David Pomerenke commited on
Commit
0638620
·
1 Parent(s): ba2a2f0
Files changed (2) hide show
  1. evals.py +1 -1
  2. results.json +244 -244
evals.py CHANGED
@@ -36,7 +36,7 @@ models = [
36
  "google/gemini-2.0-flash-001", # 0.4$/M tokens
37
  # "qwen/qwen-turbo", # 0.2$/M tokens; recognizes "inappropriate content"
38
  # "deepseek/deepseek-chat", # 0.9$/M tokens
39
- "microsoft/phi-4", # 0.07$/M tokens
40
  ]
41
  model_fast = "meta-llama/llama-3.3-70b-instruct"
42
  transcription_models = [
 
36
  "google/gemini-2.0-flash-001", # 0.4$/M tokens
37
  # "qwen/qwen-turbo", # 0.2$/M tokens; recognizes "inappropriate content"
38
  # "deepseek/deepseek-chat", # 0.9$/M tokens
39
+ "microsoft/phi-4", # 0.07$/M tokens; only 16k tokens context
40
  ]
41
  model_fast = "meta-llama/llama-3.3-70b-instruct"
42
  transcription_models = [
results.json CHANGED
@@ -7,61 +7,61 @@
7
  {
8
  "model": "openai/gpt-4o-mini",
9
  "model_type": "text-to-text",
10
- "mt_bleu": 0.5245466124037277,
11
- "mt_chrf": 65.25187717981981,
12
  "cls_acc": 0.5666666666666667,
13
- "mlm_chrf": 97.84704595784264,
14
- "t2t_score": 0.7325519660144305
15
  },
16
  {
17
  "model": "meta-llama/llama-3.3-70b-instruct",
18
  "model_type": "text-to-text",
19
- "mt_bleu": 0.48750797044187216,
20
- "mt_chrf": 63.24229348441665,
21
- "cls_acc": 0.6,
22
- "mlm_chrf": 93.62602669879945,
23
- "t2t_score": 0.7228944006107203
24
  },
25
  {
26
  "model": "mistralai/mistral-small-24b-instruct-2501",
27
  "model_type": "text-to-text",
28
- "mt_bleu": 0.486501959595472,
29
- "mt_chrf": 63.8187259254881,
30
- "cls_acc": 0.5333333333333333,
31
- "mlm_chrf": 79.91140615317198,
32
- "t2t_score": 0.656878218039978
33
  },
34
  {
35
  "model": "google/gemini-2.0-flash-001",
36
  "model_type": "text-to-text",
37
- "mt_bleu": 0.6060954569411976,
38
- "mt_chrf": 71.2288943066563,
39
  "cls_acc": 0.8666666666666667,
40
- "mlm_chrf": 98.79868693366329,
41
- "t2t_score": 0.8556474930232877
42
  },
43
  {
44
  "model": "microsoft/phi-4",
45
  "model_type": "text-to-text",
46
- "mt_bleu": 0.5199836121545649,
47
- "mt_chrf": 66.05410510011644,
48
- "cls_acc": 0.7,
49
- "mlm_chrf": 97.2290729316734,
50
- "t2t_score": 0.7776105934392995
51
  },
52
  {
53
  "model": "elevenlabs/scribe_v1",
54
  "model_type": "speech-to-text",
55
- "asr_wer": 0.34916319968417603,
56
- "asr_chrf": 78.55986690446153,
57
- "s2t_score": 39.45451505207285
58
  },
59
  {
60
  "model": "openai/whisper-large-v3",
61
  "model_type": "speech-to-text",
62
  "asr_wer": 0.25418986127300397,
63
- "asr_chrf": 86.52016887049808,
64
- "s2t_score": 43.387179365885544
65
  }
66
  ],
67
  "commonvoice_hours": 2651.0,
@@ -224,14 +224,14 @@
224
  "ZW": 6109446
225
  },
226
  "language_family": "Indo-European",
227
- "mt_bleu": 0.5249271223073668,
228
- "mt_chrf": 65.91917919929946,
229
