gsaltintas commited on
Commit
c118787
·
verified ·
1 Parent(s): 4bd0912

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +0 -1154
README.md CHANGED
@@ -7,1160 +7,6 @@ pretty_name: Tokenization Robustness Math
7
  tags:
8
  - multilingual
9
  - tokenization
10
- dataset_info:
11
- - config_name: tokenizer_robustness_completion_math_canonical
12
- features:
13
- - name: question
14
- dtype: string
15
- - name: choices
16
- list: string
17
- - name: answer
18
- dtype: int64
19
- - name: answer_label
20
- dtype: string
21
- - name: split
22
- dtype: string
23
- - name: subcategories
24
- dtype: string
25
- - name: category
26
- dtype: string
27
- - name: lang
28
- dtype: string
29
- - name: second_lang
30
- dtype: string
31
- - name: notes
32
- dtype: string
33
- - name: id
34
- dtype: string
35
- - name: set_id
36
- dtype: string
37
- - name: variation_id
38
- dtype: string
39
- - name: vanilla_cos_sim_to_canonical
40
- struct:
41
- - name: CohereLabs/aya-expanse-8b
42
- dtype: float64
43
- - name: Qwen/Qwen3-8B
44
- dtype: float64
45
- - name: bigscience/bloom
46
- dtype: float64
47
- - name: common-pile/comma-v0.1-1t
48
- dtype: float64
49
- - name: facebook/xglm-564M
50
- dtype: float64
51
- - name: google-bert/bert-base-multilingual-cased
52
- dtype: float64
53
- - name: google/byt5-small
54
- dtype: float64
55
- - name: google/gemma-2-2b
56
- dtype: float64
57
- - name: gpt2
58
- dtype: float64
59
- - name: meta-llama/Llama-3.2-1B
60
- dtype: float64
61
- - name: microsoft/Phi-3-mini-4k-instruct
62
- dtype: float64
63
- - name: mistralai/tekken
64
- dtype: float64
65
- - name: tiktoken/gpt-4o
66
- dtype: float64
67
- - name: tokenmonster/englishcode-32000-consistent-v1
68
- dtype: float64
69
- - name: trimmed_cos_sim_to_canonical
70
- struct:
71
- - name: CohereLabs/aya-expanse-8b
72
- dtype: float64
73
- - name: Qwen/Qwen3-8B
74
- dtype: float64
75
- - name: bigscience/bloom
76
- dtype: float64
77
- - name: common-pile/comma-v0.1-1t
78
- dtype: float64
79
- - name: facebook/xglm-564M
80
- dtype: float64
81
- - name: google-bert/bert-base-multilingual-cased
82
- dtype: float64
83
- - name: google/byt5-small
84
- dtype: float64
85
- - name: google/gemma-2-2b
86
- dtype: float64
87
- - name: gpt2
88
- dtype: float64
89
- - name: meta-llama/Llama-3.2-1B
90
- dtype: float64
91
- - name: microsoft/Phi-3-mini-4k-instruct
92
- dtype: float64
93
- - name: mistralai/tekken
94
- dtype: float64
95
- - name: tiktoken/gpt-4o
96
- dtype: float64
97
- - name: tokenmonster/englishcode-32000-consistent-v1
98
- dtype: float64
99
- - name: token_counts
100
- struct:
101
- - name: CohereLabs/aya-expanse-8b
102
- dtype: int64
103
- - name: Qwen/Qwen3-8B
104
- dtype: int64
105
- - name: bigscience/bloom
106
- dtype: int64
107
- - name: common-pile/comma-v0.1-1t
108
- dtype: int64
109
- - name: facebook/xglm-564M
110
- dtype: int64
111
- - name: google-bert/bert-base-multilingual-cased
112
- dtype: int64
113
- - name: google/byt5-small
114
- dtype: int64
115
- - name: google/gemma-2-2b
116
- dtype: int64
117
- - name: gpt2
118
- dtype: int64
119
- - name: meta-llama/Llama-3.2-1B
120
- dtype: int64
121
- - name: microsoft/Phi-3-mini-4k-instruct
122
- dtype: int64
123
- - name: mistralai/tekken
124
- dtype: int64
125
- - name: tiktoken/gpt-4o
126
- dtype: int64
127
- - name: tokenmonster/englishcode-32000-consistent-v1
128
- dtype: int64
129
- splits:
130
- - name: test
131
- num_bytes: 11202
132
- num_examples: 21
133
- download_size: 29983
134
- dataset_size: 11202
135
- - config_name: tokenizer_robustness_completion_math_chinese
136
- features:
137
- - name: question
138
- dtype: string
139
- - name: choices
140
- list: string
141
- - name: answer
142
- dtype: int64
143
- - name: answer_label
144
- dtype: string
145
- - name: split
146
- dtype: string
147
- - name: subcategories
148
- dtype: string
149
- - name: category
150
- dtype: string
151
- - name: lang
152
- dtype: string
153
- - name: second_lang
154
- dtype: string
155
- - name: notes
156
