSpecial": 9,
|
|
93
|
+
"goldNonSpecial": 15,
|
|
94
|
+
"specialPrecision": 0.9,
|
|
95
|
+
"specialRecall": 1,
|
|
96
|
+
"falseSpecialRate": 0.06666666666666667,
|
|
97
|
+
"falseSpecialDenominator": "all gold non-special cases, including gold uncertain",
|
|
98
|
+
"abstentionRate": 0.16666666666666666,
|
|
99
|
+
"providerFailureRate": 0,
|
|
100
|
+
"providerFailures": 0,
|
|
101
|
+
"latencyMs": {
|
|
102
|
+
"mean": 181.37440104166672,
|
|
103
|
+
"p50": 166.34016699999984,
|
|
104
|
+
"p95": 290.30345799999986
|
|
105
|
+
},
|
|
106
|
+
"inputTokens": {
|
|
107
|
+
"reportedCases": 24,
|
|
108
|
+
"total": 14745,
|
|
109
|
+
"mean": 614.375
|
|
110
|
+
},
|
|
111
|
+
"denominatorZero": "null, never zero or success"
|
|
112
|
+
},
|
|
113
|
+
"en": {
|
|
114
|
+
"attempted": 24,
|
|
115
|
+
"correct": 22,
|
|
116
|
+
"confusion": {
|
|
117
|
+
"capability_compatibility": {
|
|
118
|
+
"capability_compatibility": 9,
|
|
119
|
+
"reference": 0,
|
|
120
|
+
"uncertain": 0,
|
|
121
|
+
"provider_failure": 0
|
|
122
|
+
},
|
|
123
|
+
"reference": {
|
|
124
|
+
"capability_compatibility": 1,
|
|
125
|
+
"reference": 10,
|
|
126
|
+
"uncertain": 1,
|
|
127
|
+
"provider_failure": 0
|
|
128
|
+
},
|
|
129
|
+
"uncertain": {
|
|
130
|
+
"capability_compatibility": 0,
|
|
131
|
+
"reference": 0,
|
|
132
|
+
"uncertain": 3,
|
|
133
|
+
"provider_failure": 0
|
|
134
|
+
}
|
|
135
|
+
},
|
|
136
|
+
"accuracy": 0.9166666666666666,
|
|
137
|
+
"semanticOnlyAccuracy": 0.95,
|
|
138
|
+
"coverage": 0.8333333333333334,
|
|
139
|
+
"trueSpecial": 9,
|
|
140
|
+
"falseSpecial": 1,
|
|
141
|
+
"falseNonSpecial": 0,
|
|
142
|
+
"trueNonSpecial": 14,
|
|
143
|
+
"goldSpecial": 9,
|
|
144
|
+
"goldNonSpecial": 15,
|
|
145
|
+
"specialPrecision": 0.9,
|
|
146
|
+
"specialRecall": 1,
|
|
147
|
+
"falseSpecialRate": 0.06666666666666667,
|
|
148
|
+
"falseSpecialDenominator": "all gold non-special cases, including gold uncertain",
|
|
149
|
+
"abstentionRate": 0.16666666666666666,
|
|
150
|
+
"providerFailureRate": 0,
|
|
151
|
+
"providerFailures": 0,
|
|
152
|
+
"latencyMs": {
|
|
153
|
+
"mean": 157.6844443749999,
|
|
154
|
+
"p50": 158.0545000000002,
|
|
155
|
+
"p95": 171.3648750000002
|
|
156
|
+
},
|
|
157
|
+
"inputTokens": {
|
|
158
|
+
"reportedCases": 24,
|
|
159
|
+
"total": 13941,
|
|
160
|
+
"mean": 580.875
|
|
161
|
+
},
|
|
162
|
+
"denominatorZero": "null, never zero or success"
|
|
163
|
+
},
|
|
164
|
+
"zh": {
|
|
165
|
+
"attempted": 24,
|
|
166
|
+
"correct": 20,
|
|
167
|
+
"confusion": {
|
|
168
|
+
"capability_compatibility": {
|
|
169
|
+
"capability_compatibility": 9,
|
|
170
|
+
"reference": 0,
|
|
171
|
+
"uncertain": 0,
|
|
172
|
+
"provider_failure": 0