- "cls_acc": 0.6533333333333333,
230
- "mlm_chrf": 93.48244773503015,
231
- "asr_wer": 0.30167653047859,
232
- "asr_chrf": 82.5400178874798,
233
- "t2t_score": 0.7491165342255433,
234
- "s2t_score": 41.4208472089792
235
  },
236
  {
237
  "language_name": "Chinese",
@@ -241,61 +241,61 @@
241
  {
242
  "model": "openai/gpt-4o-mini",
243
  "model_type": "text-to-text",
244
- "mt_bleu": 0.4887591978835982,
245
- "mt_chrf": 70.23584222732792,
246
- "cls_acc": 0.43333333333333335,
247
- "mlm_chrf": 91.54177259254774,
248
- "t2t_score": 0.6837031605106967
249
  },
250
  {
251
  "model": "meta-llama/llama-3.3-70b-instruct",
252
  "model_type": "text-to-text",
253
- "mt_bleu": 0.38557580495281013,
254
- "mt_chrf": 61.11151378837755,
255
  "cls_acc": 0.5666666666666667,
256
- "mlm_chrf": 94.55849047452216,
257
- "t2t_score": 0.7077889030985546
258
  },
259
  {
260
  "model": "mistralai/mistral-small-24b-instruct-2501",
261
  "model_type": "text-to-text",
262
- "mt_bleu": 0.48838341708028316,
263
- "mt_chrf": 71.48446350451079,
264
- "cls_acc": 0.7666666666666667,
265
- "mlm_chrf": 91.82889304045095,
266
- "t2t_score": 0.799933410705428
267
  },
268
  {
269
  "model": "google/gemini-2.0-flash-001",
270
  "model_type": "text-to-text",
271
- "mt_bleu": 0.392664198976078,
272
- "mt_chrf": 58.52460212908651,
273
  "cls_acc": 0.8666666666666667,
274
- "mlm_chrf": 94.26439448464122,
275
- "t2t_score": 0.7981855442679814
276
  },
277
  {
278
  "model": "microsoft/phi-4",
279
  "model_type": "text-to-text",
280
- "mt_bleu": 0.4892212172376435,
281
- "mt_chrf": 69.23023880157619,
282
- "cls_acc": 0.7,
283
- "mlm_chrf": 93.53779735205161,
284
- "t2t_score": 0.7758934538454261
285
  },
286
  {
287
  "model": "elevenlabs/scribe_v1",
288
  "model_type": "speech-to-text",
289
  "asr_wer": 1.0,
290
- "asr_chrf": 70.77419107011707,
291
- "s2t_score": 35.887095535058535
292
  },
293
  {
294
  "model": "openai/whisper-large-v3",
295
  "model_type": "speech-to-text",
296
  "asr_wer": 1.0,
297
- "asr_chrf": 77.48220275963784,
298
- "s2t_score": 39.24110137981892
299
  }
300
  ],
301
  "commonvoice_hours": 422.0,
@@ -323,14 +323,14 @@
323
  "VN": 1085934
324
  },
325
  "language_family": "Sino-Tibetan",
326
- "mt_bleu": 0.4489207672260826,
327
- "mt_chrf": 66.1173320901758,
328
- "cls_acc": 0.6666666666666666,
329
- "mlm_chrf": 93.14626958884273,
330
  "asr_wer": 1.0,
331
- "asr_chrf": 74.12819691487746,
332
- "t2t_score": 0.7531008944856173,
333
- "s2t_score": 37.56409845743873
334
  },
335
  {
336
  "language_name": "Hindi",
@@ -340,61 +340,61 @@
340
  {
341
  "model": "openai/gpt-4o-mini",
342
  "model_type": "text-to-text",
343
- "mt_bleu": 0.3647010036099328,
344
- "mt_chrf": 55.294100726869324,
345
  "cls_acc": 0.5,
346
- "mlm_chrf": 93.01665261992896,
347
- "t2t_score": 0.6610358444893277
348
  },
349