- dtype: string
157
- - name: id
158
- dtype: string
159
- - name: set_id
160
- dtype: string
161
- - name: variation_id
162
- dtype: string
163
- - name: vanilla_cos_sim_to_canonical
164
- struct:
165
- - name: CohereLabs/aya-expanse-8b
166
- dtype: float64
167
- - name: Qwen/Qwen3-8B
168
- dtype: float64
169
- - name: bigscience/bloom
170
- dtype: float64
171
- - name: common-pile/comma-v0.1-1t
172
- dtype: float64
173
- - name: facebook/xglm-564M
174
- dtype: float64
175
- - name: google-bert/bert-base-multilingual-cased
176
- dtype: float64
177
- - name: google/byt5-small
178
- dtype: float64
179
- - name: google/gemma-2-2b
180
- dtype: float64
181
- - name: gpt2
182
- dtype: float64
183
- - name: meta-llama/Llama-3.2-1B
184
- dtype: float64
185
- - name: microsoft/Phi-3-mini-4k-instruct
186
- dtype: float64
187
- - name: mistralai/tekken
188
- dtype: float64
189
- - name: tiktoken/gpt-4o
190
- dtype: float64
191
- - name: tokenmonster/englishcode-32000-consistent-v1
192
- dtype: float64
193
- - name: trimmed_cos_sim_to_canonical
194
- struct:
195
- - name: CohereLabs/aya-expanse-8b
196
- dtype: float64
197
- - name: Qwen/Qwen3-8B
198
- dtype: float64
199
- - name: bigscience/bloom
200
- dtype: float64
201
- - name: common-pile/comma-v0.1-1t
202
- dtype: float64
203
- - name: facebook/xglm-564M
204
- dtype: float64
205
- - name: google-bert/bert-base-multilingual-cased
206
- dtype: float64
207
- - name: google/byt5-small
208
- dtype: float64
209
- - name: google/gemma-2-2b
210
- dtype: float64
211
- - name: gpt2
212
- dtype: float64
213
- - name: meta-llama/Llama-3.2-1B
214
- dtype: float64
215
- - name: microsoft/Phi-3-mini-4k-instruct
216
- dtype: float64
217
- - name: mistralai/tekken
218
- dtype: float64
219
- - name: tiktoken/gpt-4o
220
- dtype: float64
221
- - name: tokenmonster/englishcode-32000-consistent-v1
222
- dtype: float64
223
- - name: token_counts
224
- struct:
225
- - name: CohereLabs/aya-expanse-8b
226
- dtype: int64
227
- - name: Qwen/Qwen3-8B
228
- dtype: int64
229
- - name: bigscience/bloom
230
- dtype: int64
231
- - name: common-pile/comma-v0.1-1t
232
- dtype: int64
233
- - name: facebook/xglm-564M
234
- dtype: int64
235
- - name: google-bert/bert-base-multilingual-cased
236
- dtype: int64
237
- - name: google/byt5-small
238
- dtype: int64
239
- - name: google/gemma-2-2b
240
- dtype: int64
241
- - name: gpt2
242
- dtype: int64
243
- - name: meta-llama/Llama-3.2-1B
244
- dtype: int64
245
- - name: microsoft/Phi-3-mini-4k-instruct
246
- dtype: int64
247
- - name: mistralai/tekken
248
- dtype: int64
249
- - name: tiktoken/gpt-4o
250
- dtype: int64
251
- - name: tokenmonster/englishcode-32000-consistent-v1
252
- dtype: int64
253
- splits:
254
- - name: test
255
- num_bytes: 11147
256
- num_examples: 21
257
- download_size: 34446
258
- dataset_size: 11147
259
- - config_name: tokenizer_robustness_completion_math_decorative_unicode
260
- features:
261
- - name: question
262
- dtype: string
263
- - name: choices
264
- list: string
265
- - name: answer
266
- dtype: int64
267
- - name: answer_label
268
- dtype: string
269
- - name: split
270
- dtype: string
271
- - name: subcategories
272
- dtype: string
273
- - name: category
274
- dtype: string
275
- - name: lang
276
- dtype: string
277
- - name: second_lang
278
- dtype: string
279
- - name: notes
280
- dtype: string
281
- - name: id
282
- dtype: string
283
- - name: set_id
284
- dtype: string
285
- - name: variation_id
286
- dtype: string
287
- - name: vanilla_cos_sim_to_canonical
288
- struct:
289
- - name: CohereLabs/aya-expanse-8b
290
- dtype: float64
291
- - name: Qwen/Qwen3-8B
292
- dtype: float64
293
- - name: bigscience/bloom
294
- dtype: float64
295
- - name: common-pile/comma-v0.1-1t
296
- dtype: float64
297
- - name: facebook/xglm-564M
298
- dtype: float64
299
- - name: google-bert/bert-base-multilingual-cased
300
- dtype: float64
301
- - name: google/byt5-small
302
- dtype: float64
303
- - name: google/gemma-2-2b
304
- dtype: float64
305
- - name: gpt2
