ameforge commited on
Commit
f9e77d6
·
verified ·
1 Parent(s): afe3f06

Upload benchmark_results.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. benchmark_results.json +128 -128
benchmark_results.json CHANGED
@@ -5,48 +5,48 @@
5
  {
6
  "name": "SAM",
7
  "is_sam": true,
8
- "params": "35.9M",
9
- "params_raw": 35911302,
10
- "size_mb": 137.0,
11
- "json_valid": 84.0,
12
- "op_correct": 31.5,
13
- "exact_match": 8.5,
14
- "latency_ms": 691.6,
15
  "n_cases": 200,
16
  "per_tier": [
17
  {
18
  "tier": 1,
19
  "n": 40,
20
- "json_valid": 92.5,
21
- "op_correct": 75.0,
22
- "exact_match": 20.0
23
  },
24
  {
25
  "tier": 2,
26
  "n": 40,
27
- "json_valid": 80.0,
28
- "op_correct": 32.5,
29
- "exact_match": 0.0
30
  },
31
  {
32
  "tier": 3,
33
  "n": 40,
34
- "json_valid": 77.5,
35
- "op_correct": 50.0,
36
- "exact_match": 22.5
37
  },
38
  {
39
  "tier": 4,
40
  "n": 40,
41
- "json_valid": 70.0,
42
  "op_correct": 0.0,
43
  "exact_match": 0.0
44
  },
45
  {
46
  "tier": 5,
47
  "n": 40,
48
- "json_valid": 100.0,
49
- "op_correct": 0.0,
50
  "exact_match": 0.0
51
  }
52
  ],
@@ -56,14 +56,14 @@
56
  "n": 2,
57
  "json_valid": 100.0,
58
  "op_correct": 100.0,
59
- "exact_match": 50.0
60
  },
61
  {
62
  "domain": "DB",
63
  "n": 27,
64
- "json_valid": 77.8,
65
- "op_correct": 48.1,
66
- "exact_match": 3.7
67
  },
68
  {
69
  "domain": "ECOMMERCE",
@@ -76,8 +76,8 @@
76
  "domain": "FILE",
77
  "n": 4,
78
  "json_valid": 100.0,
79
- "op_correct": 50.0,
80
- "exact_match": 0.0
81
  },
82
  {
83
  "domain": "HOME",
@@ -89,23 +89,23 @@
89
  {
90
  "domain": "HTTP",
91
  "n": 25,
92
- "json_valid": 96.0,
93
- "op_correct": 60.0,
94
- "exact_match": 40.0
95
  },
96
  {
97
  "domain": "MQTT",
98
  "n": 15,
99
- "json_valid": 33.3,
100
- "op_correct": 33.3,
101
- "exact_match": 0.0
102
  },
103
  {
104
  "domain": "ROS",
105
  "n": 46,
106
- "json_valid": 87.0,
107
- "op_correct": 26.1,
108
- "exact_match": 4.3
109
  },
110
  {
111
  "domain": "VEHICLE",
@@ -117,9 +117,9 @@
117
  {
118
  "domain": "WORKFLOW",
119
  "n": 49,
120
- "json_valid": 81.6,
121
- "op_correct": 0.0,
122
- "exact_match": 0.0
123
  }
124
  ]
125
  },
@@ -129,10 +129,10 @@
129
  "params": "361.8M",
130
  "params_raw": 361821120,
131
  "size_mb": 690.1,
132
- "json_valid": 98.5,
133
  "op_correct": 0.0,
134
  "exact_match": 0.0,
135
- "latency_ms": 1027.1,
136
  "n_cases": 200,
137
  "per_tier": [
138
  {
@@ -159,7 +159,7 @@
159
  {
160
  "tier": 4,
161
  "n": 40,
162
- "json_valid": 92.5,
163
  "op_correct": 0.0,
164
  "exact_match": 0.0
165
  },
@@ -224,14 +224,14 @@
224
  {
225
  "domain": "ROS",
226
  "n": 46,
227
- "json_valid": 93.5,
228
  "op_correct": 0.0,
229
  "exact_match": 0.0
230
  },
231
  {
232
  "domain": "VEHICLE",
233
  "n": 17,
234
- "json_valid": 100.0,
235
  "op_correct": 0.0,
236
  "exact_match": 0.0
237
  },
@@ -250,10 +250,10 @@
250
  "params": "494.0M",
251
  "params_raw": 494032768,
252
  "size_mb": 942.3,
253
- "json_valid": 71.5,
254
- "op_correct": 28.5,
255
  "exact_match": 2.5,
256
- "latency_ms": 2108.7,
257
  "n_cases": 200,
258
  "per_tier": [
259
  {
@@ -280,8 +280,8 @@
280
  {
281
  "tier": 4,
282
  "n": 40,
283
- "json_valid": 100.0,
284
- "op_correct": 20.0,
285
  "exact_match": 0.0
286
  },
287
  {
@@ -303,7 +303,7 @@
303
  {
304
  "domain": "DB",
305
  "n": 27,
306
- "json_valid": 51.9,
307
  "op_correct": 7.4,
308
  "exact_match": 0.0
309
  },
@@ -353,7 +353,7 @@
353
  "domain": "VEHICLE",
354
  "n": 17,
355
  "json_valid": 82.4,
356
- "op_correct": 64.7,
357
  "exact_match": 17.6
358
  },
359
  {
@@ -371,10 +371,10 @@
