gsaltintas commited on
Commit
2219bd6
·
verified ·
1 Parent(s): e649c08

Uploading tokenizer_robustness_completion_italian_capitalization subset

Browse files
README.md CHANGED
@@ -260,6 +260,132 @@ dataset_info:
260
  num_examples: 40
261
  download_size: 33668
262
  dataset_size: 22185
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
263
  configs:
264
  - config_name: tokenizer_robustness_completion_italian_abbreviations
265
  data_files:
@@ -269,6 +395,10 @@ configs:
269
  data_files:
270
  - split: test
271
  path: tokenizer_robustness_completion_italian_canonical/test-*
 
 
 
 
272
  ---
273
 
274
  # Dataset Card for Tokenization Robustness
 
260
  num_examples: 40
261
  download_size: 33668
262
  dataset_size: 22185
263
+ - config_name: tokenizer_robustness_completion_italian_capitalization
264
+ features:
265
+ - name: question
266
+ dtype: string
267
+ - name: choices
268
+ list: string
269
+ - name: answer
270
+ dtype: int64
271
+ - name: answer_label
272
+ dtype: string
273
+ - name: split
274
+ dtype: string
275
+ - name: subcategories
276
+ dtype: string
277
+ - name: category
278
+ dtype: string
279
+ - name: lang
280
+ dtype: string
281
+ - name: second_lang
282
+ dtype: string
283
+ - name: notes
284
+ dtype: string
285
+ - name: id
286
+ dtype: string
287
+ - name: set_id
288
+ dtype: string
289
+ - name: variation_id
290
+ dtype: string
291
+ - name: perturbed_word
292
+ dtype: string
293
+ - name: vanilla_cos_sim_to_canonical
294
+ struct:
295
+ - name: CohereLabs/aya-expanse-8b
296
+ dtype: float64
297
+ - name: Qwen/Qwen3-8B
298
+ dtype: float64
299
+ - name: bigscience/bloom
300
+ dtype: float64
301
+ - name: common-pile/comma-v0.1-1t
302
+ dtype: float64
303
+ - name: facebook/xglm-564M
304
+ dtype: float64
305
+ - name: google-bert/bert-base-multilingual-cased
306
+ dtype: float64
307
+ - name: google/byt5-small
308
+ dtype: float64
309
+ - name: google/gemma-2-2b
310
+ dtype: float64
311
+ - name: gpt2
312
+ dtype: float64
313
+ - name: meta-llama/Llama-3.2-1B
314
+ dtype: float64
315
+ - name: microsoft/Phi-3-mini-4k-instruct
316
+ dtype: float64
317
+ - name: mistralai/tekken
318
+ dtype: float64
319
+ - name: tiktoken/gpt-4o
320
+ dtype: float64
321
+ - name: tokenmonster/englishcode-32000-consistent-v1
322
+ dtype: float64
323
+ - name: trimmed_cos_sim_to_canonical
324
+ struct:
325
+ - name: CohereLabs/aya-expanse-8b
326
+ dtype: float64
327
+ - name: Qwen/Qwen3-8B
328
+ dtype: float64
329
+ - name: bigscience/bloom
330
+ dtype: float64
331
+ - name: common-pile/comma-v0.1-1t
332
+ dtype: float64
333
+ - name: facebook/xglm-564M
334
+ dtype: float64
335
+ - name: google-bert/bert-base-multilingual-cased
336
+ dtype: float64
337
+ - name: google/byt5-small
338
+ dtype: float64
339
+ - name: google/gemma-2-2b
340
+ dtype: float64
341
+ - name: gpt2
342
+ dtype: float64
343
+ - name: meta-llama/Llama-3.2-1B
344
+ dtype: float64
345
+ - name: microsoft/Phi-3-mini-4k-instruct
346
+ dtype: float64
347
+ - name: mistralai/tekken
348
+ dtype: float64
349
+ - name: tiktoken/gpt-4o
350
+ dtype: float64
351
+ - name: tokenmonster/englishcode-32000-consistent-v1
352
+ dtype: float64
353
+ - name: token_counts
354
+ struct:
355
+ - name: CohereLabs/aya-expanse-8b
356
+ dtype: int64
357
+ - name: Qwen/Qwen3-8B
358
+ dtype: int64
359
+ - name: bigscience/bloom
360
+ dtype: int64
361
+ - name: common-pile/comma-v0.1-1t
362
+ dtype: int64
363
+ - name: facebook/xglm-564M
364
+ dtype: int64
365
+ - name: google-bert/bert-base-multilingual-cased
366
+ dtype: int64
367
+ - name: google/byt5-small
368
+ dtype: int64
369
+ - name: google/gemma-2-2b
370
+ dtype: int64
371
+ - name: gpt2
372
+ dtype: int64
373
+ - name: meta-llama/Llama-3.2-1B
374
+ dtype: int64
375
+ - name: microsoft/Phi-3-mini-4k-instruct
376
+ dtype: int64
377
+ - name: mistralai/tekken
378
+ dtype: int64
379
+ - name: tiktoken/gpt-4o
380
+ dtype: int64
381
+ - name: tokenmonster/englishcode-32000-consistent-v1
382
+ dtype: int64
383
+ splits:
384
+ - name: test
385
+ num_bytes: 4891
386
+ num_examples: 9
387
+ download_size: 30739
388
+ dataset_size: 4891
389
  configs:
390
  - config_name: tokenizer_robustness_completion_italian_abbreviations
391
  data_files:
 
395
  data_files:
396
  - split: test
397
  path: tokenizer_robustness_completion_italian_canonical/test-*
398
+ - config_name: tokenizer_robustness_completion_italian_capitalization
399
+ data_files:
400
+ - split: test
401
+ path: tokenizer_robustness_completion_italian_capitalization/test-*
402
  ---
403
 
404
  # Dataset Card for Tokenization Robustness
tokenizer_robustness_completion_italian_capitalization/test-00000-of-00001.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:24e0b94bb628aae02b394ea216752af76ea0f1279dab911a9726b5a1e3c9916e
3
- size 30749
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d275d0c1583d76d5e3645d7ab2c039c75b7e32b4c78e75ef88930ad78864b59
3
+ size 30739