306
- dtype: float64
307
- - name: meta-llama/Llama-3.2-1B
308
- dtype: float64
309
- - name: microsoft/Phi-3-mini-4k-instruct
310
- dtype: float64
311
- - name: mistralai/tekken
312
- dtype: float64
313
- - name: tiktoken/gpt-4o
314
- dtype: float64
315
- - name: tokenmonster/englishcode-32000-consistent-v1
316
- dtype: float64
317
- - name: trimmed_cos_sim_to_canonical
318
- struct:
319
- - name: CohereLabs/aya-expanse-8b
320
- dtype: float64
321
- - name: Qwen/Qwen3-8B
322
- dtype: float64
323
- - name: bigscience/bloom
324
- dtype: float64
325
- - name: common-pile/comma-v0.1-1t
326
- dtype: float64
327
- - name: facebook/xglm-564M
328
- dtype: float64
329
- - name: google-bert/bert-base-multilingual-cased
330
- dtype: float64
331
- - name: google/byt5-small
332
- dtype: float64
333
- - name: google/gemma-2-2b
334
- dtype: float64
335
- - name: gpt2
336
- dtype: float64
337
- - name: meta-llama/Llama-3.2-1B
338
- dtype: float64
339
- - name: microsoft/Phi-3-mini-4k-instruct
340
- dtype: float64
341
- - name: mistralai/tekken
342
- dtype: float64
343
- - name: tiktoken/gpt-4o
344
- dtype: float64
345
- - name: tokenmonster/englishcode-32000-consistent-v1
346
- dtype: float64
347
- - name: token_counts
348
- struct:
349
- - name: CohereLabs/aya-expanse-8b
350
- dtype: int64
351
- - name: Qwen/Qwen3-8B
352
- dtype: int64
353
- - name: bigscience/bloom
354
- dtype: int64
355
- - name: common-pile/comma-v0.1-1t
356
- dtype: int64
357
- - name: facebook/xglm-564M
358
- dtype: int64
359
- - name: google-bert/bert-base-multilingual-cased
360
- dtype: int64
361
- - name: google/byt5-small
362
- dtype: int64
363
- - name: google/gemma-2-2b
364
- dtype: int64
365
- - name: gpt2
366
- dtype: int64
367
- - name: meta-llama/Llama-3.2-1B
368
- dtype: int64
369
- - name: microsoft/Phi-3-mini-4k-instruct
370
- dtype: int64
371
- - name: mistralai/tekken
372
- dtype: int64
373
- - name: tiktoken/gpt-4o
374
- dtype: int64
375
- - name: tokenmonster/englishcode-32000-consistent-v1
376
- dtype: int64
377
- splits:
378
- - name: test
379
- num_bytes: 11986
380
- num_examples: 21
381
- download_size: 34661
382
- dataset_size: 11986
383
- - config_name: tokenizer_robustness_completion_math_farsi
384
- features:
385
- - name: question
386
- dtype: string
387
- - name: choices
388
- list: string
389
- - name: answer
390
- dtype: int64
391
- - name: answer_label
392
- dtype: string
393
- - name: split
394
- dtype: string
395
- - name: subcategories
396
- dtype: string
397
- - name: category
398
- dtype: string
399
- - name: lang
400
- dtype: string
401
- - name: second_lang
402
- dtype: string
403
- - name: notes
404
- dtype: string
405
- - name: id
406
- dtype: string
407
- - name: set_id
408
- dtype: string
409
- - name: variation_id
410
- dtype: string
411
- - name: vanilla_cos_sim_to_canonical
412
- struct:
413
- - name: CohereLabs/aya-expanse-8b
414
- dtype: float64
415
- - name: Qwen/Qwen3-8B
416
- dtype: float64
417
- - name: bigscience/bloom
418
- dtype: float64
419
- - name: common-pile/comma-v0.1-1t
420
- dtype: float64
421
- - name: facebook/xglm-564M
422
- dtype: float64
423
- - name: google-bert/bert-base-multilingual-cased
424
- dtype: float64
425
- - name: google/byt5-small
426
- dtype: float64
427
- - name: google/gemma-2-2b
428
- dtype: float64
429
- - name: gpt2
430
- dtype: float64
431
- - name: meta-llama/Llama-3.2-1B
432
- dtype: float64
433
- - name: microsoft/Phi-3-mini-4k-instruct
434
- dtype: float64
435
- - name: mistralai/tekken
436
- dtype: float64
437
- - name: tiktoken/gpt-4o
438
- dtype: float64
439
- - name: tokenmonster/englishcode-32000-consistent-v1
440
- dtype: float64
441
- - name: trimmed_cos_sim_to_canonical
442
- struct:
443
- - name: CohereLabs/aya-expanse-8b
444
- dtype: float64
445
- - name: Qwen/Qwen3-8B
446
- dtype: float64
447
- - name: bigscience/bloom
448
- dtype: float64
449
- - name: common-pile/comma-v0.1-1t
450
- dtype: float64
451
- - name: facebook/xglm-564M
452
- dtype: float64
453