371
  "params": "1100.0M",
372
  "params_raw": 1100048384,
373
  "size_mb": 2098.2,
374
- "json_valid": 93.5,
375
- "op_correct": 57.0,
376
  "exact_match": 9.0,
377
- "latency_ms": 4484.0,
378
  "n_cases": 200,
379
  "per_tier": [
380
  {
@@ -388,7 +388,7 @@
388
  "tier": 2,
389
  "n": 40,
390
  "json_valid": 90.0,
391
- "op_correct": 90.0,
392
  "exact_match": 12.5
393
  },
394
  {
@@ -401,7 +401,7 @@
401
  {
402
  "tier": 4,
403
  "n": 40,
404
- "json_valid": 77.5,
405
  "op_correct": 22.5,
406
  "exact_match": 0.0
407
  },
@@ -466,14 +466,14 @@
466
  {
467
  "domain": "ROS",
468
  "n": 46,
469
- "json_valid": 80.4,
470
- "op_correct": 54.3,
471
  "exact_match": 17.4
472
  },
473
  {
474
  "domain": "VEHICLE",
475
  "n": 17,
476
- "json_valid": 100.0,
477
  "op_correct": 35.3,
478
  "exact_match": 23.5
479
  },
@@ -505,9 +505,9 @@
505
  "params_raw": 1543714304,
506
  "size_mb": 2944.4,
507
  "json_valid": 100.0,
508
- "op_correct": 75.0,
509
- "exact_match": 9.5,
510
- "latency_ms": 1976.9,
511
  "n_cases": 200,
512
  "per_tier": [
513
  {
@@ -521,8 +521,8 @@
521
  "tier": 2,
522
  "n": 40,
523
  "json_valid": 100.0,
524
- "op_correct": 90.0,
525
- "exact_match": 15.0
526
  },
527
  {
528
  "tier": 3,
@@ -535,7 +535,7 @@
535
  "tier": 4,
536
  "n": 40,
537
  "json_valid": 100.0,
538
- "op_correct": 50.0,
539
  "exact_match": 0.0
540
  },
541
  {
@@ -558,7 +558,7 @@
558
  "domain": "DB",
559
  "n": 27,
560
  "json_valid": 100.0,
561
- "op_correct": 66.7,
562
  "exact_match": 18.5
563
  },
564
  {
@@ -600,8 +600,8 @@
600
  "domain": "ROS",
601
  "n": 46,
602
  "json_valid": 100.0,
603
- "op_correct": 76.1,
604
- "exact_match": 21.7
605
  },
606
  {
607
  "domain": "VEHICLE",
@@ -625,10 +625,10 @@
625
  "params": "1711.4M",
626
  "params_raw": 1711376384,
627
  "size_mb": 3264.2,
628
- "json_valid": 85.0,
629
- "op_correct": 47.0,
630
  "exact_match": 2.5,
631
- "latency_ms": 1884.5,
632
  "n_cases": 200,
633
  "per_tier": [
634
  {
@@ -641,7 +641,7 @@
641
  {
642
  "tier": 2,
643
  "n": 40,
644
- "json_valid": 72.5,
645
  "op_correct": 62.5,
646
  "exact_match": 0.0
647
  },
@@ -655,15 +655,15 @@
655
  {
656
  "tier": 4,
657
  "n": 40,
658
- "json_valid": 60.0,
659
- "op_correct": 2.5,
660
  "exact_match": 0.0
661
  },
662
  {
663
  "tier": 5,
664
  "n": 40,
665
  "json_valid": 100.0,
666
- "op_correct": 40.0,
667
  "exact_match": 0.0
668
  }
669
  ],
@@ -720,22 +720,22 @@
720
  {
721
  "domain": "ROS",
722
  "n": 46,
723
- "json_valid": 89.1,
724
  "op_correct": 47.8,
725
  "exact_match": 4.3
726
  },
727
  {
728
  "domain": "VEHICLE",
729
  "n": 17,
730
- "json_valid": 58.8,
731
- "op_correct": 29.4,
732
  "exact_match": 17.6
733
  },
734
  {
735
  "domain": "WORKFLOW",
736
  "n": 49,
737
  "json_valid": 83.7,
738
- "op_correct": 34.7,
739
  "exact_match": 0.0
740
  }
741
  ]
@@ -759,9 +759,9 @@
759
  "params_raw": 3085938688,
760
  "size_mb": 5886.0,
761
  "json_valid": 100.0,
762
- "op_correct": 81.5,
763
- "exact_match": 17.5,
764
- "latency_ms": 2716.7,
765
  "n_cases": 200,
766
  "per_tier": [
767
  {
@@ -776,7 +776,7 @@
776
  "n": 40,
777
  "json_valid": 100.0,
778
  "op_correct": 100.0,
779
- "exact_match": 20.0
780
  },
781
  {
782
  "tier": 3,
@@ -789,7 +789,7 @@
789
  "tier": 4,
790
  "n": 40,
791
  "json_valid": 100.0,
792
- "op_correct": 70.0,
793
  "exact_match": 0.0
794
  },
795
  {
@@ -812,7 +812,7 @@
812
  "domain": "DB",
813
  "n": 27,
814
  "json_valid": 100.0,
815
- "op_correct": 66.7,
816
  "exact_match": 18.5
817
  },
818
  {
@@ -854,8 +854,8 @@
854
  "domain": "ROS",
855
  "n": 46,