  {
350
  "model": "meta-llama/llama-3.3-70b-instruct",
351
  "model_type": "text-to-text",
352
- "mt_bleu": 0.32404902340686065,
353
- "mt_chrf": 53.54085104449268,
354
- "cls_acc": 0.5,
355
- "mlm_chrf": 96.17240172798218,
356
- "t2t_score": 0.6657108425749162
357
  },
358
  {
359
  "model": "mistralai/mistral-small-24b-instruct-2501",
360
  "model_type": "text-to-text",
361
- "mt_bleu": 0.346304931512649,
362
- "mt_chrf": 52.59738987149503,
363
  "cls_acc": 0.5,
364
- "mlm_chrf": 90.59820426203353,
365
- "t2t_score": 0.6439853137784286
366
  },
367
  {
368
  "model": "google/gemini-2.0-flash-001",
369
  "model_type": "text-to-text",
370
- "mt_bleu": 0.43820082761650603,
371
- "mt_chrf": 61.68241787594198,
372
  "cls_acc": 0.8666666666666667,
373
- "mlm_chrf": 96.5534261573122,
374
- "t2t_score": 0.8163417023330695
375
  },
376
  {
377
  "model": "microsoft/phi-4",
378
  "model_type": "text-to-text",
379
- "mt_bleu": 0.388038924689894,
380
- "mt_chrf": 56.13147190849697,
381
  "cls_acc": 0.8,
382
- "mlm_chrf": 94.47772196307584,
383
- "t2t_score": 0.7686973129052427
384
  },
385
  {
386
  "model": "elevenlabs/scribe_v1",
387
  "model_type": "speech-to-text",
388
- "asr_wer": 0.2338948365728121,
389
- "asr_chrf": 83.13389660250954,
390
- "s2t_score": 41.68389571954118
391
  },
392
  {
393
  "model": "openai/whisper-large-v3",
394
  "model_type": "speech-to-text",
395
  "asr_wer": 0.43522263872986894,
396
- "asr_chrf": 63.049286642268754,
397
- "s2t_score": 31.74225464049931
398
  }
399
  ],
400
  "commonvoice_hours": 16.0,
@@ -408,14 +408,14 @@
408
  "ZA": 1129272
409
  },
410
  "language_family": "Indo-European",
411
- "mt_bleu": 0.3722589421671684,
412
- "mt_chrf": 55.849246285459195,
413
- "cls_acc": 0.6333333333333334,
414
- "mlm_chrf": 94.16368134606655,
415
- "asr_wer": 0.33455873765134053,
416
- "asr_chrf": 73.09159162238915,
417
- "t2t_score": 0.7111542032161969,
418
- "s2t_score": 36.71307518002024
419
  },
420
  {
421
  "language_name": "Spanish",
@@ -425,61 +425,61 @@
425
  {
426
  "model": "openai/gpt-4o-mini",
427
  "model_type": "text-to-text",
428
- "mt_bleu": 0.3313892289629975,
429
- "mt_chrf": 52.582336167000754,
430
  "cls_acc": 0.6,
431
- "mlm_chrf": 96.34068599793908,
432
- "t2t_score": 0.6964100738831327
433
  },
434
  {
435
  "model": "meta-llama/llama-3.3-70b-instruct",
436
  "model_type": "text-to-text",
437
- "mt_bleu": 0.31587937116142056,
438
- "mt_chrf": 52.142851262301726,
439
- "cls_acc": 0.5,
440
- "mlm_chrf": 96.92768852306384,
441
- "t2t_score": 0.6635684659512185
442
  },
443
  {
444
  "model": "mistralai/mistral-small-24b-instruct-2501",
445
  "model_type": "text-to-text",
446
- "mt_bleu": 0.3120516523386294,
447
- "mt_chrf": 50.14036662376262,
448
  "cls_acc": 0.6333333333333333,
449
- "mlm_chrf": 93.9463761308063,
450
- "t2t_score": 0.6914002536263407