|
|
173
|
+
},
|
|
174
|
+
"reference": {
|
|
175
|
+
"capability_compatibility": 2,
|
|
176
|
+
"reference": 8,
|
|
177
|
+
"uncertain": 2,
|
|
178
|
+
"provider_failure": 0
|
|
179
|
+
},
|
|
180
|
+
"uncertain": {
|
|
181
|
+
"capability_compatibility": 0,
|
|
182
|
+
"reference": 0,
|
|
183
|
+
"uncertain": 3,
|
|
184
|
+
"provider_failure": 0
|
|
185
|
+
}
|
|
186
|
+
},
|
|
187
|
+
"accuracy": 0.8333333333333334,
|
|
188
|
+
"semanticOnlyAccuracy": 0.8947368421052632,
|
|
189
|
+
"coverage": 0.7916666666666666,
|
|
190
|
+
"trueSpecial": 9,
|
|
191
|
+
"falseSpecial": 2,
|
|
192
|
+
"falseNonSpecial": 0,
|
|
193
|
+
"trueNonSpecial": 13,
|
|
194
|
+
"goldSpecial": 9,
|
|
195
|
+
"goldNonSpecial": 15,
|
|
196
|
+
"specialPrecision": 0.8181818181818182,
|
|
197
|
+
"specialRecall": 1,
|
|
198
|
+
"falseSpecialRate": 0.13333333333333333,
|
|
199
|
+
"falseSpecialDenominator": "all gold non-special cases, including gold uncertain",
|
|
200
|
+
"abstentionRate": 0.20833333333333334,
|
|
201
|
+
"providerFailureRate": 0,
|
|
202
|
+
"providerFailures": 0,
|
|
203
|
+
"latencyMs": {
|
|
204
|
+
"mean": 163.86785587500003,
|
|
205
|
+
"p50": 155.27112500000112,
|
|
206
|
+
"p95": 177.3324590000011
|
|
207
|
+
},
|
|
208
|
+
"inputTokens": {
|
|
209
|
+
"reportedCases": 24,
|
|
210
|
+
"total": 14267,
|
|
211
|
+
"mean": 594.4583333333334
|
|
212
|
+
},
|
|
213
|
+
"denominatorZero": "null, never zero or success"
|
|
214
|
+
},
|
|
215
|
+
"ko": {
|
|
216
|
+
"attempted": 24,
|
|
217
|
+
"correct": 22,
|
|
218
|
+
"confusion": {
|
|
219
|
+
"capability_compatibility": {
|
|
220
|
+
"capability_compatibility": 9,
|
|
221
|
+
"reference": 0,
|
|
222
|
+
"uncertain": 0,
|
|
223
|
+
"provider_failure": 0
|
|
224
|
+
},
|
|
225
|
+
"reference": {
|
|
226
|
+
"capability_compatibility": 2,
|
|
227
|
+
"reference": 10,
|
|
228
|
+
"uncertain": 0,
|
|
229
|
+
"provider_failure": 0
|
|
230
|
+
},
|
|
231
|
+
"uncertain": {
|
|
232
|
+
"capability_compatibility": 0,
|
|
233
|
+
"reference": 0,
|
|
234
|
+
"uncertain": 3,
|
|
235
|
+
"provider_failure": 0
|
|
236
|
+
}
|
|
237
|
+
},
|
|
238
|
+
"accuracy": 0.9166666666666666,
|
|
239
|
+
"semanticOnlyAccuracy": 0.9047619047619048,
|
|
240
|
+
"coverage": 0.875,
|
|
241
|
+
"trueSpecial": 9,
|
|
242
|
+
"falseSpecial": 2,
|
|
243
|
+
"falseNonSpecial": 0,
|
|
244
|
+
"trueNonSpecial": 13,
|
|
245
|
+
"goldSpecial": 9,
|
|
246
|
+
"goldNonSpecial": 15,
|
|
247
|
+
"specialPrecision": 0.8181818181818182,
|
|
248
|
+
"specialRecall": 1,
|
|
249
|
+
"falseSpecialRate": 0.13333333333333333,
|
|
250
|
+
"falseSpecialDenominator": "all gold non-special cases, including gold uncertain",
|
|
251
|
+
"abstentionRate": 0.125,
|
|