- - name: google-bert/bert-base-multilingual-cased
454
- dtype: float64
455
- - name: google/byt5-small
456
- dtype: float64
457
- - name: google/gemma-2-2b
458
- dtype: float64
459
- - name: gpt2
460
- dtype: float64
461
- - name: meta-llama/Llama-3.2-1B
462
- dtype: float64
463
- - name: microsoft/Phi-3-mini-4k-instruct
464
- dtype: float64
465
- - name: mistralai/tekken
466
- dtype: float64
467
- - name: tiktoken/gpt-4o
468
- dtype: float64
469
- - name: tokenmonster/englishcode-32000-consistent-v1
470
- dtype: float64
471
- - name: token_counts
472
- struct:
473
- - name: CohereLabs/aya-expanse-8b
474
- dtype: int64
475
- - name: Qwen/Qwen3-8B
476
- dtype: int64
477
- - name: bigscience/bloom
478
- dtype: int64
479
- - name: common-pile/comma-v0.1-1t
480
- dtype: int64
481
- - name: facebook/xglm-564M
482
- dtype: int64
483
- - name: google-bert/bert-base-multilingual-cased
484
- dtype: int64
485
- - name: google/byt5-small
486
- dtype: int64
487
- - name: google/gemma-2-2b
488
- dtype: int64
489
- - name: gpt2
490
- dtype: int64
491
- - name: meta-llama/Llama-3.2-1B
492
- dtype: int64
493
- - name: microsoft/Phi-3-mini-4k-instruct
494
- dtype: int64
495
- - name: mistralai/tekken
496
- dtype: int64
497
- - name: tiktoken/gpt-4o
498
- dtype: int64
499
- - name: tokenmonster/englishcode-32000-consistent-v1
500
- dtype: int64
501
- splits:
502
- - name: test
503
- num_bytes: 12034
504
- num_examples: 21
505
- download_size: 34853
506
- dataset_size: 12034
507
- - config_name: tokenizer_robustness_completion_math_italian
508
- features:
509
- - name: question
510
- dtype: string
511
- - name: choices
512
- list: string
513
- - name: answer
514
- dtype: int64
515
- - name: answer_label
516
- dtype: string
517
- - name: split
518
- dtype: string
519
- - name: subcategories
520
- dtype: string
521
- - name: category
522
- dtype: string
523
- - name: lang
524
- dtype: string
525
- - name: second_lang
526
- dtype: string
527
- - name: notes
528
- dtype: string
529
- - name: id
530
- dtype: string
531
- - name: set_id
532
- dtype: string
533
- - name: variation_id
534
- dtype: string
535
- - name: vanilla_cos_sim_to_canonical
536
- struct:
537
- - name: CohereLabs/aya-expanse-8b
538
- dtype: float64
539
- - name: Qwen/Qwen3-8B
540
- dtype: float64
541
- - name: bigscience/bloom
542
- dtype: float64
543
- - name: common-pile/comma-v0.1-1t
544
- dtype: float64
545
- - name: facebook/xglm-564M
546
- dtype: float64
547
- - name: google-bert/bert-base-multilingual-cased
548
- dtype: float64
549
- - name: google/byt5-small
550
- dtype: float64
551
- - name: google/gemma-2-2b
552
- dtype: float64
553
- - name: gpt2
554
- dtype: float64
555
- - name: meta-llama/Llama-3.2-1B
556
- dtype: float64
557
- - name: microsoft/Phi-3-mini-4k-instruct
558
- dtype: float64
559
- - name: mistralai/tekken
560
- dtype: float64
561
- - name: tiktoken/gpt-4o
562
- dtype: float64
563
- - name: tokenmonster/englishcode-32000-consistent-v1
564
- dtype: float64
565
- - name: trimmed_cos_sim_to_canonical
566
- struct:
567
- - name: CohereLabs/aya-expanse-8b
568
- dtype: float64
569
- - name: Qwen/Qwen3-8B
570
- dtype: float64
571
- - name: bigscience/bloom
572
- dtype: float64
573
- - name: common-pile/comma-v0.1-1t
574
- dtype: float64
575
- - name: facebook/xglm-564M
576
- dtype: float64
577
- - name: google-bert/bert-base-multilingual-cased
578
- dtype: float64
579
- - name: google/byt5-small
580
- dtype: float64
581
- - name: google/gemma-2-2b
582
- dtype: float64
583
- - name: gpt2
584
- dtype: float64
585
- - name: meta-llama/Llama-3.2-1B
586
- dtype: float64
587
- - name: microsoft/Phi-3-mini-4k-instruct
588
- dtype: float64
589
- - name: mistralai/tekken
590
- dtype: float64
591
- - name: tiktoken/gpt-4o
592
- dtype: float64
593
- - name: tokenmonster/englishcode-32000-consistent-v1
594
- dtype: float64
595
- - name: token_counts
596
- struct:
597
- - name: CohereLabs/aya-expanse-8b
598
- dtype: int64
599
- - name: Qwen/Qwen3-8B
600
- dtype: int64
601
- - name: bigscience/bloom
602