856
  "json_valid": 100.0,
857
- "op_correct": 93.5,
858
- "exact_match": 28.3
859
  },
860
  {
861
  "domain": "VEHICLE",
@@ -916,9 +916,9 @@
916
  "params_raw": 270000000,
917
  "size_mb": 241.4,
918
  "json_valid": 92.0,
919
- "op_correct": 2.5,
920
  "exact_match": 0.0,
921
- "latency_ms": 6871.5,
922
  "n_cases": 200,
923
  "per_tier": [
924
  {
@@ -932,7 +932,7 @@
932
  "tier": 2,
933
  "n": 40,
934
  "json_valid": 97.5,
935
- "op_correct": 7.5,
936
  "exact_match": 0.0
937
  },
938
  {
@@ -968,7 +968,7 @@
968
  {
969
  "domain": "DB",
970
  "n": 27,
971
- "json_valid": 85.2,
972
  "op_correct": 0.0,
973
  "exact_match": 0.0
974
  },
@@ -1010,8 +1010,8 @@
1010
  {
1011
  "domain": "ROS",
1012
  "n": 46,
1013
- "json_valid": 89.1,
1014
- "op_correct": 2.2,
1015
  "exact_match": 0.0
1016
  },
1017
  {
@@ -1048,10 +1048,10 @@
1048
  "params": "8000.0M",
1049
  "params_raw": 8000000000,
1050
  "size_mb": 4794.9,
1051
- "json_valid": 98.5,
1052
- "op_correct": 79.5,
1053
  "exact_match": 11.5,
1054
- "latency_ms": 37642.4,
1055
  "n_cases": 200,
1056
  "per_tier": [
1057
  {
@@ -1078,8 +1078,8 @@
1078
  {
1079
  "tier": 4,
1080
  "n": 40,
1081
- "json_valid": 97.5,
1082
- "op_correct": 95.0,
1083
  "exact_match": 0.0
1084
  },
1085
  {
@@ -1143,8 +1143,8 @@
1143
  {
1144
  "domain": "ROS",
1145
  "n": 46,
1146
- "json_valid": 97.8,
1147
- "op_correct": 71.7,
1148
  "exact_match": 10.9
1149
  },
1150
  {
@@ -1168,31 +1168,31 @@
1168
  {
1169
  "tier": 1,
1170
  "n": 40,
1171
- "json_valid": 92.5,
1172
- "op_correct": 75.0,
1173
- "exact_match": 20.0,
1174
  "description": "Atomic \u2014 single op"
1175
  },
1176
  {
1177
  "tier": 2,
1178
  "n": 40,
1179
- "json_valid": 80.0,
1180
- "op_correct": 32.5,
1181
- "exact_match": 0.0,
1182
  "description": "Compound \u2014 2-3 ops same domain"
1183
  },
1184
  {
1185
  "tier": 3,
1186
  "n": 40,
1187
- "json_valid": 77.5,
1188
- "op_correct": 50.0,
1189
- "exact_match": 22.5,
1190
  "description": "Noisy \u2014 typos / casing / fillers"
1191
  },
1192
  {
1193
  "tier": 4,
1194
  "n": 40,
1195
- "json_valid": 70.0,
1196
  "op_correct": 0.0,
1197
  "exact_match": 0.0,
1198
  "description": "Long chain \u2014 4+ ops"
@@ -1200,8 +1200,8 @@
1200
  {
1201
  "tier": 5,
1202
  "n": 40,
1203
- "json_valid": 100.0,
1204
- "op_correct": 0.0,
1205
  "exact_match": 0.0,
1206
  "description": "Cross-domain workflow"
1207
  }
@@ -1212,14 +1212,14 @@
1212
  "n": 2,
1213
  "json_valid": 100.0,
1214
  "op_correct": 100.0,
1215
- "exact_match": 50.0
1216
  },
1217
  {
1218
  "domain": "DB",
1219
  "n": 27,
1220
- "json_valid": 77.8,
1221
- "op_correct": 48.1,
1222
- "exact_match": 3.7
1223
  },
1224
  {
1225
  "domain": "ECOMMERCE",
@@ -1232,8 +1232,8 @@
1232
  "domain": "FILE",
1233
  "n": 4,
1234
  "json_valid": 100.0,
1235
- "op_correct": 50.0,
1236
- "exact_match": 0.0
1237
  },
1238
  {
1239
  "domain": "HOME",
@@ -1245,23 +1245,23 @@
1245
  {
1246
  "domain": "HTTP",
1247
  "n": 25,
1248
- "json_valid": 96.0,
1249
- "op_correct": 60.0,
1250
- "exact_match": 40.0
1251
  },
1252
  {
1253
  "domain": "MQTT",
1254
  "n": 15,
1255
- "json_valid": 33.3,
1256
- "op_correct": 33.3,
1257
- "exact_match": 0.0
1258
  },
1259
  {
1260
  "domain": "ROS",
1261
  "n": 46,
1262
- "json_valid": 87.0,
1263
- "op_correct": 26.1,
1264
- "exact_match": 4.3
1265
  },
1266
  {
1267
  "domain": "VEHICLE",
@@ -1273,9 +1273,9 @@
1273
  {
1274
  "domain": "WORKFLOW",
1275
  "n": 49,
1276
- "json_valid": 81.6,
1277
- "op_correct": 0.0,
1278
- "exact_match": 0.0
1279
  }
1280
  ],
1281
  "tier_distribution": {
 
