Datasets:

Modalities:
Text
Languages:
English
ArXiv:
Libraries:
Datasets
License:
Clémentine commited on
Commit
27d6bc8
1 Parent(s): 0a46935
Files changed (5) hide show
  1. .gitattributes +1 -0
  2. README.md +1836 -0
  3. data.tar +3 -0
  4. dataset_infos.json +0 -0
  5. mmlu.py +171 -0
.gitattributes CHANGED
@@ -52,3 +52,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
52
  *.jpg filter=lfs diff=lfs merge=lfs -text
53
  *.jpeg filter=lfs diff=lfs merge=lfs -text
54
  *.webp filter=lfs diff=lfs merge=lfs -text
 
 
52
  *.jpg filter=lfs diff=lfs merge=lfs -text
53
  *.jpeg filter=lfs diff=lfs merge=lfs -text
54
  *.webp filter=lfs diff=lfs merge=lfs -text
55
+ *.tar filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,1836 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ annotations_creators:
3
+ - no-annotation
4
+ language_creators:
5
+ - expert-generated
6
+ language:
7
+ - en
8
+ license:
9
+ - mit
10
+ multilinguality:
11
+ - monolingual
12
+ size_categories:
13
+ - 10K<n<100K
14
+ source_datasets:
15
+ - original
16
+ task_categories:
17
+ - question-answering
18
+ task_ids:
19
+ - multiple-choice-qa
20
+ paperswithcode_id: mmlu
21
+ pretty_name: Measuring Massive Multitask Language Understanding
22
+ language_bcp47:
23
+ - en-US
24
+ dataset_info:
25
+ - config_name: abstract_algebra
26
+ features:
27
+ - name: question
28
+ dtype: string
29
+ - name: choices
30
+ sequence: string
31
+ - name: answer
32
+ dtype:
33
+ class_label:
34
+ names:
35
+ '0': A
36
+ '1': B
37
+ '2': C
38
+ '3': D
39
+ splits:
40
+ - name: auxiliary_train
41
+ num_bytes: 160601377
42
+ num_examples: 99842
43
+ - name: test
44
+ num_bytes: 19328
45
+ num_examples: 100
46
+ - name: validation
47
+ num_bytes: 2024
48
+ num_examples: 11
49
+ - name: dev
50
+ num_bytes: 830
51
+ num_examples: 5
52
+ download_size: 166184960
53
+ dataset_size: 160623559
54
+ - config_name: anatomy
55
+ features:
56
+ - name: question
57
+ dtype: string
58
+ - name: choices
59
+ sequence: string
60
+ - name: answer
61
+ dtype:
62
+ class_label:
63
+ names:
64
+ '0': A
65
+ '1': B
66
+ '2': C
67
+ '3': D
68
+ splits:
69
+ - name: auxiliary_train
70
+ num_bytes: 160601377
71
+ num_examples: 99842
72
+ - name: test
73
+ num_bytes: 33121
74
+ num_examples: 135
75
+ - name: validation
76
+ num_bytes: 3140
77
+ num_examples: 14
78
+ - name: dev
79
+ num_bytes: 967
80
+ num_examples: 5
81
+ download_size: 166184960
82
+ dataset_size: 160638605
83
+ - config_name: astronomy
84
+ features:
85
+ - name: question
86
+ dtype: string
87
+ - name: choices
88
+ sequence: string
89
+ - name: answer
90
+ dtype:
91
+ class_label:
92
+ names:
93
+ '0': A
94
+ '1': B
95
+ '2': C
96
+ '3': D
97
+ splits:
98
+ - name: auxiliary_train
99
+ num_bytes: 160601377
100
+ num_examples: 99842
101
+ - name: test
102
+ num_bytes: 46771
103
+ num_examples: 152
104
+ - name: validation
105
+ num_bytes: 5027
106
+ num_examples: 16
107
+ - name: dev
108
+ num_bytes: 2076
109
+ num_examples: 5
110
+ download_size: 166184960
111
+ dataset_size: 160655251
112
+ - config_name: business_ethics
113
+ features:
114
+ - name: question
115
+ dtype: string
116
+ - name: choices
117
+ sequence: string
118
+ - name: answer
119
+ dtype:
120
+ class_label:
121
+ names:
122
+ '0': A
123
+ '1': B
124
+ '2': C
125
+ '3': D
126
+ splits:
127
+ - name: auxiliary_train
128
+ num_bytes: 160601377
129
+ num_examples: 99842
130
+ - name: test
131
+ num_bytes: 33252
132
+ num_examples: 100
133
+ - name: validation
134
+ num_bytes: 3038
135
+ num_examples: 11
136
+ - name: dev
137
+ num_bytes: 2190
138
+ num_examples: 5
139
+ download_size: 166184960
140
+ dataset_size: 160639857
141
+ - config_name: clinical_knowledge
142
+ features:
143
+ - name: question
144
+ dtype: string
145
+ - name: choices
146
+ sequence: string
147
+ - name: answer
148
+ dtype:
149
+ class_label:
150
+ names:
151
+ '0': A
152
+ '1': B
153
+ '2': C
154
+ '3': D
155
+ splits:
156
+ - name: auxiliary_train
157
+ num_bytes: 160601377
158
+ num_examples: 99842
159
+ - name: test
160
+ num_bytes: 62754
161
+ num_examples: 265
162
+ - name: validation
163
+ num_bytes: 6664
164
+ num_examples: 29
165
+ - name: dev
166
+ num_bytes: 1210
167
+ num_examples: 5
168
+ download_size: 166184960
169
+ dataset_size: 160672005
170
+ - config_name: college_biology
171
+ features:
172
+ - name: question
173
+ dtype: string
174
+ - name: choices
175
+ sequence: string
176
+ - name: answer
177
+ dtype:
178
+ class_label:
179
+ names:
180
+ '0': A
181
+ '1': B
182
+ '2': C
183
+ '3': D
184
+ splits:
185
+ - name: auxiliary_train
186
+ num_bytes: 160601377
187
+ num_examples: 99842
188
+ - name: test
189
+ num_bytes: 48797
190
+ num_examples: 144
191
+ - name: validation
192
+ num_bytes: 4819
193
+ num_examples: 16
194
+ - name: dev
195
+ num_bytes: 1532
196
+ num_examples: 5
197
+ download_size: 166184960
198
+ dataset_size: 160656525
199
+ - config_name: college_chemistry
200
+ features:
201
+ - name: question
202
+ dtype: string
203
+ - name: choices
204
+ sequence: string
205
+ - name: answer
206
+ dtype:
207
+ class_label:
208
+ names:
209
+ '0': A
210
+ '1': B
211
+ '2': C
212
+ '3': D
213
+ splits:
214
+ - name: auxiliary_train
215
+ num_bytes: 160601377
216
+ num_examples: 99842
217
+ - name: test
218
+ num_bytes: 24708
219
+ num_examples: 100
220
+ - name: validation
221
+ num_bytes: 2328
222
+ num_examples: 8
223
+ - name: dev
224
+ num_bytes: 1331
225
+ num_examples: 5
226
+ download_size: 166184960
227
+ dataset_size: 160629744
228
+ - config_name: college_computer_science
229
+ features:
230
+ - name: question
231
+ dtype: string
232
+ - name: choices
233
+ sequence: string
234
+ - name: answer
235
+ dtype:
236
+ class_label:
237
+ names:
238
+ '0': A
239
+ '1': B
240
+ '2': C
241
+ '3': D
242
+ splits:
243
+ - name: auxiliary_train
244
+ num_bytes: 160601377
245
+ num_examples: 99842
246
+ - name: test
247
+ num_bytes: 42641
248
+ num_examples: 100
249
+ - name: validation
250
+ num_bytes: 4663
251
+ num_examples: 11
252
+ - name: dev
253
+ num_bytes: 2765
254
+ num_examples: 5
255
+ download_size: 166184960
256
+ dataset_size: 160651446
257
+ - config_name: college_mathematics