- dtype: int64
603
- - name: common-pile/comma-v0.1-1t
604
- dtype: int64
605
- - name: facebook/xglm-564M
606
- dtype: int64
607
- - name: google-bert/bert-base-multilingual-cased
608
- dtype: int64
609
- - name: google/byt5-small
610
- dtype: int64
611
- - name: google/gemma-2-2b
612
- dtype: int64
613
- - name: gpt2
614
- dtype: int64
615
- - name: meta-llama/Llama-3.2-1B
616
- dtype: int64
617
- - name: microsoft/Phi-3-mini-4k-instruct
618
- dtype: int64
619
- - name: mistralai/tekken
620
- dtype: int64
621
- - name: tiktoken/gpt-4o
622
- dtype: int64
623
- - name: tokenmonster/englishcode-32000-consistent-v1
624
- dtype: int64
625
- splits:
626
- - name: test
627
- num_bytes: 11219
628
- num_examples: 21
629
- download_size: 34633
630
- dataset_size: 11219
631
- - config_name: tokenizer_robustness_completion_math_latex
632
- features:
633
- - name: question
634
- dtype: string
635
- - name: choices
636
- list: string
637
- - name: answer
638
- dtype: int64
639
- - name: answer_label
640
- dtype: string
641
- - name: split
642
- dtype: string
643
- - name: subcategories
644
- dtype: string
645
- - name: category
646
- dtype: string
647
- - name: lang
648
- dtype: string
649
- - name: second_lang
650
- dtype: string
651
- - name: notes
652
- dtype: string
653
- - name: id
654
- dtype: string
655
- - name: set_id
656
- dtype: string
657
- - name: variation_id
658
- dtype: string
659
- - name: vanilla_cos_sim_to_canonical
660
- struct:
661
- - name: CohereLabs/aya-expanse-8b
662
- dtype: float64
663
- - name: Qwen/Qwen3-8B
664
- dtype: float64
665
- - name: bigscience/bloom
666
- dtype: float64
667
- - name: common-pile/comma-v0.1-1t
668
- dtype: float64
669
- - name: facebook/xglm-564M
670
- dtype: float64
671
- - name: google-bert/bert-base-multilingual-cased
672
- dtype: float64
673
- - name: google/byt5-small
674
- dtype: float64
675
- - name: google/gemma-2-2b
676
- dtype: float64
677
- - name: gpt2
678
- dtype: float64
679
- - name: meta-llama/Llama-3.2-1B
680
- dtype: float64
681
- - name: microsoft/Phi-3-mini-4k-instruct
682
- dtype: float64
683
- - name: mistralai/tekken
684
- dtype: float64
685
- - name: tiktoken/gpt-4o
686
- dtype: float64
687
- - name: tokenmonster/englishcode-32000-consistent-v1
688
- dtype: float64
689
- - name: trimmed_cos_sim_to_canonical
690
- struct:
691
- - name: CohereLabs/aya-expanse-8b
692
- dtype: float64
693
- - name: Qwen/Qwen3-8B
694
- dtype: float64
695
- - name: bigscience/bloom
696
- dtype: float64
697
- - name: common-pile/comma-v0.1-1t
698
- dtype: float64
699
- - name: facebook/xglm-564M
700
- dtype: float64
701
- - name: google-bert/bert-base-multilingual-cased
702
- dtype: float64
703
- - name: google/byt5-small
704
- dtype: float64
705
- - name: google/gemma-2-2b
706
- dtype: float64
707
- - name: gpt2
708
- dtype: float64
709
- - name: meta-llama/Llama-3.2-1B
710
- dtype: float64
711
- - name: microsoft/Phi-3-mini-4k-instruct
712
- dtype: float64
713
- - name: mistralai/tekken
714
- dtype: float64
715
- - name: tiktoken/gpt-4o
716
- dtype: float64
717
- - name: tokenmonster/englishcode-32000-consistent-v1
718
- dtype: float64
719
- - name: token_counts
720
- struct:
721
- - name: CohereLabs/aya-expanse-8b
722
- dtype: int64
723
- - name: Qwen/Qwen3-8B
724
- dtype: int64
725
- - name: bigscience/bloom
726
- dtype: int64
727
- - name: common-pile/comma-v0.1-1t
728
- dtype: int64
729
- - name: facebook/xglm-564M
730
- dtype: int64
731
- - name: google-bert/bert-base-multilingual-cased
732
- dtype: int64
733
- - name: google/byt5-small
734
- dtype: int64
735
- - name: google/gemma-2-2b
736
- dtype: int64
737
- - name: gpt2
738
- dtype: int64
739
- - name: meta-llama/Llama-3.2-1B
740
- dtype: int64
741
- - name: microsoft/Phi-3-mini-4k-instruct
742
- dtype: int64
743
- - name: mistralai/tekken
744
- dtype: int64
745
- - name: tiktoken/gpt-4o
746
- dtype: int64
747
- - name: tokenmonster/englishcode-32000-consistent-v1
748
- dtype: int64
749
- splits:
750
- - name: test
751
- num_bytes: 11494