252
|
+
"providerFailureRate": 0,
|
|
253
|
+
"providerFailures": 0,
|
|
254
|
+
"latencyMs": {
|
|
255
|
+
"mean": 159.04671354166643,
|
|
256
|
+
"p50": 153.05062499999985,
|
|
257
|
+
"p95": 192.79079099999944
|
|
258
|
+
},
|
|
259
|
+
"inputTokens": {
|
|
260
|
+
"reportedCases": 24,
|
|
261
|
+
"total": 14452,
|
|
262
|
+
"mean": 602.1666666666666
|
|
263
|
+
},
|
|
264
|
+
"denominatorZero": "null, never zero or success"
|
|
265
|
+
},
|
|
266
|
+
"es": {
|
|
267
|
+
"attempted": 24,
|
|
268
|
+
"correct": 21,
|
|
269
|
+
"confusion": {
|
|
270
|
+
"capability_compatibility": {
|
|
271
|
+
"capability_compatibility": 9,
|
|
272
|
+
"reference": 0,
|
|
273
|
+
"uncertain": 0,
|
|
274
|
+
"provider_failure": 0
|
|
275
|
+
},
|
|
276
|
+
"reference": {
|
|
277
|
+
"capability_compatibility": 1,
|
|
278
|
+
"reference": 9,
|
|
279
|
+
"uncertain": 2,
|
|
280
|
+
"provider_failure": 0
|
|
281
|
+
},
|
|
282
|
+
"uncertain": {
|
|
283
|
+
"capability_compatibility": 0,
|
|
284
|
+
"reference": 0,
|
|
285
|
+
"uncertain": 3,
|
|
286
|
+
"provider_failure": 0
|
|
287
|
+
}
|
|
288
|
+
},
|
|
289
|
+
"accuracy": 0.875,
|
|
290
|
+
"semanticOnlyAccuracy": 0.9473684210526315,
|
|
291
|
+
"coverage": 0.7916666666666666,
|
|
292
|
+
"trueSpecial": 9,
|
|
293
|
+
"falseSpecial": 1,
|
|
294
|
+
"falseNonSpecial": 0,
|
|
295
|
+
"trueNonSpecial": 14,
|
|
296
|
+
"goldSpecial": 9,
|
|
297
|
+
"goldNonSpecial": 15,
|
|
298
|
+
"specialPrecision": 0.9,
|
|
299
|
+
"specialRecall": 1,
|
|
300
|
+
"falseSpecialRate": 0.06666666666666667,
|
|
301
|
+
"falseSpecialDenominator": "all gold non-special cases, including gold uncertain",
|
|
302
|
+
"abstentionRate": 0.20833333333333334,
|
|
303
|
+
"providerFailureRate": 0,
|
|
304
|
+
"providerFailures": 0,
|
|
305
|
+
"latencyMs": {
|
|
306
|
+
"mean": 166.47277600000038,
|
|
307
|
+
"p50": 155.445499999998,
|
|
308
|
+
"p95": 230.20254200000272
|
|
309
|
+
},
|
|
310
|
+
"inputTokens": {
|
|
311
|
+
"reportedCases": 24,
|
|
312
|
+
"total": 14033,
|
|
313
|
+
"mean": 584.7083333333334
|
|
314
|
+
},
|
|
315
|
+
"denominatorZero": "null, never zero or success"
|
|
316
|
+
}
|
|
317
|
+
},
|
|
318
|
+
"gates": {
|
|
319
|
+
"specialPrecision": 0.95,
|
|
320
|
+
"specialRecall": 0.9,
|
|
321
|
+
"falseSpecialRate": 0.02,
|
|
322
|
+
"noSharplyFailingLanguage": "each tested language meets the precision and recall gates",
|
|
323
|
+
"qualified": false
|
|
324
|
+
}
|
|
325
|
+
},
|
|
326
|
+
"baseline": {
|
|
327
|
+
"overall": {
|
|
328
|
+
"attempted": 120,
|
|
329
|
+
"correct": 52,
|
|
330
|
+
"confusion": {
|
|
331
|
+
"capability_compatibility": {
|
|
332
|
+
"capability_compatibility": 5,
|
|
333
|
+
"reference": 40,
|
|
334