5
  {
6
  "name": "SAM",
7
  "is_sam": true,
8
+ "params": "33.4M",
9
+ "params_raw": 33400324,
10
+ "size_mb": 127.4,
11
+ "json_valid": 88.0,
12
+ "op_correct": 63.5,
13
+ "exact_match": 18.0,
14
+ "latency_ms": 715.4,
15
  "n_cases": 200,
16
  "per_tier": [
17
  {
18
  "tier": 1,
19
  "n": 40,
20
+ "json_valid": 100.0,
21
+ "op_correct": 82.5,
22
+ "exact_match": 37.5
23
  },
24
  {
25
  "tier": 2,
26
  "n": 40,
27
+ "json_valid": 100.0,
28
+ "op_correct": 70.0,
29
+ "exact_match": 27.5
30
  },
31
  {
32
  "tier": 3,
33
  "n": 40,
34
+ "json_valid": 100.0,
35
+ "op_correct": 100.0,
36
+ "exact_match": 25.0
37
  },
38
  {
39
  "tier": 4,
40
  "n": 40,
41
+ "json_valid": 75.0,
42
  "op_correct": 0.0,
43
  "exact_match": 0.0
44
  },
45
  {
46
  "tier": 5,
47
  "n": 40,
48
+ "json_valid": 65.0,
49
+ "op_correct": 65.0,
50
  "exact_match": 0.0
51
  }
52
  ],
 