258
+ features:
259
+ - name: question
260
+ dtype: string
261
+ - name: choices
262
+ sequence: string
263
+ - name: answer
264
+ dtype:
265
+ class_label:
266
+ names:
267
+ '0': A
268
+ '1': B
269
+ '2': C
270
+ '3': D
271
+ splits:
272
+ - name: auxiliary_train
273
+ num_bytes: 160601377
274
+ num_examples: 99842
275
+ - name: test
276
+ num_bytes: 24711
277
+ num_examples: 100
278
+ - name: validation
279
+ num_bytes: 2668
280
+ num_examples: 11
281
+ - name: dev
282
+ num_bytes: 1493
283
+ num_examples: 5
284
+ download_size: 166184960
285
+ dataset_size: 160630249
286
+ - config_name: college_medicine
287
+ features:
288
+ - name: question
289
+ dtype: string
290
+ - name: choices
291
+ sequence: string
292
+ - name: answer
293
+ dtype:
294
+ class_label:
295
+ names:
296
+ '0': A
297
+ '1': B
298
+ '2': C
299
+ '3': D
300
+ splits:
301
+ - name: auxiliary_train
302
+ num_bytes: 160601377
303
+ num_examples: 99842
304
+ - name: test
305
+ num_bytes: 82397
306
+ num_examples: 173
307
+ - name: validation
308
+ num_bytes: 7909
309
+ num_examples: 22
310
+ - name: dev
311
+ num_bytes: 1670
312
+ num_examples: 5
313
+ download_size: 166184960
314
+ dataset_size: 160693353
315
+ - config_name: college_physics
316
+ features:
317
+ - name: question
318
+ dtype: string
319
+ - name: choices
320
+ sequence: string
321
+ - name: answer
322
+ dtype:
323
+ class_label:
324
+ names:
325
+ '0': A
326
+ '1': B
327
+ '2': C
328
+ '3': D
329
+ splits:
330
+ - name: auxiliary_train
331
+ num_bytes: 160601377
332
+ num_examples: 99842
333
+ - name: test
334
+ num_bytes: 30181
335
+ num_examples: 102
336
+ - name: validation
337
+ num_bytes: 3490
338
+ num_examples: 11
339
+ - name: dev
340
+ num_bytes: 1412
341
+ num_examples: 5
342
+ download_size: 166184960
343
+ dataset_size: 160636460
344
+ - config_name: computer_security
345
+ features:
346
+ - name: question
347
+ dtype: string
348
+ - name: choices
349
+ sequence: string
350
+ - name: answer
351
+ dtype:
352
+ class_label:
353
+ names:
354
+ '0': A
355
+ '1': B
356
+ '2': C
357
+ '3': D
358
+ splits:
359
+ - name: auxiliary_train
360
+ num_bytes: 160601377
361
+ num_examples: 99842
362
+ - name: test
363
+ num_bytes: 27124
364
+ num_examples: 100
365
+ - name: validation
366
+ num_bytes: 4549
367
+ num_examples: 11
368
+ - name: dev
369
+ num_bytes: 1101
370
+ num_examples: 5
371
+ download_size: 166184960
372
+ dataset_size: 160634151
373
+ - config_name: conceptual_physics
374
+ features:
375
+ - name: question
376
+ dtype: string
377
+ - name: choices
378
+ sequence: string
379
+ - name: answer
380
+ dtype:
381
+ class_label:
382
+ names:
383
+ '0': A
384
+ '1': B
385
+ '2': C
386
+ '3': D
387
+ splits:
388
+ - name: auxiliary_train
389
+ num_bytes: 160601377
390
+ num_examples: 99842
391
+ - name: test
392
+ num_bytes: 40709
393
+ num_examples: 235
394
+ - name: validation
395
+ num_bytes: 4474
396
+ num_examples: 26
397
+ - name: dev
398
+ num_bytes: 934
399
+ num_examples: 5
400
+ download_size: 166184960
401
+ dataset_size: 160647494
402
+ - config_name: econometrics
403
+ features:
404
+ - name: question
405
+ dtype: string
406
+ - name: choices
407
+ sequence: string
408
+ - name: answer
409
+ dtype:
410
+ class_label:
411
+ names:
412
+ '0': A
413
+ '1': B
414
+ '2': C
415
+ '3': D
416
+ splits:
417
+ - name: auxiliary_train
418
+ num_bytes: 160601377
419
+ num_examples: 99842
420
+ - name: test
421
+ num_bytes: 46547
422
+ num_examples: 114
423
+ - name: validation
424
+ num_bytes: 4967
425
+ num_examples: 12
426
+ - name: dev
427
+ num_bytes: 1644
428
+ num_examples: 5
429
+ download_size: 166184960
430
+ dataset_size: 160654535
431
+ - config_name: electrical_engineering
432
+ features:
433
+ - name: question
434
+ dtype: string
435
+ - name: choices
436
+ sequence: string
437
+ - name: answer
438
+ dtype:
439
+ class_label:
440
+ names:
441
+ '0': A
442
+ '1': B
443
+ '2': C
444
+ '3': D
445
+ splits:
446
+ - name: auxiliary_train
447
+ num_bytes: 160601377
448
+ num_examples: 99842
449
+ - name: test
450
+ num_bytes: 25142
451
+ num_examples: 145
452
+ - name: validation
453
+ num_bytes: 2903
454
+ num_examples: 16
455
+ - name: dev
456
+ num_bytes: 972
457
+ num_examples: 5
458
+ download_size: 166184960
459
+ dataset_size: 160630394
460
+ - config_name: elementary_mathematics
461
+ features:
462
+ - name: question
463
+ dtype: string
464
+ - name: choices
465
+ sequence: string
466
+ - name: answer
467
+ dtype:
468
+ class_label:
469
+ names:
470
+ '0': A
471
+ '1': B
472
+ '2': C
473
+ '3': D
474
+ splits:
475
+ - name: auxiliary_train
476
+ num_bytes: 160601377
477
+ num_examples: 99842
478
+ - name: test
479
+ num_bytes: 70108
480
+ num_examples: 378
481
+ - name: validation
482
+ num_bytes: 8988
483
+ num_examples: 41
484
+ - name: dev
485
+ num_bytes: 1440
486
+ num_examples: 5
487
+ download_size: 166184960
488
+ dataset_size: 160681913
489
+ - config_name: formal_logic
490
+ features:
491
+ - name: question
492
+ dtype: string
493
+ - name: choices
494
+ sequence: string
495
+ - name: answer
496
+ dtype:
497
+ class_label:
498
+ names:
499
+ '0': A
500
+ '1': B
501
+ '2': C
502
+ '3': D
503
+ splits:
504
+ - name: auxiliary_train
505
+ num_bytes: 160601377
506
+ num_examples: 99842
507
+ - name: test
508
+ num_bytes: 49785
509
+ num_examples: 126
510
+ - name: validation
511
+ num_bytes: 6252
512
+ num_examples: 14
513
+ - name: dev
514
+ num_bytes: 1757
515
+ num_examples: 5
516
+ download_size: 166184960
517
+ dataset_size: 160659171
518
+ - config_name: global_facts
519
+ features:
520
+ - name: question
521
+ dtype: string
522
+ - name: choices
523
+ sequence: string
524
+ - name: answer
525
+ dtype:
526
+ class_label:
527
+ names:
528
+ '0': A
529
+ '1': B
530
+ '2': C
531
+ '3': D
532
+ splits:
533
+ - name: auxiliary_train
534
+ num_bytes: 160601377
535
+ num_examples: 99842
536
+ - name: test
537
+ num_bytes: 18403
538
+ num_examples: 100
539
+ - name: validation
540
+ num_bytes: 1865
541
+ num_examples: 10
542
+ - name: dev
543
+ num_bytes: 1229
544
+ num_examples: 5
545
+ download_size: 166184960
546
+ dataset_size: 160622874
547
+ - config_name: high_school_biology
548
+ features:
549
+ - name: question
550
+ dtype: string
551
+ - name: choices
552
+ sequence: string
553
+ - name: answer
554
+ dtype:
555
+ class_label:
556
+ names:
557
+ '0': A
558
+ '1': B
559
+ '2': C
560
+ '3': D
561
+ splits:
562
+ - name: auxiliary_train
563
+ num_bytes: 160601377
564
+ num_examples: 99842
565
+ - name: test
566
+ num_bytes: 109732
567
+ num_examples: 310
568
+ - name: validation
569
+ num_bytes: 11022
570
+ num_examples: 32
571
+ - name: dev
572
+ num_bytes: 1673
573
+ num_examples: 5
574
+ download_size: 166184960
575
+ dataset_size: 160723804
576
+ - config_name: high_school_chemistry
577
+ features:
578
+ - name: question
579
+ dtype: string
580
+ - name: choices
581
+ sequence: string
582
+ - name: answer
583
+ dtype:
584
+ class_label:
585
+ names:
586
+ '0': A
587
+ '1': B
588
+ '2': C
589
+ '3': D
590
+ splits:
591
+ - name: auxiliary_train
592
+ num_bytes: 160601377
593
+ num_examples: 99842
594
+ - name: test
595
+ num_bytes: 58464
596
+ num_examples: 203
597
+ - name: validation
598
+ num_bytes: 7092
599
+ num_examples: 22
600
+ - name: dev
601
+ num_bytes: 1220
602
+ num_examples: 5
603
+ download_size: 166184960
604
+ dataset_size: 160668153
605
+ - config_name: high_school_computer_science
606
+ features:
607
+ - name: question
608
+ dtype: string
609
+ - name: choices
610
+ sequence: string
611
+ - name: answer
612
+ dtype:
613
+ class_label:
614
+ names:
615
+ '0': A
616
+ '1': B
617
+ '2': C
618
+ '3': D
619
+ splits:
620
+ - name: auxiliary_train
621
+ num_bytes: 160601377
622
+ num_examples: 99842
623
+ - name: test
624
+ num_bytes: 44476
625
+ num_examples: 100
626
+ - name: validation
627
+ num_bytes: 3343
628
+ num_examples: 9
629
+ - name: dev
630
+ num_bytes: 2918
631
+ num_examples: 5
632
+ download_size: 166184960
633
+ dataset_size: 160652114
634
+ - config_name: high_school_european_history
635
+ features:
636
+ - name: question
637
+ dtype: string
638
+ - name: choices
639
+ sequence: string
640
+ - name: answer
641
+ dtype:
642
+ class_label:
643
+ names:
644
+ '0': A
645
+ '1': B
646
+ '2': C
647
+ '3': D
648
+ splits:
649
+ - name: auxiliary_train
650
+ num_bytes: 160601377
651
+ num_examples: 99842
652
+ - name: test
653
+ num_bytes: 270300
654
+ num_examples: 165
655
+ - name: validation
656
+ num_bytes: 29632
657
+ num_examples: 18
658
+ - name: dev
659
+ num_bytes: 11564
660
+ num_examples: 5
661
+ download_size: 166184960
662
+ dataset_size: 160912873
663
+ - config_name: high_school_geography
664
+ features:
665
+ - name: question
666
+ dtype: string
667
+ - name: choices
668
+ sequence: string
669
+ - name: answer
670
+ dtype:
671
+ class_label:
672
+ names:
673
+ '0': A
674
+ '1': B
675
+ '2': C
676
+ '3': D
677
+ splits:
678
+ - name: auxiliary_train
679
+ num_bytes: 160601377
680
+ num_examples: 99842
681
+ - name: test
682
+ num_bytes: 42034
683
+ num_examples: 198
684
+ - name: validation
685
+ num_bytes: 4332
686
+ num_examples: 22
687
+ - name: dev
688
+ num_bytes: 1403
689
+ num_examples: 5
690
+ download_size: 166184960
691
+ dataset_size: 160649146
692
+ - config_name: high_school_government_and_politics
693
+ features:
694
+ - name: question
695
+ dtype: string
696
+ - name: choices
697
+ sequence: string
698
+ - name: answer
699
+ dtype:
700
+ class_label:
701
+ names:
702
+ '0': A
703
+ '1': B
704
+ '2': C
705
+ '3': D
706
+ splits:
707
+ - name: auxiliary_train
708
+ num_bytes: 160601377
709
+ num_examples: 99842
710
+ - name: test
711
+ num_bytes: 66074
712
+ num_examples: 193
713
+ - name: validation
714
+ num_bytes: 7063
715
+ num_examples: 21
716
+ - name: dev
717
+ num_bytes: 1779
718
+ num_examples: 5
719
+ download_size: 166184960
720
+ dataset_size: 160676293
721
+ - config_name: high_school_macroeconomics
722
+ features:
723
+ - name: question
724
+ dtype: string
725
+ - name: choices
726
+ sequence: string
727
+ - name: answer
728
+ dtype:
729
+ class_label:
730
+ names:
731
+ '0': A
732
+ '1': B
733
+ '2': C
734
+ '3': D
735
+ splits:
736
+ - name: auxiliary_train
737
+ num_bytes: 160601377
738
+ num_examples: 99842
739
+ - name: test
740
+ num_bytes: 117687
741
+ num_examples: 390
742
+ - name: validation
743
+ num_bytes: 13020
744
+ num_examples: 43
745
+ - name: dev
746
+ num_bytes: 1328
747
+ num_examples: 5
748
+ download_size: 166184960
749
+ dataset_size: 160733412
750
+ - config_name: high_school_mathematics
751
+ features:
752
+ - name: question
753
+ dtype: string
754
+ - name: choices
755
+ sequence: string
756
+ - name: answer
757
+ dtype:
758
+ class_label:
759
+ names:
760
+ '0': A
761
+ '1': B
762
+ '2': C
763
+ '3': D
764
+ splits:
765
+ - name: auxiliary_train
766
+ num_bytes: 160601377
767
+ num_examples: 99842
768
+ - name: test
769
+ num_bytes: 54854
770
+ num_examples: 270
771
+ - name: validation
772
+ num_bytes: 5765
773
+ num_examples: 29
774
+ - name: dev
775
+ num_bytes: 1297
776
+ num_examples: 5
777
+ download_size: 166184960
778
+ dataset_size: 160663293
779
+ - config_name: high_school_microeconomics
780
+ features:
781
+ - name: question
782
+ dtype: string
783
+ - name: choices
784
+ sequence: string
785
+ - name: answer
786
+ dtype:
787
+ class_label:
788
+ names:
789
+ '0': A
790
+ '1': B
791
+ '2': C
792
+ '3': D
793
+ splits:
794
+ - name: auxiliary_train
795
+ num_bytes: 160601377
796
+ num_examples: 99842
797
+ - name: test
798
+ num_bytes: 75703
799
+ num_examples: 238
800
+ - name: validation
801
+ num_bytes: 7553
802
+ num_examples: 26
803
+ - name: dev
804
+ num_bytes: 1298
805
+ num_examples: 5
806
+ download_size: 166184960
807
+ dataset_size: 160685931
808
+ - config_name: high_school_physics
809
+ features:
810
+ - name: question
811
+ dtype: string
812
+ - name: choices
813
+ sequence: string
814
+ - name: answer
815
+ dtype:
816
+ class_label:
817
+ names:
818
+ '0': A
819
+ '1': B
820
+ '2': C
821
+ '3': D
822
+ splits:
823
+ - name: auxiliary_train
824
+ num_bytes: 160601377
825
+ num_examples: 99842
826
+ - name: test
827
+ num_bytes: 59538
828
+ num_examples: 151
829
+ - name: validation
830
+ num_bytes: 6771
831
+ num_examples: 17
832
+ - name: dev
833
+ num_bytes: 1489
834
+ num_examples: 5
835
+ download_size: 166184960
836