|
+
"uncertain": 0,
|
|
335
|
+
"provider_failure": 0
|
|
336
|
+
},
|
|
337
|
+
"reference": {
|
|
338
|
+
"capability_compatibility": 13,
|
|
339
|
+
"reference": 47,
|
|
340
|
+
"uncertain": 0,
|
|
341
|
+
"provider_failure": 0
|
|
342
|
+
},
|
|
343
|
+
"uncertain": {
|
|
344
|
+
"capability_compatibility": 1,
|
|
345
|
+
"reference": 14,
|
|
346
|
+
"uncertain": 0,
|
|
347
|
+
"provider_failure": 0
|
|
348
|
+
}
|
|
349
|
+
},
|
|
350
|
+
"accuracy": 0.43333333333333335,
|
|
351
|
+
"semanticOnlyAccuracy": 0.43333333333333335,
|
|
352
|
+
"coverage": 1,
|
|
353
|
+
"trueSpecial": 5,
|
|
354
|
+
"falseSpecial": 14,
|
|
355
|
+
"falseNonSpecial": 40,
|
|
356
|
+
"trueNonSpecial": 61,
|
|
357
|
+
"goldSpecial": 45,
|
|
358
|
+
"goldNonSpecial": 75,
|
|
359
|
+
"specialPrecision": 0.2631578947368421,
|
|
360
|
+
"specialRecall": 0.1111111111111111,
|
|
361
|
+
"falseSpecialRate": 0.18666666666666668,
|
|
362
|
+
"falseSpecialDenominator": "all gold non-special cases, including gold uncertain",
|
|
363
|
+
"abstentionRate": 0,
|
|
364
|
+
"providerFailureRate": 0,
|
|
365
|
+
"providerFailures": 0,
|
|
366
|
+
"latencyMs": {
|
|
367
|
+
"mean": 0.0001807618249999986,
|
|
368
|
+
"p50": 0.0001452910000000429,
|
|
369
|
+
"p95": 0.00026858299999997823
|
|
370
|
+
},
|
|
371
|
+
"inputTokens": {
|
|
372
|
+
"reportedCases": 120,
|
|
373
|
+
"total": 0,
|
|
374
|
+
"mean": 0
|
|
375
|
+
},
|
|
376
|
+
"denominatorZero": "null, never zero or success"
|
|
377
|
+
},
|
|
378
|
+
"languages": {
|
|
379
|
+
"ja": {
|
|
380
|
+
"attempted": 24,
|
|
381
|
+
"correct": 10,
|
|
382
|
+
"confusion": {
|
|
383
|
+
"capability_compatibility": {
|
|
384
|
+
"capability_compatibility": 1,
|
|
385
|
+
"reference": 8,
|
|
386
|
+
"uncertain": 0,
|
|
387
|
+
"provider_failure": 0
|
|
388
|
+
},
|
|
389
|
+
"reference": {
|
|
390
|
+
"capability_compatibility": 3,
|
|
391
|
+
"reference": 9,
|
|
392
|
+
"uncertain": 0,
|
|
393
|
+
"provider_failure": 0
|
|
394
|
+
},
|
|
395
|
+
"uncertain": {
|
|
396
|
+
"capability_compatibility": 0,
|
|
397
|
+
"reference": 3,
|
|
398
|
+
"uncertain": 0,
|
|
399
|
+
"provider_failure": 0
|
|
400
|
+
}
|
|
401
|
+
},
|
|
402
|
+
"accuracy": 0.4166666666666667,
|
|
403
|
+
"semanticOnlyAccuracy": 0.4166666666666667,
|
|
404
|
+
"coverage": 1,
|
|
405
|
+
"trueSpecial": 1,
|
|
406
|
+
"falseSpecial": 3,
|
|
407
|
+
"falseNonSpecial": 8,
|
|
408
|
+
"trueNonSpecial": 12,
|
|
409
|
+
"goldSpecial": 9,
|
|
410
|
+
"goldNonSpecial": 15,
|
|
411
|
+
"specialPrecision": 0.25,
|
|
412
|
+
"specialRecall": 0.1111111111111111,
|
|
413
|
+
"falseSpecialRate": 0.2,
|
|
414
|
+
"falseSpecialDenominator": "all gold non-special cases, including gold uncertain",
|