56
  "n": 2,
57
  "json_valid": 100.0,
58
  "op_correct": 100.0,
59
+ "exact_match": 100.0
60
  },
61
  {
62
  "domain": "DB",
63
  "n": 27,
64
+ "json_valid": 66.7,
65
+ "op_correct": 66.7,
66
+ "exact_match": 11.1
67
  },
68
  {
69
  "domain": "ECOMMERCE",
 
76
  "domain": "FILE",
77
  "n": 4,
78
  "json_valid": 100.0,
79
+ "op_correct": 100.0,
80
+ "exact_match": 75.0
81
  },
82
  {
83
  "domain": "HOME",
 
89
  {
90
  "domain": "HTTP",
91
  "n": 25,
92
+ "json_valid": 100.0,
93
+ "op_correct": 64.0,
94
+ "exact_match": 44.0
95
  },
96
  {
97
  "domain": "MQTT",
98
  "n": 15,
99
+ "json_valid": 100.0,
100
+ "op_correct": 100.0,
101
+ "exact_match": 6.7
102
  },
103
  {
104
  "domain": "ROS",
105
  "n": 46,
106
+ "json_valid": 97.8,
107
+ "op_correct": 50.0,
108
+ "exact_match": 10.9
109
  },
110
  {
111
  "domain": "VEHICLE",
 
117
  {
118
  "domain": "WORKFLOW",
119
  "n": 49,
120
+ "json_valid": 71.4,
121
+ "op_correct": 71.4,
122
+ "exact_match": 16.3
123
  }
124
  ]
125
  },
 
129
  "params": "361.8M",
130
  "params_raw": 361821120,
131
  "size_mb": 690.1,
132
+ "json_valid": 89.0,
133
  "op_correct": 0.0,
134
  "exact_match": 0.0,
135
+ "latency_ms": 2042.9,
136
  "n_cases": 200,
137
  "per_tier": [
138
  {
 