451
  },
452
  {
453
  "model": "google/gemini-2.0-flash-001",
454
  "model_type": "text-to-text",
455
- "mt_bleu": 0.3318101035886167,
456
- "mt_chrf": 53.80973313056067,
457
  "cls_acc": 0.8666666666666667,
458
- "mlm_chrf": 96.46210398041012,
459
- "t2t_score": 0.7897950125921249
460
  },
461
  {
462
  "model": "microsoft/phi-4",
463
  "model_type": "text-to-text",
464
- "mt_bleu": 0.2808051746473726,
465
- "mt_chrf": 49.858821063770044,
466
  "cls_acc": 0.7666666666666667,
467
- "mlm_chrf": 96.28984542933327,
468
- "t2t_score": 0.7427177771992333
469
  },
470
  {
471
  "model": "elevenlabs/scribe_v1",
472
  "model_type": "speech-to-text",
473
- "asr_wer": 0.19653905528613333,
474
- "asr_chrf": 89.3473231669277,
475
- "s2t_score": 44.771931111106916
476
  },
477
  {
478
  "model": "openai/whisper-large-v3",
479
  "model_type": "speech-to-text",
480
  "asr_wer": 0.17561491933862197,
481
- "asr_chrf": 91.85915393819565,
482
- "s2t_score": 46.01738442876714
483
  }
484
  ],
485
  "commonvoice_hours": 446.0,
@@ -526,14 +526,14 @@
526
  "VE": 23488572
527
  },
528
  "language_family": "Indo-European",
529
- "mt_bleu": 0.31438710613980736,
530
- "mt_chrf": 51.70682164947916,
531
- "cls_acc": 0.6733333333333333,
532
- "mlm_chrf": 95.99334001231053,
533
- "asr_wer": 0.18607698731237765,
534
- "asr_chrf": 90.60323855256166,
535
- "t2t_score": 0.71677831665041,
536
- "s2t_score": 45.39465776993703
537
  },
538
  {
539
  "language_name": "Arabic",
@@ -543,61 +543,61 @@
543
  {
544
  "model": "openai/gpt-4o-mini",
545
  "model_type": "text-to-text",
546
- "mt_bleu": 0.312307215788044,
547
- "mt_chrf": 51.50028782321851,
548
  "cls_acc": 0.6,
549
- "mlm_chrf": 96.57677413527875,
550
- "t2t_score": 0.6935902065283241
551
  },
552
  {
553
  "model": "meta-llama/llama-3.3-70b-instruct",
554
  "model_type": "text-to-text",
555
- "mt_bleu": 0.39547934933771334,
556
- "mt_chrf": 57.51652731936118,
557
  "cls_acc": 0.4666666666666667,
558
- "mlm_chrf": 94.97026443937914,
559
- "t2t_score": 0.6638448614180232
560
  },
561
  {
562
  "model": "mistralai/mistral-small-24b-instruct-2501",
563
  "model_type": "text-to-text",
564
- "mt_bleu": 0.26940670420361396,
565
- "mt_chrf": 46.07654794208784,
566
  "cls_acc": 0.7,
567
- "mlm_chrf": 90.92876728332028,
568
- "t2t_score": 0.690017717418027
569
  },
570
  {
571
  "model": "google/gemini-2.0-flash-001",
572
  "model_type": "text-to-text",
573
- "mt_bleu": 0.3928034519711188,
574
- "mt_chrf": 57.47507639595937,
575
  "cls_acc": 0.9,
576
- "mlm_chrf": 96.82551362297947,
577
- "t2t_score": 0.8143353000631294
578
  },
579
  {
580
  "model": "microsoft/phi-4",
581
  "model_type": "text-to-text",
582
- "mt_bleu": 0.2700754695807907,
583
- "mt_chrf": 45.896533062231306,
584
  "cls_acc": 0.7,
585
- "mlm_chrf": 93.60903960225609,
586