|
415
|
+
"abstentionRate": 0,
|
|
416
|
+
"providerFailureRate": 0,
|
|
417
|
+
"providerFailures": 0,
|
|
418
|
+
"latencyMs": {
|
|
419
|
+
"mean": 0.00019138024999999694,
|
|
420
|
+
"p50": 0.00016729099999997744,
|
|
421
|
+
"p95": 0.00026858299999997823
|
|
422
|
+
},
|
|
423
|
+
"inputTokens": {
|
|
424
|
+
"reportedCases": 24,
|
|
425
|
+
"total": 0,
|
|
426
|
+
"mean": 0
|
|
427
|
+
},
|
|
428
|
+
"denominatorZero": "null, never zero or success"
|
|
429
|
+
},
|
|
430
|
+
"en": {
|
|
431
|
+
"attempted": 24,
|
|
432
|
+
"correct": 6,
|
|
433
|
+
"confusion": {
|
|
434
|
+
"capability_compatibility": {
|
|
435
|
+
"capability_compatibility": 2,
|
|
436
|
+
"reference": 7,
|
|
437
|
+
"uncertain": 0,
|
|
438
|
+
"provider_failure": 0
|
|
439
|
+
},
|
|
440
|
+
"reference": {
|
|
441
|
+
"capability_compatibility": 8,
|
|
442
|
+
"reference": 4,
|
|
443
|
+
"uncertain": 0,
|
|
444
|
+
"provider_failure": 0
|
|
445
|
+
},
|
|
446
|
+
"uncertain": {
|
|
447
|
+
"capability_compatibility": 1,
|
|
448
|
+
"reference": 2,
|
|
449
|
+
"uncertain": 0,
|
|
450
|
+
"provider_failure": 0
|
|
451
|
+
}
|
|
452
|
+
},
|
|
453
|
+
"accuracy": 0.25,
|
|
454
|
+
"semanticOnlyAccuracy": 0.25,
|
|
455
|
+
"coverage": 1,
|
|
456
|
+
"trueSpecial": 2,
|
|
457
|
+
"falseSpecial": 9,
|
|
458
|
+
"falseNonSpecial": 7,
|
|
459
|
+
"trueNonSpecial": 6,
|
|
460
|
+
"goldSpecial": 9,
|
|
461
|
+
"goldNonSpecial": 15,
|
|
462
|
+
"specialPrecision": 0.18181818181818182,
|
|
463
|
+
"specialRecall": 0.2222222222222222,
|
|
464
|
+
"falseSpecialRate": 0.6,
|
|
465
|
+
"falseSpecialDenominator": "all gold non-special cases, including gold uncertain",
|
|
466
|
+
"abstentionRate": 0,
|
|
467
|
+
"providerFailureRate": 0,
|
|
468
|
+
"providerFailures": 0,
|
|
469
|
+
"latencyMs": {
|
|
470
|
+
"mean": 0.00022948783333333016,
|
|
471
|
+
"p50": 0.00017116699999996854,
|
|
472
|
+
"p95": 0.00026358300000003964
|
|
473
|
+
},
|
|
474
|
+
"inputTokens": {
|
|
475
|
+
"reportedCases": 24,
|
|
476
|
+
"total": 0,
|
|
477
|
+
"mean": 0
|
|
478
|
+
},
|
|
479
|
+
"denominatorZero": "null, never zero or success"
|
|
480
|
+
},
|
|
481
|
+
"zh": {
|
|
482
|
+
"attempted": 24,
|
|
483
|
+
"correct": 11,
|
|
484
|
+
"confusion": {
|
|
485
|
+
"capability_compatibility": {
|
|
486
|
+
"capability_compatibility": 1,
|
|
487
|
+
"reference": 8,
|
|
488
|
+
"uncertain": 0,
|
|
489
|
+
"provider_failure": 0
|
|
490
|
+
},
|
|
491
|
+
"reference": {
|
|
492
|
+
"capability_compatibility": 2,
|
|
493
|
+
"reference": 10,
|
|
494
|
+
"uncertain": 0,
|
|
495
|
+
"provider_failure": 0
|
|
496
|
+
},
|
|
497
|
+
"uncertain": {
|
|
498
|
+
"capability_compatibility": 0,
|
|
499
|
+
"reference": 3,
|
|
500
|