159
  {
160
  "tier": 4,
161
  "n": 40,
162
+ "json_valid": 45.0,
163
  "op_correct": 0.0,
164
  "exact_match": 0.0
165
  },
 
224
  {
225
  "domain": "ROS",
226
  "n": 46,
227
+ "json_valid": 76.1,
228
  "op_correct": 0.0,
229
  "exact_match": 0.0
230
  },
231
  {
232
  "domain": "VEHICLE",
233
  "n": 17,
234
+ "json_valid": 35.3,
235
  "op_correct": 0.0,
236
  "exact_match": 0.0
237
  },
 
250
  "params": "494.0M",
251
  "params_raw": 494032768,
252
  "size_mb": 942.3,
253
+ "json_valid": 70.5,
254
+ "op_correct": 30.0,
255
  "exact_match": 2.5,
256
+ "latency_ms": 1937.2,
257
  "n_cases": 200,
258
  "per_tier": [
259
  {
 
280
  {
281
  "tier": 4,
282
  "n": 40,
283
+ "json_valid": 95.0,
284
+ "op_correct": 27.5,
285
  "exact_match": 0.0
286
  },
287
  {
 
303
  {
304
  "domain": "DB",
305
  "n": 27,
306
+ "json_valid": 44.4,
307
  "op_correct": 7.4,
308
  "exact_match": 0.0
309
  },
 
353
  "domain": "VEHICLE",
354
  "n": 17,
355
  "json_valid": 82.4,
356
+ "op_correct": 82.4,
357
  "exact_match": 17.6
358
  },
359
  {
 
371
  "params": "1100.0M",
372
  "params_raw": 1100048384,
373
  "size_mb": 2098.2,
374
+ "json_valid": 92.0,
375
+ "op_correct": 55.5,
376
  "exact_match": 9.0,
377
+ "latency_ms": 4199.3,
378
  "n_cases": 200,
379
  "per_tier": [
380
  {
 
388
  "tier": 2,
389
  "n": 40,
390
  "json_valid": 90.0,
391
+ "op_correct": 82.5,
392
  "exact_match": 12.5
393
  },
394
  {
 
401
  {
402
  "tier": 4,
403
  "n": 40,
404
+ "json_valid": 70.0,
405
  "op_correct": 22.5,
406
  "exact_match": 0.0
407
  },
 
466
  {
467
  "domain": "ROS",
468
  "n": 46,
469
+ "json_valid": 76.1,
470
+ "op_correct": 47.8,
471
  "exact_match": 17.4
472
  },
473
  {
474
  "domain": "VEHICLE",
475
  "n": 17,
476
+ "json_valid": 94.1,
477
  "op_correct": 35.3,
478
  "exact_match": 23.5
479
  },
 
505
  "params_raw": 1543714304,
506
  "size_mb": 2944.4,
507
  "json_valid": 100.0,
508
+ "op_correct": 78.5,
509
+ "exact_match": 9.0,
510
+ "latency_ms": 1852.4,
511
  "n_cases": 200,
512
  "per_tier": [
513
  {
 
521
  "tier": 2,
522
  "n": 40,
523
  "json_valid": 100.0,
524
+ "op_correct": 85.0,
525
+ "exact_match": 12.5
526
  },
527
  {
528
  "tier": 3,
 
535
  "tier": 4,
536
  "n": 40,
537
  "json_valid": 100.0,
538
+ "op_correct": 72.5,
539
  "exact_match": 0.0
540
  },
541
  {
 
558
  "domain": "DB",
559
  "n": 27,
560
  "json_valid": 100.0,
561
+ "op_correct": 100.0,
562
  "exact_match": 18.5
563
  },
564
  {
 