752
- num_examples: 21
753
- download_size: 34232
754
- dataset_size: 11494
755
- - config_name: tokenizer_robustness_completion_math_space_removal
756
- features:
757
- - name: question
758
- dtype: string
759
- - name: choices
760
- list: string
761
- - name: answer
762
- dtype: int64
763
- - name: answer_label
764
- dtype: string
765
- - name: split
766
- dtype: string
767
- - name: subcategories
768
- dtype: string
769
- - name: category
770
- dtype: string
771
- - name: lang
772
- dtype: string
773
- - name: second_lang
774
- dtype: string
775
- - name: notes
776
- dtype: string
777
- - name: id
778
- dtype: string
779
- - name: set_id
780
- dtype: string
781
- - name: variation_id
782
- dtype: string
783
- - name: vanilla_cos_sim_to_canonical
784
- struct:
785
- - name: CohereLabs/aya-expanse-8b
786
- dtype: float64
787
- - name: Qwen/Qwen3-8B
788
- dtype: float64
789
- - name: bigscience/bloom
790
- dtype: float64
791
- - name: common-pile/comma-v0.1-1t
792
- dtype: float64
793
- - name: facebook/xglm-564M
794
- dtype: float64
795
- - name: google-bert/bert-base-multilingual-cased
796
- dtype: float64
797
- - name: google/byt5-small
798
- dtype: float64
799
- - name: google/gemma-2-2b
800
- dtype: float64
801
- - name: gpt2
802
- dtype: float64
803
- - name: meta-llama/Llama-3.2-1B
804
- dtype: float64
805
- - name: microsoft/Phi-3-mini-4k-instruct
806
- dtype: float64
807
- - name: mistralai/tekken
808
- dtype: float64
809
- - name: tiktoken/gpt-4o
810
- dtype: float64
811
- - name: tokenmonster/englishcode-32000-consistent-v1
812
- dtype: float64
813
- - name: trimmed_cos_sim_to_canonical
814
- struct:
815
- - name: CohereLabs/aya-expanse-8b
816
- dtype: float64
817
- - name: Qwen/Qwen3-8B
818
- dtype: float64
819
- - name: bigscience/bloom
820
- dtype: float64
821
- - name: common-pile/comma-v0.1-1t
822
- dtype: float64
823
- - name: facebook/xglm-564M
824
- dtype: float64
825
- - name: google-bert/bert-base-multilingual-cased
826
- dtype: float64
827
- - name: google/byt5-small
828
- dtype: float64
829
- - name: google/gemma-2-2b
830
- dtype: float64
831
- - name: gpt2
832
- dtype: float64
833
- - name: meta-llama/Llama-3.2-1B
834
- dtype: float64
835
- - name: microsoft/Phi-3-mini-4k-instruct
836
- dtype: float64
837
- - name: mistralai/tekken
838
- dtype: float64
839
- - name: tiktoken/gpt-4o
840
- dtype: float64
841
- - name: tokenmonster/englishcode-32000-consistent-v1
842
- dtype: float64
843
- - name: token_counts
844
- struct:
845
- - name: CohereLabs/aya-expanse-8b
846
- dtype: int64
847
- - name: Qwen/Qwen3-8B
848
- dtype: int64
849
- - name: bigscience/bloom
850
- dtype: int64
851
- - name: common-pile/comma-v0.1-1t
852
- dtype: int64
853
- - name: facebook/xglm-564M
854
- dtype: int64
855
- - name: google-bert/bert-base-multilingual-cased
856
- dtype: int64
857
- - name: google/byt5-small
858
- dtype: int64
859
- - name: google/gemma-2-2b
860
- dtype: int64
861
- - name: gpt2
862
- dtype: int64
863
- - name: meta-llama/Llama-3.2-1B
864
- dtype: int64
865
- - name: microsoft/Phi-3-mini-4k-instruct
866
- dtype: int64
867
- - name: mistralai/tekken
868
- dtype: int64
869
- - name: tiktoken/gpt-4o
870
- dtype: int64
871
- - name: tokenmonster/englishcode-32000-consistent-v1
872
- dtype: int64
873
- splits:
874
- - name: test
875
- num_bytes: 11559
876
- num_examples: 21
877
- download_size: 34055
878
- dataset_size: 11559
879
- - config_name: tokenizer_robustness_completion_math_spelled_out
880
- features:
881
- - name: question
882
- dtype: string
883
- - name: choices
884
- list: string
885
- - name: answer
886
- dtype: int64
887
- - name: answer_label
888
- dtype: string
889
- - name: split
890
- dtype: string
891
- - name: subcategories
892
- dtype: string
893
- - name: category
894
- dtype: string
895
- - name: lang
896
- dtype: string
897
- - name: second_lang
898
- dtype: string
899
- - name: notes
900
- dtype: string
901
- - name: id
902
- dtype: string
903
- - name: set_id
904
- dtype: string
905