+
"uncertain": 0,
|
|
501
|
+
"provider_failure": 0
|
|
502
|
+
}
|
|
503
|
+
},
|
|
504
|
+
"accuracy": 0.4583333333333333,
|
|
505
|
+
"semanticOnlyAccuracy": 0.4583333333333333,
|
|
506
|
+
"coverage": 1,
|
|
507
|
+
"trueSpecial": 1,
|
|
508
|
+
"falseSpecial": 2,
|
|
509
|
+
"falseNonSpecial": 8,
|
|
510
|
+
"trueNonSpecial": 13,
|
|
511
|
+
"goldSpecial": 9,
|
|
512
|
+
"goldNonSpecial": 15,
|
|
513
|
+
"specialPrecision": 0.3333333333333333,
|
|
514
|
+
"specialRecall": 0.1111111111111111,
|
|
515
|
+
"falseSpecialRate": 0.13333333333333333,
|
|
516
|
+
"falseSpecialDenominator": "all gold non-special cases, including gold uncertain",
|
|
517
|
+
"abstentionRate": 0,
|
|
518
|
+
"providerFailureRate": 0,
|
|
519
|
+
"providerFailures": 0,
|
|
520
|
+
"latencyMs": {
|
|
521
|
+
"mean": 0.00018282125000000348,
|
|
522
|
+
"p50": 0.000136333000000036,
|
|
523
|
+
"p95": 0.0005663339999999835
|
|
524
|
+
},
|
|
525
|
+
"inputTokens": {
|
|
526
|
+
"reportedCases": 24,
|
|
527
|
+
"total": 0,
|
|
528
|
+
"mean": 0
|
|
529
|
+
},
|
|
530
|
+
"denominatorZero": "null, never zero or success"
|
|
531
|
+
},
|
|
532
|
+
"ko": {
|
|
533
|
+
"attempted": 24,
|
|
534
|
+
"correct": 13,
|
|
535
|
+
"confusion": {
|
|
536
|
+
"capability_compatibility": {
|
|
537
|
+
"capability_compatibility": 1,
|
|
538
|
+
"reference": 8,
|
|
539
|
+
"uncertain": 0,
|
|
540
|
+
"provider_failure": 0
|
|
541
|
+
},
|
|
542
|
+
"reference": {
|
|
543
|
+
"capability_compatibility": 0,
|
|
544
|
+
"reference": 12,
|
|
545
|
+
"uncertain": 0,
|
|
546
|
+
"provider_failure": 0
|
|
547
|
+
},
|
|
548
|
+
"uncertain": {
|
|
549
|
+
"capability_compatibility": 0,
|
|
550
|
+
"reference": 3,
|
|
551
|
+
"uncertain": 0,
|
|
552
|
+
"provider_failure": 0
|
|
553
|
+
}
|
|
554
|
+
},
|
|
555
|
+
"accuracy": 0.5416666666666666,
|
|
556
|
+
"semanticOnlyAccuracy": 0.5416666666666666,
|
|
557
|
+
"coverage": 1,
|
|
558
|
+
"trueSpecial": 1,
|
|
559
|
+
"falseSpecial": 0,
|
|
560
|
+
"falseNonSpecial": 8,
|
|
561
|
+
"trueNonSpecial": 15,
|
|
562
|
+
"goldSpecial": 9,
|
|
563
|
+
"goldNonSpecial": 15,
|
|
564
|
+
"specialPrecision": 1,
|
|
565
|
+
"specialRecall": 0.1111111111111111,
|
|
566
|
+
"falseSpecialRate": 0,
|
|
567
|
+
"falseSpecialDenominator": "all gold non-special cases, including gold uncertain",
|
|
568
|
+
"abstentionRate": 0,
|
|
569
|
+
"providerFailureRate": 0,
|
|
570
|
+
"providerFailures": 0,
|
|
571
|
+
"latencyMs": {
|
|
572
|
+
"mean": 0.00015534366666666461,
|
|
573
|
+
"p50": 0.00012225000000000818,
|
|
574
|
+
"p95": 0.00020162500000003546
|
|
575
|
+
},
|
|
576
|
+
"inputTokens": {
|
|
577
|
+
"reportedCases": 24,
|
|
578
|
+
"total": 0,
|
|
579
|
+
"mean": 0
|
|
580
|