- "t2t_score": 0.6983519088816247
587
  },
588
  {
589
  "model": "elevenlabs/scribe_v1",
590
  "model_type": "speech-to-text",
591
- "asr_wer": 0.2685436379713873,
592
- "asr_chrf": 83.61712973768607,
593
- "s2t_score": 41.94283668782873
594
  },
595
  {
596
  "model": "openai/whisper-large-v3",
597
  "model_type": "speech-to-text",
598
  "asr_wer": 0.17370718156523782,
599
- "asr_chrf": 88.68973505122798,
600
- "s2t_score": 44.43172111639661
601
  }
602
  ],
603
  "commonvoice_hours": 91.0,
@@ -643,14 +643,14 @@
643
  "YE": 22114456
644
  },
645
  "language_family": "Afro-Asiatic",
646
- "mt_bleu": 0.32801443817625614,
647
- "mt_chrf": 51.692994508571644,
648
  "cls_acc": 0.6733333333333332,
649
- "mlm_chrf": 94.58207181664275,
650
- "asr_wer": 0.22112540976831258,
651
- "asr_chrf": 86.15343239445703,
652
- "t2t_score": 0.7120279988618257,
653
- "s2t_score": 43.187278902112666
654
  },
655
  {
656
  "language_name": "Urdu",
@@ -660,18 +660,18 @@
660
  {
661
  "model": "meta-llama/llama-3.3-70b-instruct",
662
  "model_type": "text-to-text",
663
- "mt_bleu": 0.26585004461425726,
664
- "mt_chrf": 47.37157150967947,
665
- "cls_acc": 0.43333333333333335,
666
- "mlm_chrf": 94.38802161979918,
667
- "t2t_score": 0.6169764215427066
668
  },
669
  {
670
  "model": "elevenlabs/scribe_v1",
671
  "model_type": "speech-to-text",
672
- "asr_wer": 0.2982973325975355,
673
- "asr_chrf": 78.04754030614318,
674
- "s2t_score": 39.17291881937036
675
  }
676
  ],
677
  "commonvoice_hours": 77.0,
@@ -684,14 +684,14 @@
684
  "PK": 221825950
685
  },
686
  "language_family": "Indo-European",
687
- "mt_bleu": 0.26585004461425726,
688
- "mt_chrf": 47.37157150967947,
689
- "cls_acc": 0.43333333333333335,
690
- "mlm_chrf": 94.38802161979918,
691
- "asr_wer": 0.2982973325975355,
692
- "asr_chrf": 78.04754030614318,
693
- "t2t_score": 0.6169764215427066,
694
- "s2t_score": 39.17291881937036
695
  },
696
  {
697
  "language_name": "French",
@@ -701,18 +701,18 @@
701
  {
702
  "model": "meta-llama/llama-3.3-70b-instruct",
703
  "model_type": "text-to-text",
704
- "mt_bleu": 0.3510210872150948,
705
- "mt_chrf": 55.795595938804894,
706
  "cls_acc": 0.5666666666666667,
707
- "mlm_chrf": 97.12318847922649,
708
- "t2t_score": 0.6986181702823268
709
  },
710
  {
711
  "model": "elevenlabs/scribe_v1",
712
  "model_type": "speech-to-text",
713
  "asr_wer": 0.2610754929736961,
714
- "asr_chrf": 87.75590287945104,
715
- "s2t_score": 44.00848918621237
716
  }
717
  ],
718
  "commonvoice_hours": 1052.0,
@@ -782,14 +782,14 @@
782
  "YT": 110580
783
  },
784
  "language_family": "Indo-European",
785
- "mt_bleu": 0.3510210872150948,
786
- "mt_chrf": 55.795595938804894,
787
  "cls_acc": 0.5666666666666667,
788
- "mlm_chrf": 97.12318847922649,
789
  "asr_wer": 0.2610754929736961,
790
- "asr_chrf": 87.75590287945104,
791