+ dataset_size: 160669175
837
+ - config_name: high_school_psychology
838
+ features:
839
+ - name: question
840
+ dtype: string
841
+ - name: choices
842
+ sequence: string
843
+ - name: answer
844
+ dtype:
845
+ class_label:
846
+ names:
847
+ '0': A
848
+ '1': B
849
+ '2': C
850
+ '3': D
851
+ splits:
852
+ - name: auxiliary_train
853
+ num_bytes: 160601377
854
+ num_examples: 99842
855
+ - name: test
856
+ num_bytes: 159407
857
+ num_examples: 545
858
+ - name: validation
859
+ num_bytes: 17269
860
+ num_examples: 60
861
+ - name: dev
862
+ num_bytes: 1905
863
+ num_examples: 5
864
+ download_size: 166184960
865
+ dataset_size: 160779958
866
+ - config_name: high_school_statistics
867
+ features:
868
+ - name: question
869
+ dtype: string
870
+ - name: choices
871
+ sequence: string
872
+ - name: answer
873
+ dtype:
874
+ class_label:
875
+ names:
876
+ '0': A
877
+ '1': B
878
+ '2': C
879
+ '3': D
880
+ splits:
881
+ - name: auxiliary_train
882
+ num_bytes: 160601377
883
+ num_examples: 99842
884
+ - name: test
885
+ num_bytes: 110702
886
+ num_examples: 216
887
+ - name: validation
888
+ num_bytes: 9997
889
+ num_examples: 23
890
+ - name: dev
891
+ num_bytes: 2528
892
+ num_examples: 5
893
+ download_size: 166184960
894
+ dataset_size: 160724604
895
+ - config_name: high_school_us_history
896
+ features:
897
+ - name: question
898
+ dtype: string
899
+ - name: choices
900
+ sequence: string
901
+ - name: answer
902
+ dtype:
903
+ class_label:
904
+ names:
905
+ '0': A
906
+ '1': B
907
+ '2': C
908
+ '3': D
909
+ splits:
910
+ - name: auxiliary_train
911
+ num_bytes: 160601377
912
+ num_examples: 99842
913
+ - name: test
914
+ num_bytes: 296734
915
+ num_examples: 204
916
+ - name: validation
917
+ num_bytes: 31706
918
+ num_examples: 22
919
+ - name: dev
920
+ num_bytes: 8864
921
+ num_examples: 5
922
+ download_size: 166184960
923
+ dataset_size: 160938681
924
+ - config_name: high_school_world_history
925
+ features:
926
+ - name: question
927
+ dtype: string
928
+ - name: choices
929
+ sequence: string
930
+ - name: answer
931
+ dtype:
932
+ class_label:
933
+ names:
934
+ '0': A
935
+ '1': B
936
+ '2': C
937
+ '3': D
938
+ splits:
939
+ - name: auxiliary_train
940
+ num_bytes: 160601377
941
+ num_examples: 99842
942
+ - name: test
943
+ num_bytes: 378617
944
+ num_examples: 237
945
+ - name: validation
946
+ num_bytes: 45501
947
+ num_examples: 26
948
+ - name: dev
949
+ num_bytes: 4882
950
+ num_examples: 5
951
+ download_size: 166184960
952
+ dataset_size: 161030377
953
+ - config_name: human_aging
954
+ features:
955
+ - name: question
956
+ dtype: string
957
+ - name: choices
958
+ sequence: string
959
+ - name: answer
960
+ dtype:
961
+ class_label:
962
+ names:
963
+ '0': A
964
+ '1': B
965
+ '2': C
966
+ '3': D
967
+ splits:
968
+ - name: auxiliary_train
969
+ num_bytes: 160601377
970
+ num_examples: 99842
971
+ - name: test
972
+ num_bytes: 46098
973
+ num_examples: 223
974
+ - name: validation
975
+ num_bytes: 4707
976
+ num_examples: 23
977
+ - name: dev
978
+ num_bytes: 1008
979
+ num_examples: 5
980
+ download_size: 166184960
981
+ dataset_size: 160653190
982
+ - config_name: human_sexuality
983
+ features:
984
+ - name: question
985
+ dtype: string
986
+ - name: choices
987
+ sequence: string
988
+ - name: answer
989
+ dtype:
990
+ class_label:
991
+ names:
992
+ '0': A
993
+ '1': B
994
+ '2': C
995
+ '3': D
996
+ splits:
997
+ - name: auxiliary_train
998
+ num_bytes: 160601377
999
+ num_examples: 99842
1000
+ - name: test
1001
+ num_bytes: 32110
1002
+ num_examples: 131
1003
+ - name: validation
1004
+ num_bytes: 2421
1005
+ num_examples: 12
1006
+ - name: dev
1007
+ num_bytes: 1077
1008
+ num_examples: 5
1009
+ download_size: 166184960
1010
+ dataset_size: 160636985
1011
+ - config_name: international_law
1012
+ features:
1013
+ - name: question
1014
+ dtype: string
1015
+ - name: choices
1016
+ sequence: string
1017
+ - name: answer
1018
+ dtype:
1019
+ class_label:
1020
+ names:
1021
+ '0': A
1022
+ '1': B
1023
+ '2': C
1024
+ '3': D
1025
+ splits:
1026
+ - name: auxiliary_train
1027
+ num_bytes: 160601377
1028
+ num_examples: 99842
1029
+ - name: test
1030
+ num_bytes: 53531
1031
+ num_examples: 121
1032
+ - name: validation
1033
+ num_bytes: 6473
1034
+ num_examples: 13
1035
+ - name: dev
1036
+ num_bytes: 2418
1037
+ num_examples: 5
1038
+ download_size: 166184960
1039
+ dataset_size: 160663799
1040
+ - config_name: jurisprudence
1041
+ features:
1042
+ - name: question
1043
+ dtype: string
1044
+ - name: choices
1045
+ sequence: string
1046
+ - name: answer
1047
+ dtype:
1048
+ class_label:
1049
+ names:
1050
+ '0': A
1051
+ '1': B
1052
+ '2': C
1053
+ '3': D
1054
+ splits:
1055
+ - name: auxiliary_train
1056
+ num_bytes: 160601377
1057
+ num_examples: 99842
1058
+ - name: test
1059
+ num_bytes: 33986
1060
+ num_examples: 108
1061
+ - name: validation
1062
+ num_bytes: 3729
1063
+ num_examples: 11
1064
+ - name: dev
1065
+ num_bytes: 1303
1066
+ num_examples: 5
1067
+ download_size: 166184960
1068
+ dataset_size: 160640395
1069
+ - config_name: logical_fallacies
1070
+ features:
1071
+ - name: question
1072
+ dtype: string
1073
+ - name: choices
1074
+ sequence: string
1075
+ - name: answer
1076
+ dtype:
1077
+ class_label:
1078
+ names:
1079
+ '0': A
1080
+ '1': B
1081
+ '2': C
1082
+ '3': D
1083
+ splits:
1084
+ - name: auxiliary_train
1085
+ num_bytes: 160601377
1086
+ num_examples: 99842
1087
+ - name: test
1088
+ num_bytes: 50117
1089
+ num_examples: 163
1090
+ - name: validation
1091
+ num_bytes: 5103
1092
+ num_examples: 18
1093
+ - name: dev
1094
+ num_bytes: 1573
1095
+ num_examples: 5
1096
+ download_size: 166184960
1097
+ dataset_size: 160658170
1098
+ - config_name: machine_learning
1099
+ features:
1100
+ - name: question
1101
+ dtype: string
1102
+ - name: choices
1103
+ sequence: string
1104
+ - name: answer
1105
+ dtype:
1106
+ class_label:
1107
+ names:
1108
+ '0': A
1109
+ '1': B
1110
+ '2': C
1111
+ '3': D
1112
+ splits:
1113
+ - name: auxiliary_train
1114
+ num_bytes: 160601377
1115
+ num_examples: 99842
1116
+ - name: test
1117
+ num_bytes: 33880
1118
+ num_examples: 112
1119
+ - name: validation
1120
+ num_bytes: 3232