+
},
|
|
581
|
+
"denominatorZero": "null, never zero or success"
|
|
582
|
+
},
|
|
583
|
+
"es": {
|
|
584
|
+
"attempted": 24,
|
|
585
|
+
"correct": 12,
|
|
586
|
+
"confusion": {
|
|
587
|
+
"capability_compatibility": {
|
|
588
|
+
"capability_compatibility": 0,
|
|
589
|
+
"reference": 9,
|
|
590
|
+
"uncertain": 0,
|
|
591
|
+
"provider_failure": 0
|
|
592
|
+
},
|
|
593
|
+
"reference": {
|
|
594
|
+
"capability_compatibility": 0,
|
|
595
|
+
"reference": 12,
|
|
596
|
+
"uncertain": 0,
|
|
597
|
+
"provider_failure": 0
|
|
598
|
+
},
|
|
599
|
+
"uncertain": {
|
|
600
|
+
"capability_compatibility": 0,
|
|
601
|
+
"reference": 3,
|
|
602
|
+
"uncertain": 0,
|
|
603
|
+
"provider_failure": 0
|
|
604
|
+
}
|
|
605
|
+
},
|
|
606
|
+
"accuracy": 0.5,
|
|
607
|
+
"semanticOnlyAccuracy": 0.5,
|
|
608
|
+
"coverage": 1,
|
|
609
|
+
"trueSpecial": 0,
|
|
610
|
+
"falseSpecial": 0,
|
|
611
|
+
"falseNonSpecial": 9,
|
|
612
|
+
"trueNonSpecial": 15,
|
|
613
|
+
"goldSpecial": 9,
|
|
614
|
+
"goldNonSpecial": 15,
|
|
615
|
+
"specialPrecision": null,
|
|
616
|
+
"specialRecall": 0,
|
|
617
|
+
"falseSpecialRate": 0,
|
|
618
|
+
"falseSpecialDenominator": "all gold non-special cases, including gold uncertain",
|
|
619
|
+
"abstentionRate": 0,
|
|
620
|
+
"providerFailureRate": 0,
|
|
621
|
+
"providerFailures": 0,
|
|
622
|
+
"latencyMs": {
|
|
623
|
+
"mean": 0.00014477612499999755,
|
|
624
|
+
"p50": 0.00013620800000001054,
|
|
625
|
+
"p95": 0.00021516699999995126
|
|
626
|
+
},
|
|
627
|
+
"inputTokens": {
|
|
628
|
+
"reportedCases": 24,
|
|
629
|
+
"total": 0,
|
|
630
|
+
"mean": 0
|
|
631
|
+
},
|
|
632
|
+
"denominatorZero": "null, never zero or success"
|
|
633
|
+
}
|
|
634
|
+
},
|
|
635
|
+
"latencyMethod": "Frozen regex CPU time: 20 warmup calls and mean of 1000 repetitions per each of the same 120 inputs; repeats are timing only, not extra semantic coverage.",
|
|
636
|
+
"inputTokens": "zero: no model request"
|
|
637
|
+
},
|
|
638
|
+
"accuracyDefinition": "Three-way accuracy counts a genuine abstention on gold uncertain as correct. Gold special abstentions and provider failures are misses. Conditional semantic-only accuracy uses classified answers only. Provider failures are a separate confusion column, never gold uncertain successes.",
|
|
639
|
+
"cacheEditEvidence": {
|
|
640
|
+
"originalReportPreserved": true,
|
|
641
|
+
"finalModelContractMatchesOriginal": true,
|
|
642
|
+
"changes": "Cache current-entry checks before lookup and inside publish; no prompt, provider, model, acceptance threshold, corpus or gold-label change. Final-source live MCP handoff and timing prototype were re-run; the original quality measurement is not relabeled."