600
  "domain": "ROS",
601
  "n": 46,
602
  "json_valid": 100.0,
603
+ "op_correct": 71.7,
604
+ "exact_match": 19.6
605
  },
606
  {
607
  "domain": "VEHICLE",
 
625
  "params": "1711.4M",
626
  "params_raw": 1711376384,
627
  "size_mb": 3264.2,
628
+ "json_valid": 89.0,
629
+ "op_correct": 38.5,
630
  "exact_match": 2.5,
631
+ "latency_ms": 1694.9,
632
  "n_cases": 200,
633
  "per_tier": [
634
  {
 
641
  {
642
  "tier": 2,
643
  "n": 40,
644
+ "json_valid": 80.0,
645
  "op_correct": 62.5,
646
  "exact_match": 0.0
647
  },
 
655
  {
656
  "tier": 4,
657
  "n": 40,
658
+ "json_valid": 72.5,
659
+ "op_correct": 0.0,
660
  "exact_match": 0.0
661
  },
662
  {
663
  "tier": 5,
664
  "n": 40,
665
  "json_valid": 100.0,
666
+ "op_correct": 0.0,
667
  "exact_match": 0.0
668
  }
669
  ],
 
720
  {
721
  "domain": "ROS",
722
  "n": 46,
723
+ "json_valid": 95.7,
724
  "op_correct": 47.8,
725
  "exact_match": 4.3
726
  },
727
  {
728
  "domain": "VEHICLE",
729
  "n": 17,
730
+ "json_valid": 88.2,
731
+ "op_correct": 23.5,
732
  "exact_match": 17.6
733
  },
734
  {
735
  "domain": "WORKFLOW",
736
  "n": 49,
737
  "json_valid": 83.7,
738
+ "op_correct": 2.0,
739
  "exact_match": 0.0
740
  }
741
  ]
 
759
  "params_raw": 3085938688,
760
  "size_mb": 5886.0,
761
  "json_valid": 100.0,
762
+ "op_correct": 82.5,
763
+ "exact_match": 16.0,
764
+ "latency_ms": 2061.1,
765
  "n_cases": 200,
766
  "per_tier": [
767
  {
 
776
  "n": 40,
777
  "json_valid": 100.0,
778
  "op_correct": 100.0,
779
+ "exact_match": 12.5
780
  },
781
  {
782
  "tier": 3,
 
789
  "tier": 4,
790
  "n": 40,
791
  "json_valid": 100.0,
792
+ "op_correct": 75.0,
793
  "exact_match": 0.0
794
  },
795
  {
 
812
  "domain": "DB",
813
  "n": 27,
814
  "json_valid": 100.0,
815
+ "op_correct": 100.0,
816
  "exact_match": 18.5
817
  },
818
  {
 
854
  "domain": "ROS",
855
  "n": 46,
856
  "json_valid": 100.0,
857
+ "op_correct": 78.3,
858
+ "exact_match": 21.7
859
  },
860
  {
861
  "domain": "VEHICLE",
 
916
  "params_raw": 270000000,
917
  "size_mb": 241.4,
918
  "json_valid": 92.0,
919
+ "op_correct": 2.0,
920
  "exact_match": 0.0,
921
+ "latency_ms": 6322.5,
922
  "n_cases": 200,
923
  "per_tier": [
924
  {
 
932
  "tier": 2,
933
  "n": 40,
934
  "json_valid": 97.5,
935
+ "op_correct": 5.0,
936
  "exact_match": 0.0
937
  },
938
  {
 
968
  {
969
  "domain": "DB",
970
  "n": 27,
971
+ "json_valid": 88.9,
972
  "op_correct": 0.0,
973
  "exact_match": 0.0
974
  },
 
1010
  {
1011
  "domain": "ROS",
1012
  "n": 46,
1013
+ "json_valid": 87.0,
1014
+ "op_correct": 0.0,
1015
  "exact_match": 0.0
1016
  },
1017
  {
 