- "t2t_score": 0.6986181702823268,
792
- "s2t_score": 44.00848918621237
793
  },
794
  {
795
  "language_name": "Bangla",
@@ -799,18 +799,18 @@
799
  {
800
  "model": "meta-llama/llama-3.3-70b-instruct",
801
  "model_type": "text-to-text",
802
- "mt_bleu": 0.2874920154082786,
803
- "mt_chrf": 51.49279116112809,
804
- "cls_acc": 0.4,
805
- "mlm_chrf": 90.6067262108039,
806
- "t2t_score": 0.6069983912397733
807
  },
808
  {
809
  "model": "elevenlabs/scribe_v1",
810
  "model_type": "speech-to-text",
811
- "asr_wer": 0.26686188207927336,
812
- "asr_chrf": 85.49430743996201,
813
- "s2t_score": 42.88058466102064
814
  }
815
  ],
816
  "commonvoice_hours": 49.0,
@@ -823,14 +823,14 @@
823
  "NP": 28508
824
  },
825
  "language_family": "Indo-European",
826
- "mt_bleu": 0.2874920154082786,
827
- "mt_chrf": 51.49279116112809,
828
- "cls_acc": 0.4,
829
- "mlm_chrf": 90.6067262108039,
830
- "asr_wer": 0.26686188207927336,
831
- "asr_chrf": 85.49430743996201,
832
- "t2t_score": 0.6069983912397733,
833
- "s2t_score": 42.88058466102064
834
  },
835
  {
836
  "language_name": "Portuguese",
@@ -840,18 +840,18 @@
840
  {
841
  "model": "meta-llama/llama-3.3-70b-instruct",
842
  "model_type": "text-to-text",
843
- "mt_bleu": 0.33491649454450034,
844
- "mt_chrf": 54.60211868234021,
845
  "cls_acc": 0.5666666666666667,
846
- "mlm_chrf": 96.52676764996336,
847
- "t2t_score": 0.6926518433299008
848
  },
849
  {
850
  "model": "elevenlabs/scribe_v1",
851
  "model_type": "speech-to-text",
852
- "asr_wer": 0.22967756370402836,
853
- "asr_chrf": 89.03888828875101,
854
- "s2t_score": 44.63428292622752
855
  }
856
  ],
857
  "commonvoice_hours": 177.0,
@@ -875,14 +875,14 @@
875
  "TL": 816395
876
  },
877
  "language_family": "Indo-European",
878
- "mt_bleu": 0.33491649454450034,
879
- "mt_chrf": 54.60211868234021,
880
  "cls_acc": 0.5666666666666667,
881
- "mlm_chrf": 96.52676764996336,
882
- "asr_wer": 0.22967756370402836,
883
- "asr_chrf": 89.03888828875101,
884
- "t2t_score": 0.6926518433299008,
885
- "s2t_score": 44.63428292622752
886
  },
887
  {
888
  "language_name": "Punjabi",
@@ -892,18 +892,18 @@
892
  {
893
  "model": "meta-llama/llama-3.3-70b-instruct",
894
  "model_type": "text-to-text",
895
- "mt_bleu": 0.3078917767345886,
896
- "mt_chrf": 50.505686987696365,
897
- "cls_acc": 0.5333333333333333,
898
- "mlm_chrf": 90.10119297923285,
899
- "t2t_score": 0.6464673776675418
900
  },
901
  {
902
  "model": "elevenlabs/scribe_v1",
903
  "model_type": "speech-to-text",
904
- "asr_wer": 0.20953788908863977,
905
- "asr_chrf": 84.56957135963033,
906
- "s2t_score": 42.389554624359484
907
  }
908
  ],
909
  "commonvoice_hours": 2.3,
@@ -917,14 +917,14 @@
917
  "SG": 9314
918
  },
919
  "language_family": "Indo-European",
920
- "mt_bleu": 0.3078917767345886,
921
- "mt_chrf": 50.505686987696365,
922
- "cls_acc": 0.5333333333333333,