|
|
643
|
+
},
|
|
644
|
+
"finalPrototype": {
|
|
645
|
+
"capture": {
|
|
646
|
+
"writeMs": 4.670458999999994,
|
|
647
|
+
"totalMs": 233.71345900000006,
|
|
648
|
+
"inferenceMs": 229.04137500000007,
|
|
649
|
+
"inputTokens": 571
|
|
650
|
+
},
|
|
651
|
+
"firstUse": {
|
|
652
|
+
"captureMs": 2.964375000000018,
|
|
653
|
+
"coldLookupMs": 155.61066600000004,
|
|
654
|
+
"cachedLookupMeanMs": 0.9485958499999981,
|
|
655
|
+
"repeatedLookups": 20,
|
|
656
|
+
"repeatedProviderCalls": 0,
|
|
657
|
+
"inputTokens": 570
|
|
658
|
+
},
|
|
659
|
+
"changedEntry": {
|
|
660
|
+
"providerCalls": 1,
|
|
661
|
+
"elapsedMs": 160.63212499999997,
|
|
662
|
+
"oldObjectState": "stale"
|
|
663
|
+
},
|
|
664
|
+
"selection": "first-use with durable content/revision/policy/configuration cache",
|
|
665
|
+
"rationale": "Capture does not wait for inference; unused corrections cause no model calls. First use pays one inference; unchanged repeated use pays cache I/O. A changed entry is reclassified."
|
|
666
|
+
},
|
|
667
|
+
"finalLiveMcp": {
|
|
668
|
+
"schema": "correction-live-mcp-v1",
|
|
669
|
+
"status": "passed",
|
|
670
|
+
"dshSourceSha": "01e27576d2efe5020c19e93bdba1a6528bfcaee1",
|
|
671
|
+
"mainSourceSha": "5289ed72e347e65eb9a17d8b1d90f51958f8b333",
|
|
672
|
+
"provider": "typesafe",
|
|
673
|
+
"model": "jev-1.13.0",
|
|
674
|
+
"syntheticEntries": 2,
|
|
675
|
+
"providerCalls": 2,
|
|
676
|
+
"repeatProviderCalls": 0,
|
|
677
|
+
"cachedResults": 2,
|
|
678
|
+
"observedRules": [
|
|
679
|
+
{
|
|
680
|
+
"rule": "uncertain",
|
|
681
|
+
"observed": "reference",
|
|
682
|
+
"model": "jev-1.13.0",
|
|
683
|
+
"mode": "shadow",
|
|
684
|
+
"inputTokens": 565
|
|
685
|
+
},
|
|
686
|
+
{
|
|
687
|
+
"rule": "uncertain",
|
|
688
|
+
"observed": "capability_compatibility",
|
|
689
|
+
"model": "jev-1.13.0",
|
|
690
|
+
"mode": "shadow",
|
|
691
|
+
"inputTokens": 571
|
|
692
|
+
}
|
|
693
|
+
],
|
|
694
|
+
"modelInput": "unobserved",
|
|
695
|
+
"verification": "unverified",
|
|
696
|
+
"authorizationGranted": false,
|
|
697
|
+
"qualityMeasurement": "separate held-out corpus, not these two handoff cases"
|
|
698
|
+
},
|
|
699
|
+
"recommendation": "Keep default shadow; provisional release gates failed. No automatic production enablement or arbitrary-language correctness claim.",
|
|
700
|
+
"scope": {
|
|
701
|
+
"heldOutCases": 120,
|
|
702
|
+
"distinctHeldOutSemanticFamilies": 24,
|
|
703
|
+
"languages": [
|
|
704
|
+
"ja",
|
|
705
|
+
"en",
|
|
706
|
+
"zh",
|
|
707
|
+
"ko",
|
|
708
|
+
"es"
|
|
709
|
+
],
|
|
710
|
+
"gold": {
|
|
711
|
+
"special": 45,
|
|
712
|
+
"reference": 60,
|
|
713
|
+
"uncertain": 15
|
|
714
|
+
},
|
|
715
|
+
"independence": "Aligned translations are not 120 independent semantic families."
|
|
716
|
+
}
|
|
717
|
+
}
|