- - name: variation_id
906
- dtype: string
907
- - name: vanilla_cos_sim_to_canonical
908
- struct:
909
- - name: CohereLabs/aya-expanse-8b
910
- dtype: float64
911
- - name: Qwen/Qwen3-8B
912
- dtype: float64
913
- - name: bigscience/bloom
914
- dtype: float64
915
- - name: common-pile/comma-v0.1-1t
916
- dtype: float64
917
- - name: facebook/xglm-564M
918
- dtype: float64
919
- - name: google-bert/bert-base-multilingual-cased
920
- dtype: float64
921
- - name: google/byt5-small
922
- dtype: float64
923
- - name: google/gemma-2-2b
924
- dtype: float64
925
- - name: gpt2
926
- dtype: float64
927
- - name: meta-llama/Llama-3.2-1B
928
- dtype: float64
929
- - name: microsoft/Phi-3-mini-4k-instruct
930
- dtype: float64
931
- - name: mistralai/tekken
932
- dtype: float64
933
- - name: tiktoken/gpt-4o
934
- dtype: float64
935
- - name: tokenmonster/englishcode-32000-consistent-v1
936
- dtype: float64
937
- - name: trimmed_cos_sim_to_canonical
938
- struct:
939
- - name: CohereLabs/aya-expanse-8b
940
- dtype: float64
941
- - name: Qwen/Qwen3-8B
942
- dtype: float64
943
- - name: bigscience/bloom
944
- dtype: float64
945
- - name: common-pile/comma-v0.1-1t
946
- dtype: float64
947
- - name: facebook/xglm-564M
948
- dtype: float64
949
- - name: google-bert/bert-base-multilingual-cased
950
- dtype: float64
951
- - name: google/byt5-small
952
- dtype: float64
953
- - name: google/gemma-2-2b
954
- dtype: float64
955
- - name: gpt2
956
- dtype: float64
957
- - name: meta-llama/Llama-3.2-1B
958
- dtype: float64
959
- - name: microsoft/Phi-3-mini-4k-instruct
960
- dtype: float64
961
- - name: mistralai/tekken
962
- dtype: float64
963
- - name: tiktoken/gpt-4o
964
- dtype: float64
965
- - name: tokenmonster/englishcode-32000-consistent-v1
966
- dtype: float64
967
- - name: token_counts
968
- struct:
969
- - name: CohereLabs/aya-expanse-8b
970
- dtype: int64
971
- - name: Qwen/Qwen3-8B
972
- dtype: int64
973
- - name: bigscience/bloom
974
- dtype: int64
975
- - name: common-pile/comma-v0.1-1t
976
- dtype: int64
977
- - name: facebook/xglm-564M
978
- dtype: int64
979
- - name: google-bert/bert-base-multilingual-cased
980
- dtype: int64
981
- - name: google/byt5-small
982
- dtype: int64
983
- - name: google/gemma-2-2b
984
- dtype: int64
985
- - name: gpt2
986
- dtype: int64
987
- - name: meta-llama/Llama-3.2-1B
988
- dtype: int64
989
- - name: microsoft/Phi-3-mini-4k-instruct
990
- dtype: int64
991
- - name: mistralai/tekken
992
- dtype: int64
993
- - name: tiktoken/gpt-4o
994
- dtype: int64
995
- - name: tokenmonster/englishcode-32000-consistent-v1
996
- dtype: int64
997
- splits:
998
- - name: test
999
- num_bytes: 12129
1000
- num_examples: 21
1001
- download_size: 34627
1002
- dataset_size: 12129
1003
- - config_name: tokenizer_robustness_completion_math_turkish
1004
- features:
1005
- - name: question
1006
- dtype: string
1007
- - name: choices
1008
- list: string
1009
- - name: answer
1010
- dtype: int64
1011
- - name: answer_label
1012
- dtype: string
1013
- - name: split
1014
- dtype: string
1015
- - name: subcategories
1016
- dtype: string
1017
- - name: category
1018
- dtype: string
1019
- - name: lang
1020
- dtype: string
1021
- - name: second_lang
1022
- dtype: string
1023
- - name: notes
1024
- dtype: string
1025
- - name: id
1026
- dtype: string
1027
- - name: set_id
1028
- dtype: string
1029
- - name: variation_id
1030
- dtype: string
1031
- - name: vanilla_cos_sim_to_canonical
1032
- struct:
1033
- - name: CohereLabs/aya-expanse-8b
1034
- dtype: float64
1035
- - name: Qwen/Qwen3-8B
1036
- dtype: float64
1037
- - name: bigscience/bloom
1038
- dtype: float64
1039
- - name: common-pile/comma-v0.1-1t
1040
- dtype: float64
1041
- - name: facebook/xglm-564M
1042
- dtype: float64
1043
- - name: google-bert/bert-base-multilingual-cased
1044
- dtype: float64
1045
- - name: google/byt5-small
1046
- dtype: float64
1047
- - name: google/gemma-2-2b
1048
- dtype: float64
1049
- - name: gpt2
1050
- dtype: float64
1051
- - name: meta-llama/Llama-3.2-1B
1052