1048
  "params": "8000.0M",
1049
  "params_raw": 8000000000,
1050
  "size_mb": 4794.9,
1051
+ "json_valid": 98.0,
1052
+ "op_correct": 79.0,
1053
  "exact_match": 11.5,
1054
+ "latency_ms": 31737.9,
1055
  "n_cases": 200,
1056
  "per_tier": [
1057
  {
 
1078
  {
1079
  "tier": 4,
1080
  "n": 40,
1081
+ "json_valid": 95.0,
1082
+ "op_correct": 92.5,
1083
  "exact_match": 0.0
1084
  },
1085
  {
 
1143
  {
1144
  "domain": "ROS",
1145
  "n": 46,
1146
+ "json_valid": 95.7,
1147
+ "op_correct": 69.6,
1148
  "exact_match": 10.9
1149
  },
1150
  {
 
1168
  {
1169
  "tier": 1,
1170
  "n": 40,
1171
+ "json_valid": 100.0,
1172
+ "op_correct": 82.5,
1173
+ "exact_match": 37.5,
1174
  "description": "Atomic \u2014 single op"
1175
  },
1176
  {
1177
  "tier": 2,
1178
  "n": 40,
1179
+ "json_valid": 100.0,
1180
+ "op_correct": 70.0,
1181
+ "exact_match": 27.5,
1182
  "description": "Compound \u2014 2-3 ops same domain"
1183
  },
1184
  {
1185
  "tier": 3,
1186
  "n": 40,
1187
+ "json_valid": 100.0,
1188
+ "op_correct": 100.0,
1189
+ "exact_match": 25.0,
1190
  "description": "Noisy \u2014 typos / casing / fillers"
1191
  },
1192
  {
1193
  "tier": 4,
1194
  "n": 40,
1195
+ "json_valid": 75.0,
1196
  "op_correct": 0.0,
1197
  "exact_match": 0.0,
1198
  "description": "Long chain \u2014 4+ ops"
 
1200
  {
1201
  "tier": 5,
1202
  "n": 40,
1203
+ "json_valid": 65.0,
1204
+ "op_correct": 65.0,
1205
  "exact_match": 0.0,
1206
  "description": "Cross-domain workflow"
1207
  }
 
1212
  "n": 2,
1213
  "json_valid": 100.0,
1214
  "op_correct": 100.0,
1215
+ "exact_match": 100.0
1216
  },
1217
  {
1218
  "domain": "DB",
1219
  "n": 27,
1220
+ "json_valid": 66.7,
1221
+ "op_correct": 66.7,
1222
+ "exact_match": 11.1
1223
  },
1224
  {
1225
  "domain": "ECOMMERCE",
 
1232
  "domain": "FILE",
1233
  "n": 4,
1234
  "json_valid": 100.0,
1235
+ "op_correct": 100.0,
1236
+ "exact_match": 75.0
1237
  },
1238
  {
1239
  "domain": "HOME",
 
1245
  {
1246
  "domain": "HTTP",
1247
  "n": 25,
1248
+ "json_valid": 100.0,
1249
+ "op_correct": 64.0,
1250
+ "exact_match": 44.0
1251
  },
1252
  {
1253
  "domain": "MQTT",
1254
  "n": 15,
1255
+ "json_valid": 100.0,
1256
+ "op_correct": 100.0,
1257
+ "exact_match": 6.7
1258
  },
1259
  {
1260
  "domain": "ROS",
1261
  "n": 46,
1262
+ "json_valid": 97.8,
1263
+ "op_correct": 50.0,
1264
+ "exact_match": 10.9
1265
  },
1266
  {
1267
  "domain": "VEHICLE",
 
1273
  {
1274
  "domain": "WORKFLOW",
1275
  "n": 49,
1276
+ "json_valid": 71.4,
1277
+ "op_correct": 71.4,
1278
+ "exact_match": 16.3
1279
  }
1280
  ],
1281
  "tier_distribution": {