923
- "mlm_chrf": 90.10119297923285,
924
- "asr_wer": 0.20953788908863977,
925
- "asr_chrf": 84.56957135963033,
926
- "t2t_score": 0.6464673776675418,
927
- "s2t_score": 42.389554624359484
928
  },
929
  {
930
  "language_name": "Russian",
 
7
  {
8
  "model": "openai/gpt-4o-mini",
9
  "model_type": "text-to-text",
10
+ "mt_bleu": 0.40476391184425853,
11
+ "mt_chrf": 0.5350771248567329,
12
  "cls_acc": 0.5666666666666667,
13
+ "mlm_chrf": 0.9792909796569321,
14
+ "t2t_score": 0.6936782570601107
15
  },
16
  {
17
  "model": "meta-llama/llama-3.3-70b-instruct",
18
  "model_type": "text-to-text",
19
+ "mt_bleu": 0.42039857566100747,
20
+ "mt_chrf": 0.5300571003447904,
21
+ "cls_acc": 0.5,
22
+ "mlm_chrf": 0.9378674673892118,
23
+ "t2t_score": 0.6559748559113341
24
  },
25
  {
26
  "model": "mistralai/mistral-small-24b-instruct-2501",
27
  "model_type": "text-to-text",
28
+ "mt_bleu": 0.39023213290403214,
29
+ "mt_chrf": 0.501310810487581,
30
+ "cls_acc": 0.43333333333333335,
31
+ "mlm_chrf": 0.7876173555317149,
32
+ "t2t_score": 0.5740871664508763
33
  },
34
  {
35
  "model": "google/gemini-2.0-flash-001",
36
  "model_type": "text-to-text",
37
+ "mt_bleu": 0.5005636231739998,
38
+ "mt_chrf": 0.6083608181074219,
39
  "cls_acc": 0.8666666666666667,
40
+ "mlm_chrf": 0.9820612175447262,
41
+ "t2t_score": 0.819029567439605
42
  },
43
  {
44
  "model": "microsoft/phi-4",
45
  "model_type": "text-to-text",
46
+ "mt_bleu": 0.37572897166941227,
47
+ "mt_chrf": 0.5070735300311086,
48
+ "cls_acc": 0.6666666666666666,
49
+ "mlm_chrf": 0.969030413937307,
50
+ "t2t_score": 0.714256870211694
51
  },
52
  {
53
  "model": "elevenlabs/scribe_v1",
54
  "model_type": "speech-to-text",
55
+ "asr_wer": 0.32249653301750936,
56
+ "asr_chrf": 0.8168177351797856,
57
+ "s2t_score": 0.5696571340986475
58
  },
59
  {
60
  "model": "openai/whisper-large-v3",
61
  "model_type": "speech-to-text",
62
  "asr_wer": 0.25418986127300397,
63
+ "asr_chrf": 0.8652016887049812,
64
+ "s2t_score": 0.5596957749889926
65
  }
66
  ],
67
  "commonvoice_hours": 2651.0,
 
224
  "ZW": 6109446
225
  },
226
  "language_family": "Indo-European",
227
+ "mt_bleu": 0.41833744305054205,
228
+ "mt_chrf": 0.5363758767655269,
229
+ "cls_acc": 0.6066666666666667,
230
+ "mlm_chrf": 0.9311734868119783,
231
+ "asr_wer": 0.2883431971452567,
232
+ "asr_chrf": 0.8410097119423834,
233
+ "t2t_score": 0.691405343414724,
234
+ "s2t_score": 0.5646764545438201
235
  },
236
  {
237
  "language_name": "Chinese",
 
241
  {
242
  "model": "openai/gpt-4o-mini",
243
  "model_type": "text-to-text",
244
+ "mt_bleu": 0.37010453186602793,
245
+ "mt_chrf": 0.5495542857606343,
246
+ "cls_acc": 0.4666666666666667,
247
+ "mlm_chrf": 0.900779581057449,
248
+ "t2t_score": 0.63900017782825
249
  },
250
  {
251
  "model": "meta-llama/llama-3.3-70b-instruct",