1121
+ num_examples: 11
1122
+ - name: dev
1123
+ num_bytes: 2323
1124
+ num_examples: 5
1125
+ download_size: 166184960
1126
+ dataset_size: 160640812
1127
+ - config_name: management
1128
+ features:
1129
+ - name: question
1130
+ dtype: string
1131
+ - name: choices
1132
+ sequence: string
1133
+ - name: answer
1134
+ dtype:
1135
+ class_label:
1136
+ names:
1137
+ '0': A
1138
+ '1': B
1139
+ '2': C
1140
+ '3': D
1141
+ splits:
1142
+ - name: auxiliary_train
1143
+ num_bytes: 160601377
1144
+ num_examples: 99842
1145
+ - name: test
1146
+ num_bytes: 20002
1147
+ num_examples: 103
1148
+ - name: validation
1149
+ num_bytes: 1820
1150
+ num_examples: 11
1151
+ - name: dev
1152
+ num_bytes: 898
1153
+ num_examples: 5
1154
+ download_size: 166184960
1155
+ dataset_size: 160624097
1156
+ - config_name: marketing
1157
+ features:
1158
+ - name: question
1159
+ dtype: string
1160
+ - name: choices
1161
+ sequence: string
1162
+ - name: answer
1163
+ dtype:
1164
+ class_label:
1165
+ names:
1166
+ '0': A
1167
+ '1': B
1168
+ '2': C
1169
+ '3': D
1170
+ splits:
1171
+ - name: auxiliary_train
1172
+ num_bytes: 160601377
1173
+ num_examples: 99842
1174
+ - name: test
1175
+ num_bytes: 63025
1176
+ num_examples: 234
1177
+ - name: validation
1178
+ num_bytes: 7394
1179
+ num_examples: 25
1180
+ - name: dev
1181
+ num_bytes: 1481
1182
+ num_examples: 5
1183
+ download_size: 166184960
1184
+ dataset_size: 160673277
1185
+ - config_name: medical_genetics
1186
+ features:
1187
+ - name: question
1188
+ dtype: string
1189
+ - name: choices
1190
+ sequence: string
1191
+ - name: answer
1192
+ dtype:
1193
+ class_label:
1194
+ names:
1195
+ '0': A
1196
+ '1': B
1197
+ '2': C
1198
+ '3': D
1199
+ splits:
1200
+ - name: auxiliary_train
1201
+ num_bytes: 160601377
1202
+ num_examples: 99842
1203
+ - name: test
1204
+ num_bytes: 20864
1205
+ num_examples: 100
1206
+ - name: validation
1207
+ num_bytes: 3005
1208
+ num_examples: 11
1209
+ - name: dev
1210
+ num_bytes: 1089
1211
+ num_examples: 5
1212
+ download_size: 166184960
1213
+ dataset_size: 160626335
1214
+ - config_name: miscellaneous
1215
+ features:
1216
+ - name: question
1217
+ dtype: string
1218
+ - name: choices
1219
+ sequence: string
1220
+ - name: answer
1221
+ dtype:
1222
+ class_label:
1223
+ names:
1224
+ '0': A
1225
+ '1': B
1226
+ '2': C
1227
+ '3': D
1228
+ splits:
1229
+ - name: auxiliary_train
1230
+ num_bytes: 160601377
1231
+ num_examples: 99842
1232
+ - name: test
1233
+ num_bytes: 147704
1234
+ num_examples: 783
1235
+ - name: validation
1236
+ num_bytes: 14330
1237
+ num_examples: 86
1238
+ - name: dev
1239
+ num_bytes: 699
1240
+ num_examples: 5
1241
+ download_size: 166184960
1242
+ dataset_size: 160764110
1243
+ - config_name: moral_disputes
1244
+ features:
1245
+ - name: question
1246
+ dtype: string
1247
+ - name: choices
1248
+ sequence: string
1249
+ - name: answer
1250
+ dtype:
1251
+ class_label:
1252
+ names:
1253
+ '0': A
1254
+ '1': B
1255
+ '2': C
1256
+ '3': D
1257
+ splits:
1258
+ - name: auxiliary_train
1259
+ num_bytes: 160601377
1260
+ num_examples: 99842
1261
+ - name: test
1262
+ num_bytes: 107818
1263
+ num_examples: 346
1264
+ - name: validation
1265
+ num_bytes: 12420
1266
+ num_examples: 38
1267
+ - name: dev
1268
+ num_bytes: 1755
1269
+ num_examples: 5
1270
+ download_size: 166184960
1271
+ dataset_size: 160723370
1272
+ - config_name: moral_scenarios
1273
+ features:
1274
+ - name: question
1275
+ dtype: string
1276
+ - name: choices
1277
+ sequence: string
1278
+ - name: answer
1279
+ dtype:
1280
+ class_label:
1281
+ names:
1282
+ '0': A
1283
+ '1': B
1284
+ '2': C
1285
+ '3': D
1286
+ splits:
1287
+ - name: auxiliary_train
1288
+ num_bytes: 160601377
1289
+ num_examples: 99842
1290
+ - name: test
1291
+ num_bytes: 374026
1292
+ num_examples: 895
1293
+ - name: validation
1294
+ num_bytes: 42338
1295
+ num_examples: 100
1296
+ - name: dev
1297
+ num_bytes: 2058
1298
+ num_examples: 5
1299
+ download_size: 166184960
1300
+ dataset_size: 161019799
1301
+ - config_name: nutrition
1302
+ features:
1303
+ - name: question
1304
+ dtype: string
1305
+ - name: choices
1306
+ sequence: string
1307
+ - name: answer
1308
+ dtype:
1309
+ class_label:
1310
+ names:
1311
+ '0': A
1312
+ '1': B
1313
+ '2': C
1314
+ '3': D
1315
+ splits:
1316
+ - name: auxiliary_train
1317
+ num_bytes: 160601377
1318
+ num_examples: 99842
1319
+ - name: test
1320
+ num_bytes: 92410
1321
+ num_examples: 306
1322
+ - name: validation
1323
+ num_bytes: 8436
1324
+ num_examples: 33
1325
+ - name: dev
1326
+ num_bytes: 2085
1327
+ num_examples: 5
1328
+ download_size: 166184960
1329
+ dataset_size: 160704308
1330
+ - config_name: philosophy
1331
+ features:
1332
+ - name: question
1333
+ dtype: string
1334
+ - name: choices
1335
+ sequence: string
1336
+ - name: answer
1337
+ dtype:
1338
+ class_label:
1339
+ names:
1340
+ '0': A
1341
+ '1': B
1342
+ '2': C
1343
+ '3': D
1344
+ splits:
1345
+ - name: auxiliary_train
1346
+ num_bytes: 160601377
1347
+ num_examples: 99842
1348
+ - name: test
1349
+ num_bytes: 80073
1350
+ num_examples: 311
1351
+ - name: validation
1352
+ num_bytes: 9184
1353
+ num_examples: 34
1354
+ - name: dev
1355
+ num_bytes: 988
1356
+ num_examples: 5
1357
+ download_size: 166184960
1358
+ dataset_size: 160691622
1359
+ - config_name: prehistory
1360
+ features:
1361
+ - name: question
1362
+ dtype: string
1363
+ - name: choices
1364
+ sequence: string
1365
+ - name: answer
1366
+ dtype:
1367
+ class_label:
1368
+ names:
1369
+ '0': A
1370
+ '1': B
1371
+ '2': C
1372
+ '3': D
1373
+ splits:
1374
+ - name: auxiliary_train
1375
+ num_bytes: 160601377
1376
+ num_examples: 99842
1377
+ - name: test
1378
+ num_bytes: 89594
1379
+ num_examples: 324
1380
+ - name: validation
1381
+ num_bytes: 10285
1382
+ num_examples: 35
1383
+ - name: dev
1384
+ num_bytes: 1878
1385
+ num_examples: 5
1386
+ download_size: 166184960
1387
+ dataset_size: 160703134
1388
+ - config_name: professional_accounting
1389
+ features:
1390
+ - name: question
1391
+ dtype: string
1392
+ - name: choices
1393
+ sequence: string
1394
+ - name: answer
1395
+ dtype:
1396
+ class_label:
1397
+ names:
1398
+ '0': A
1399