- dtype: float64
1053
- - name: microsoft/Phi-3-mini-4k-instruct
1054
- dtype: float64
1055
- - name: mistralai/tekken
1056
- dtype: float64
1057
- - name: tiktoken/gpt-4o
1058
- dtype: float64
1059
- - name: tokenmonster/englishcode-32000-consistent-v1
1060
- dtype: float64
1061
- - name: trimmed_cos_sim_to_canonical
1062
- struct:
1063
- - name: CohereLabs/aya-expanse-8b
1064
- dtype: float64
1065
- - name: Qwen/Qwen3-8B
1066
- dtype: float64
1067
- - name: bigscience/bloom
1068
- dtype: float64
1069
- - name: common-pile/comma-v0.1-1t
1070
- dtype: float64
1071
- - name: facebook/xglm-564M
1072
- dtype: float64
1073
- - name: google-bert/bert-base-multilingual-cased
1074
- dtype: float64
1075
- - name: google/byt5-small
1076
- dtype: float64
1077
- - name: google/gemma-2-2b
1078
- dtype: float64
1079
- - name: gpt2
1080
- dtype: float64
1081
- - name: meta-llama/Llama-3.2-1B
1082
- dtype: float64
1083
- - name: microsoft/Phi-3-mini-4k-instruct
1084
- dtype: float64
1085
- - name: mistralai/tekken
1086
- dtype: float64
1087
- - name: tiktoken/gpt-4o
1088
- dtype: float64
1089
- - name: tokenmonster/englishcode-32000-consistent-v1
1090
- dtype: float64
1091
- - name: token_counts
1092
- struct:
1093
- - name: CohereLabs/aya-expanse-8b
1094
- dtype: int64
1095
- - name: Qwen/Qwen3-8B
1096
- dtype: int64
1097
- - name: bigscience/bloom
1098
- dtype: int64
1099
- - name: common-pile/comma-v0.1-1t
1100
- dtype: int64
1101
- - name: facebook/xglm-564M
1102
- dtype: int64
1103
- - name: google-bert/bert-base-multilingual-cased
1104
- dtype: int64
1105
- - name: google/byt5-small
1106
- dtype: int64
1107
- - name: google/gemma-2-2b
1108
- dtype: int64
1109
- - name: gpt2
1110
- dtype: int64
1111
- - name: meta-llama/Llama-3.2-1B
1112
- dtype: int64
1113
- - name: microsoft/Phi-3-mini-4k-instruct
1114
- dtype: int64
1115
- - name: mistralai/tekken
1116
- dtype: int64
1117
- - name: tiktoken/gpt-4o
1118
- dtype: int64
1119
- - name: tokenmonster/englishcode-32000-consistent-v1
1120
- dtype: int64
1121
- splits:
1122
- - name: test
1123
- num_bytes: 11339
1124
- num_examples: 21
1125
- download_size: 34647
1126
- dataset_size: 11339
1127
- configs:
1128
- - config_name: tokenizer_robustness_completion_math_canonical
1129
- data_files:
1130
- - split: test
1131
- path: tokenizer_robustness_completion_math_canonical/test-*
1132
- - config_name: tokenizer_robustness_completion_math_chinese
1133
- data_files:
1134
- - split: test
1135
- path: tokenizer_robustness_completion_math_chinese/test-*
1136
- - config_name: tokenizer_robustness_completion_math_decorative_unicode
1137
- data_files:
1138
- - split: test
1139
- path: tokenizer_robustness_completion_math_decorative_unicode/test-*
1140
- - config_name: tokenizer_robustness_completion_math_farsi
1141
- data_files:
1142
- - split: test
1143
- path: tokenizer_robustness_completion_math_farsi/test-*
1144
- - config_name: tokenizer_robustness_completion_math_italian
1145
- data_files:
1146
- - split: test
1147
- path: tokenizer_robustness_completion_math_italian/test-*
1148
- - config_name: tokenizer_robustness_completion_math_latex
1149
- data_files:
1150
- - split: test
1151
- path: tokenizer_robustness_completion_math_latex/test-*
1152
- - config_name: tokenizer_robustness_completion_math_space_removal
1153
- data_files:
1154
- - split: test
1155
- path: tokenizer_robustness_completion_math_space_removal/test-*
1156
- - config_name: tokenizer_robustness_completion_math_spelled_out
1157
- data_files:
1158
- - split: test
1159
- path: tokenizer_robustness_completion_math_spelled_out/test-*
1160
- - config_name: tokenizer_robustness_completion_math_turkish
1161
- data_files:
1162
- - split: test
1163
- path: tokenizer_robustness_completion_math_turkish/test-*
1164
  ---
1165
 
1166
  # Dataset Card for Tokenization Robustness Math
 
7
  tags:
8
  - multilingual
9
  - tokenization
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
10
  ---
11
 
12
  # Dataset Card for Tokenization Robustness Math