+ '1': B
1400
+ '2': C
1401
+ '3': D
1402
+ splits:
1403
+ - name: auxiliary_train
1404
+ num_bytes: 160601377
1405
+ num_examples: 99842
1406
+ - name: test
1407
+ num_bytes: 124550
1408
+ num_examples: 282
1409
+ - name: validation
1410
+ num_bytes: 14372
1411
+ num_examples: 31
1412
+ - name: dev
1413
+ num_bytes: 2148
1414
+ num_examples: 5
1415
+ download_size: 166184960
1416
+ dataset_size: 160742447
1417
+ - config_name: professional_law
1418
+ features:
1419
+ - name: question
1420
+ dtype: string
1421
+ - name: choices
1422
+ sequence: string
1423
+ - name: answer
1424
+ dtype:
1425
+ class_label:
1426
+ names:
1427
+ '0': A
1428
+ '1': B
1429
+ '2': C
1430
+ '3': D
1431
+ splits:
1432
+ - name: auxiliary_train
1433
+ num_bytes: 160601377
1434
+ num_examples: 99842
1435
+ - name: test
1436
+ num_bytes: 1891762
1437
+ num_examples: 1534
1438
+ - name: validation
1439
+ num_bytes: 203519
1440
+ num_examples: 170
1441
+ - name: dev
1442
+ num_bytes: 6610
1443
+ num_examples: 5
1444
+ download_size: 166184960
1445
+ dataset_size: 162703268
1446
+ - config_name: professional_medicine
1447
+ features:
1448
+ - name: question
1449
+ dtype: string
1450
+ - name: choices
1451
+ sequence: string
1452
+ - name: answer
1453
+ dtype:
1454
+ class_label:
1455
+ names:
1456
+ '0': A
1457
+ '1': B
1458
+ '2': C
1459
+ '3': D
1460
+ splits:
1461
+ - name: auxiliary_train
1462
+ num_bytes: 160601377
1463
+ num_examples: 99842
1464
+ - name: test
1465
+ num_bytes: 217561
1466
+ num_examples: 272
1467
+ - name: validation
1468
+ num_bytes: 23847
1469
+ num_examples: 31
1470
+ - name: dev
1471
+ num_bytes: 3807
1472
+ num_examples: 5
1473
+ download_size: 166184960
1474
+ dataset_size: 160846592
1475
+ - config_name: professional_psychology
1476
+ features:
1477
+ - name: question
1478
+ dtype: string
1479
+ - name: choices
1480
+ sequence: string
1481
+ - name: answer
1482
+ dtype:
1483
+ class_label:
1484
+ names:
1485
+ '0': A
1486
+ '1': B
1487
+ '2': C
1488
+ '3': D
1489
+ splits:
1490
+ - name: auxiliary_train
1491
+ num_bytes: 160601377
1492
+ num_examples: 99842
1493
+ - name: test
1494
+ num_bytes: 225899
1495
+ num_examples: 612
1496
+ - name: validation
1497
+ num_bytes: 29101
1498
+ num_examples: 69
1499
+ - name: dev
1500
+ num_bytes: 2267
1501
+ num_examples: 5
1502
+ download_size: 166184960
1503
+ dataset_size: 160858644
1504
+ - config_name: public_relations
1505
+ features:
1506
+ - name: question
1507
+ dtype: string
1508
+ - name: choices
1509
+ sequence: string
1510
+ - name: answer
1511
+ dtype:
1512
+ class_label:
1513
+ names:
1514
+ '0': A
1515
+ '1': B
1516
+ '2': C
1517
+ '3': D
1518
+ splits:
1519
+ - name: auxiliary_train
1520
+ num_bytes: 160601377
1521
+ num_examples: 99842
1522
+ - name: test
1523
+ num_bytes: 28760
1524
+ num_examples: 110
1525
+ - name: validation
1526
+ num_bytes: 4566
1527
+ num_examples: 12
1528
+ - name: dev
1529
+ num_bytes: 1496
1530
+ num_examples: 5
1531
+ download_size: 166184960
1532
+ dataset_size: 160636199
1533
+ - config_name: security_studies
1534
+ features:
1535
+ - name: question
1536
+ dtype: string
1537
+ - name: choices
1538
+ sequence: string
1539
+ - name: answer
1540
+ dtype:
1541
+ class_label:
1542
+ names:
1543
+ '0': A
1544
+ '1': B
1545
+ '2': C
1546
+ '3': D
1547
+ splits:
1548
+ - name: auxiliary_train
1549
+ num_bytes: 160601377
1550
+ num_examples: 99842
1551
+ - name: test
1552
+ num_bytes: 204844
1553
+ num_examples: 245
1554
+ - name: validation
1555
+ num_bytes: 22637
1556
+ num_examples: 27
1557
+ - name: dev
1558
+ num_bytes: 5335
1559
+ num_examples: 5
1560
+ download_size: 166184960
1561
+ dataset_size: 160834193
1562
+ - config_name: sociology
1563
+ features:
1564
+ - name: question
1565
+ dtype: string
1566
+ - name: choices
1567
+ sequence: string
1568
+ - name: answer
1569
+ dtype:
1570
+ class_label:
1571
+ names:
1572
+ '0': A
1573
+ '1': B
1574
+ '2': C
1575
+ '3': D
1576
+ splits:
1577
+ - name: auxiliary_train
1578
+ num_bytes: 160601377
1579
+ num_examples: 99842
1580
+ - name: test
1581
+ num_bytes: 66243
1582
+ num_examples: 201
1583
+ - name: validation
1584
+ num_bytes: 7184
1585
+ num_examples: 22
1586
+ - name: dev
1587
+ num_bytes: 1613
1588
+ num_examples: 5
1589
+ download_size: 166184960
1590
+ dataset_size: 160676417
1591
+ - config_name: us_foreign_policy
1592
+ features:
1593
+ - name: question
1594
+ dtype: string
1595
+ - name: choices
1596
+ sequence: string
1597
+ - name: answer
1598
+ dtype:
1599
+ class_label:
1600
+ names:
1601
+ '0': A
1602
+ '1': B
1603
+ '2': C
1604
+ '3': D
1605
+ splits:
1606
+ - name: auxiliary_train
1607
+ num_bytes: 160601377
1608
+ num_examples: 99842
1609
+ - name: test
1610
+ num_bytes: 28443
1611
+ num_examples: 100
1612
+ - name: validation
1613
+ num_bytes: 3264
1614
+ num_examples: 11
1615
+ - name: dev
1616
+ num_bytes: 1611
1617
+ num_examples: 5
1618
+ download_size: 166184960
1619
+ dataset_size: 160634695
1620
+ - config_name: virology
1621
+ features:
1622
+ - name: question
1623
+ dtype: string
1624
+ - name: choices
1625
+ sequence: string
1626
+ - name: answer
1627
+ dtype:
1628
+ class_label:
1629
+ names:
1630
+ '0': A
1631
+ '1': B
1632
+ '2': C
1633
+ '3': D
1634
+ splits:
1635
+ - name: auxiliary_train
1636
+ num_bytes: 160601377
1637
+ num_examples: 99842
1638
+ - name: test
1639
+ num_bytes: 38759
1640
+ num_examples: 166
1641
+ - name: validation
1642
+ num_bytes: 5463
1643
+ num_examples: 18
1644
+ - name: dev
1645
+ num_bytes: 1096
1646
+ num_examples: 5
1647
+ download_size: 166184960
1648
+ dataset_size: 160646695
1649
+ - config_name: world_religions
1650
+ features:
1651
+ - name: question
1652
+ dtype: string
1653
+ - name: choices
1654
+ sequence: string
1655
+ - name: answer
1656
+ dtype:
1657
+ class_label:
1658
+ names:
1659
+ '0': A
1660
+ '1': B
1661
+ '2': C
1662
+ '3': D
1663
+ splits:
1664
+ - name: auxiliary_train
1665
+ num_bytes: 160601377
1666
+ num_examples: 99842
1667
+ - name: test
1668
+ num_bytes: 25274
1669
+ num_examples: 171
1670
+ - name: validation
1671
+ num_bytes: 2765
1672
+ num_examples: 19
1673
+ - name: dev
1674
+ num_bytes: 670
1675
+ num_examples: 5
1676
+ download_size: 166184960
1677
+ dataset_size: 160630086
1678
+ ---
1679
+
1680
+ # Dataset Card for MMLU
1681
+
1682
+ ## Table of Contents
1683
+ - [Table of Contents](#table-of-contents)
1684
+ - [Dataset Description](#dataset-description)
1685
+ - [Dataset Summary](#dataset-summary)
1686
+ - [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
1687
+ - [Languages](#languages)
1688
+ - [Dataset Structure](#dataset-structure)
1689
+ - [Data Instances](#data-instances)
1690
+ - [Data Fields](#data-fields)
1691
+ - [Data Splits](#data-splits)
1692
+ - [Dataset Creation](#dataset-creation)
1693
+ - [Curation Rationale](#curation-rationale)
1694
+ - [Source Data](#source-data)
1695
+ - [Annotations](#annotations)
1696
+ - [Personal and Sensitive Information](#personal-and-sensitive-information)
1697
+ - [Considerations for Using the Data](#considerations-for-using-the-data)
1698
+ - [Social Impact of Dataset](#social-impact-of-dataset)
1699
+ - [Discussion of Biases](#discussion-of-biases)
1700
+ - [Other Known Limitations](#other-known-limitations)
1701
+ - [Additional Information](#additional-information)
1702
+ - [Dataset Curators](#dataset-curators)
1703
+ - [Licensing Information](#licensing-information)
1704
+ - [Citation Information](#citation-information)
1705
+ - [Contributions](#contributions)
1706
+
1707
+ ## Dataset Description
1708
+
1709
+ - **Repository**: https://github.com/hendrycks/test
1710
+ - **Paper**: https://arxiv.org/abs/2009.03300
1711
+
1712
+ ### Dataset Summary
1713
+
1714
+ [Measuring Massive Multitask Language Understanding](https://arxiv.org/pdf/2009.03300) by [Dan Hendrycks](https://people.eecs.berkeley.edu/~hendrycks/), [Collin Burns](http://collinpburns.com), [Steven Basart](https://stevenbas.art), Andy Zou, Mantas Mazeika, [Dawn Song](https://people.eecs.berkeley.edu/~dawnsong/), and [Jacob Steinhardt](https://www.stat.berkeley.edu/~jsteinhardt/) (ICLR 2021).
1715
+
1716
+ This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge. The test spans subjects in the humanities, social sciences, hard sciences, and other areas that are important for some people to learn. This covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability.
1717
+
1718
+ A complete list of tasks: ['abstract_algebra', 'anatomy', 'astronomy', 'business_ethics', 'clinical_knowledge', 'college_biology', 'college_chemistry', 'college_computer_science', 'college_mathematics', 'college_medicine', 'college_physics', 'computer_security', 'conceptual_physics', 'econometrics', 'electrical_engineering', 'elementary_mathematics', 'formal_logic', 'global_facts', 'high_school_biology', 'high_school_chemistry', 'high_school_computer_science', 'high_school_european_history', 'high_school_geography', 'high_school_government_and_politics', 'high_school_macroeconomics', 'high_school_mathematics', 'high_school_microeconomics', 'high_school_physics', 'high_school_psychology', 'high_school_statistics', 'high_school_us_history', 'high_school_world_history', 'human_aging', 'human_sexuality', 'international_law', 'jurisprudence', 'logical_fallacies', 'machine_learning', 'management', 'marketing', 'medical_genetics', 'miscellaneous', 'moral_disputes', 'moral_scenarios', 'nutrition', 'philosophy', 'prehistory', 'professional_accounting', 'professional_law', 'professional_medicine', 'professional_psychology', 'public_relations', 'security_studies', 'sociology', 'us_foreign_policy', 'virology', 'world_religions']
1719
+
1720
+ ### Supported Tasks and Leaderboards
1721
+
1722
+ | Model | Authors | Humanities | Social Science | STEM | Other | Average |
1723
+ |------------------------------------|----------|:-------:|:-------:|:-------:|:-------:|:-------:|
1724
+ | [UnifiedQA](https://arxiv.org/abs/2005.00700) | Khashabi et al., 2020 | 45.6 | 56.6 | 40.2 | 54.6 | 48.9
1725
+ | [GPT-3](https://arxiv.org/abs/2005.14165) (few-shot) | Brown et al., 2020 | 40.8 | 50.4 | 36.7 | 48.8 | 43.9
1726
+ | [GPT-2](https://arxiv.org/abs/2005.14165) | Radford et al., 2019 | 32.8 | 33.3 | 30.2 | 33.1 | 32.4
1727
+ | Random Baseline | N/A | 25.0 | 25.0 | 25.0 | 25.0 | 25.0 | 25.0
1728
+
1729
+ ### Languages
1730
+
1731
+ English
1732
+
1733
+ ## Dataset Structure
1734
+
1735
+ ### Data Instances
1736
+
1737
+ An example from anatomy subtask looks as follows:
1738
+ ```
1739
+ {
1740
+ "question": "What is the embryological origin of the hyoid bone?",
1741
+ "choices": ["The first pharyngeal arch", "The first and second pharyngeal arches", "The second pharyngeal arch", "The second and third pharyngeal arches"],
1742
+ "answer": "D"
1743
+ }
1744
+ ```
1745
+
1746
+ ### Data Fields
1747
+
1748
+ - `question`: a string feature
1749
+ - `choices`: a list of 4 string features
1750
+ - `answer`: a ClassLabel feature
1751
+
1752
+ ### Data Splits
1753
+
1754
+ - `auxiliary_train`: auxiliary multiple-choice training questions from ARC, MC_TEST, OBQA, RACE, etc.
1755
+ - `dev`: 5 examples per subtask, meant for few-shot setting
1756
+ - `test`: there are at least 100 examples per subtask
1757
+
1758
+ | | auxiliary_train | dev | val | test |
1759
+ | ----- | :------: | :-----: | :-----: | :-----: |
1760
+ | TOTAL | 99842 | 285 | 1531 | 14042
1761
+
1762
+ ## Dataset Creation
1763
+
1764
+ ### Curation Rationale
1765
+
1766
+ Transformer models have driven this recent progress by pretraining on massive text corpora, including all of Wikipedia, thousands of books, and numerous websites. These models consequently see extensive information about specialized topics, most of which is not assessed by existing NLP benchmarks. To bridge the gap between the wide-ranging knowledge that models see during pretraining and the existing measures of success, we introduce a new benchmark for assessing models across a diverse set of subjects that humans learn.
1767
+
1768
+ ### Source Data
1769
+
1770
+ #### Initial Data Collection and Normalization
1771
+
1772
+ [More Information Needed]
1773
+
1774
+ #### Who are the source language producers?
1775
+
1776
+ [More Information Needed]
1777
+
1778
+ ### Annotations
1779
+
1780
+ #### Annotation process
1781
+
1782
+ [More Information Needed]
1783
+
1784
+ #### Who are the annotators?
1785
+
1786
+ [More Information Needed]
1787
+
1788
+ ### Personal and Sensitive Information
1789
+
1790
+ [More Information Needed]
1791
+
1792
+ ## Considerations for Using the Data
1793
+
1794
+ ### Social Impact of Dataset
1795
+
1796
+ [More Information Needed]
1797
+
1798
+ ### Discussion of Biases
1799
+
1800
+ [More Information Needed]
1801
+
1802
+ ### Other Known Limitations
1803
+
1804
+ [More Information Needed]
1805
+
1806
+ ## Additional Information
1807
+
1808
+ ### Dataset Curators
1809
+
1810
+ [More Information Needed]
1811
+
1812
+ ### Licensing Information
1813
+
1814
+ [MIT License](https://github.com/hendrycks/test/blob/master/LICENSE)
1815
+
1816
+ ### Citation Information
1817
+
1818
+ If you find this useful in your research, please consider citing the test and also the [ETHICS](https://arxiv.org/abs/2008.02275) dataset it draws from:
1819
+ ```
1820
+ @article{hendryckstest2021,
1821
+ title={Measuring Massive Multitask Language Understanding},
1822
+ author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},
1823
+ journal={Proceedings of the International Conference on Learning Representations (ICLR)},
1824
+ year={2021}
1825
+ }
1826
+
1827
+ @article{hendrycks2021ethics,
1828
+ title={Aligning AI With Shared Human Values},
1829
+ author={Dan Hendrycks and Collin Burns and Steven Basart and Andrew Critch and Jerry Li and Dawn Song and Jacob Steinhardt},
1830
+ journal={Proceedings of the International Conference on Learning Representations (ICLR)},
1831
+ year={2021}
1832
+ }
1833
+ ```
1834
+ ### Contributions
1835
+
1836
+ Thanks to [@andyzoujm](https://github.com/andyzoujm) for adding this dataset.
data.tar ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bec563ba4bac1d6aaf04141cd7d1605d7a5ca833e38f994051e818489592989b
3
+ size 166184960
dataset_infos.json ADDED
The diff for this file is too large to render. See raw diff
 
mmlu.py ADDED
@@ -0,0 +1,171 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # coding=utf-8
2
+ # Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
3
+ #
4
+ # Licensed under the Apache License, Version 2.0 (the "License");
5
+ # you may not use this file except in compliance with the License.
6
+ # You may obtain a copy of the License at
7
+ #
8
+ # http://www.apache.org/licenses/LICENSE-2.0
9
+ #
10
+ # Unless required by applicable law or agreed to in writing, software
11
+ # distributed under the License is distributed on an "AS IS" BASIS,
12
+ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13
+ # See the License for the specific language governing permissions and
14
+ # limitations under the License.
15
+
16
+
17
+ import csv
18
+
19
+ import datasets
20
+
21
+
22
+ _CITATION = """\
23
+ @article{hendryckstest2021,
24
+ title={Measuring Massive Multitask Language Understanding},
25
+ author={Dan Hendrycks and Collin Burns and Steven Basart and Andy Zou and Mantas Mazeika and Dawn Song and Jacob Steinhardt},
26
+ journal={Proceedings of the International Conference on Learning Representations (ICLR)},
27
+ year={2021}
28
+ }
29
+ """
30
+
31
+ _DESCRIPTION = """\
32
+ This is a massive multitask test consisting of multiple-choice questions from various branches of knowledge, covering 57 tasks including elementary mathematics, US history, computer science, law, and more.
33
+ """
34
+
35
+ _HOMEPAGE = "https://github.com/hendrycks/test"
36
+
37
+ _URL = "data.tar"
38
+
39
+ _SUBJECTS = [
40
+ "all",
41
+ "abstract_algebra",
42
+ "anatomy",
43
+ "astronomy",
44
+ "business_ethics",
45
+ "clinical_knowledge",
46
+ "college_biology",
47
+ "college_chemistry",
48
+ "college_computer_science",
49
+ "college_mathematics",
50
+ "college_medicine",
51
+ "college_physics",
52
+ "computer_security",
53
+ "conceptual_physics",
54
+ "econometrics",
55
+ "electrical_engineering",
56
+ "elementary_mathematics",
57
+ "formal_logic",
58
+ "global_facts",
59
+ "high_school_biology",
60
+ "high_school_chemistry",
61
+ "high_school_computer_science",
62
+ "high_school_european_history",
63
+ "high_school_geography",
64
+ "high_school_government_and_politics",
65
+ "high_school_macroeconomics",
66
+ "high_school_mathematics",
67
+ "high_school_microeconomics",
68
+ "high_school_physics",
69
+ "high_school_psychology",
70
+ "high_school_statistics",
71
+ "high_school_us_history",
72
+ "high_school_world_history",
73
+ "human_aging",
74
+ "human_sexuality",
75
+ "international_law",
76
+ "jurisprudence",
77
+ "logical_fallacies",
78
+ "machine_learning",
79
+ "management",
80
+ "marketing",
81
+ "medical_genetics",
82
+ "miscellaneous",
83
+ "moral_disputes",
84
+ "moral_scenarios",
85
+ "nutrition",
86
+ "philosophy",
87
+ "prehistory",
88
+ "professional_accounting",
89
+ "professional_law",
90
+ "professional_medicine",
91
+ "professional_psychology",
92
+ "public_relations",
93
+ "security_studies",
94
+ "sociology",
95
+ "us_foreign_policy",
96
+ "virology",
97
+ "world_religions",
98
+ ]
99
+
100
+
101
+ class Mmlu(datasets.GeneratorBasedBuilder):
102
+ """Measuring Massive Multitask Language Understanding, consisting of 57 tasks"""
103
+
104
+ BUILDER_CONFIGS = [
105
+ datasets.BuilderConfig(
106
+ name=sub, version=datasets.Version("1.0.0"), description=f"MMLU Subject {sub}"
107
+ )
108
+ for sub in _SUBJECTS
109
+ ]
110
+
111
+ def _info(self):
112
+ features = datasets.Features(
113
+ {
114
+ "question": datasets.Value("string"),
115
+ "choices": datasets.features.Sequence(datasets.Value("string")),
116
+ "answer": datasets.features.ClassLabel(num_classes=4, names=["A", "B", "C", "D"]),
117
+ }
118
+ )
119
+ return datasets.DatasetInfo(
120
+ description=_DESCRIPTION,
121
+ features=features,
122
+ homepage=_HOMEPAGE,
123
+ citation=_CITATION,
124
+ )
125
+
126
+ def _split_generators(self, dl_manager):
127
+ """Returns SplitGenerators."""
128
+ archive = dl_manager.download(_URL)
129
+ return [
130
+ datasets.SplitGenerator(
131
+ name=datasets.Split("auxiliary_train"),
132
+ gen_kwargs={
133
+ "iter_archive": dl_manager.iter_archive(archive),
134
+ "split": "auxiliary_train",
135
+ },
136
+ ),
137
+ datasets.SplitGenerator(
138
+ name=datasets.Split.TEST,
139
+ gen_kwargs={"iter_archive": dl_manager.iter_archive(archive), "split": "test"},
140
+ ),
141
+ datasets.SplitGenerator(
142
+ name=datasets.Split.VALIDATION,
143
+ gen_kwargs={
144
+ "iter_archive": dl_manager.iter_archive(archive),
145
+ "split": "val",
146
+ },
147
+ ),
148
+ datasets.SplitGenerator(
149
+ name=datasets.Split("dev"),
150
+ gen_kwargs={
151
+ "iter_archive": dl_manager.iter_archive(archive),
152
+ "split": "dev",
153
+ },
154
+ ),
155
+ ]
156
+
157
+ def _generate_examples(self, iter_archive, split):
158
+ """Yields examples as (key, example) tuples."""
159
+ n_yielded_files = 0
160
+ for id_file, (path, file) in enumerate(iter_archive):
161
+ if f"/{split}/" in path:
162
+ if split == "auxiliary_train" or (self.config.name in path or self.config.name == "all"):
163
+ n_yielded_files += 1
164
+ lines = (line.decode("utf-8") for line in file)
165
+ reader = csv.reader(lines)
166
+ for id_line, data in enumerate(reader):
167
+ yield f"{id_file}_{id_line}", {"question": data[0], "choices": data[1:5], "answer": data[5]}
168
+ #else:
169
+ #print("KO", path)
170
+ #else:
171
+ #print("KO2", split, path)