252
  "model_type": "text-to-text",
253
+ "mt_bleu": 0.3904811079178694,
254
+ "mt_chrf": 0.5665464429871689,
255
  "cls_acc": 0.5666666666666667,
256
+ "mlm_chrf": 0.9044414344868564,
257
+ "t2t_score": 0.6792181813802306
258
  },
259
  {
260
  "model": "mistralai/mistral-small-24b-instruct-2501",
261
  "model_type": "text-to-text",
262
+ "mt_bleu": 0.3631466653543813,
263
+ "mt_chrf": 0.552006597905903,
264
+ "cls_acc": 0.6666666666666666,
265
+ "mlm_chrf": 0.8851414000070698,
266
+ "t2t_score": 0.7012715548598797
267
  },
268
  {
269
  "model": "google/gemini-2.0-flash-001",
270
  "model_type": "text-to-text",
271
+ "mt_bleu": 0.41028031844473206,
272
+ "mt_chrf": 0.5607632287491137,
273
  "cls_acc": 0.8666666666666667,
274
+ "mlm_chrf": 0.9353777720326482,
275
+ "t2t_score": 0.7876025558161427
276
  },
277
  {
278
  "model": "microsoft/phi-4",
279
  "model_type": "text-to-text",
280
+ "mt_bleu": 0.3516110848479108,
281
+ "mt_chrf": 0.5536738005415017,
282
+ "cls_acc": 0.7333333333333333,
283
+ "mlm_chrf": 0.9080915180096581,
284
+ "t2t_score": 0.7316995506281644
285
  },
286
  {
287
  "model": "elevenlabs/scribe_v1",
288
  "model_type": "speech-to-text",
289
  "asr_wer": 1.0,
290
+ "asr_chrf": 0.7077419107011705,
291
+ "s2t_score": 0.8538709553505852
292
  },
293
  {
294
  "model": "openai/whisper-large-v3",
295
  "model_type": "speech-to-text",
296
  "asr_wer": 1.0,
297
+ "asr_chrf": 0.7748220275963785,
298
+ "s2t_score": 0.8874110137981892
299
  }
300
  ],
301
  "commonvoice_hours": 422.0,
 
323
  "VN": 1085934
324
  },
325
  "language_family": "Sino-Tibetan",
326
+ "mt_bleu": 0.3771247416861843,
327
+ "mt_chrf": 0.5565088711888644,
328
+ "cls_acc": 0.6599999999999999,
329
+ "mlm_chrf": 0.9067663411187363,
330
  "asr_wer": 1.0,
331
+ "asr_chrf": 0.7412819691487745,
332
+ "t2t_score": 0.7077584041025334,
333
+ "s2t_score": 0.8706409845743872
334
  },
335
  {
336
  "language_name": "Hindi",
 
340
  {
341
  "model": "openai/gpt-4o-mini",
342
  "model_type": "text-to-text",
343
+ "mt_bleu": 0.30348099344592144,
344
+ "mt_chrf": 0.46828573138186164,
345
  "cls_acc": 0.5,
346
+ "mlm_chrf": 0.9491600973221963,
347
+ "t2t_score": 0.6391486095680193
348
  },
349
  {
350
  "model": "meta-llama/llama-3.3-70b-instruct",
351
  "model_type": "text-to-text",
352
+ "mt_bleu": 0.30694468553453214,
353
+ "mt_chrf": 0.4717417673162039,
354
+ "cls_acc": 0.5666666666666667,
355
+ "mlm_chrf": 0.9225335146094515,
356
+ "t2t_score": 0.6536473161974407
357
  },
358
  {
359
  "model": "mistralai/mistral-small-24b-instruct-2501",
360
  "model_type": "text-to-text",
361
+ "mt_bleu": 0.21739461408317115,
362
+ "mt_chrf": 0.35902002421447704,
363
  "cls_acc": 0.5,
364
+ "mlm_chrf": 0.8896017448529104,
365
+ "t2t_score": 0.5828739230224625
366
  },
367
  {
368
  "model": "google/gemini-2.0-flash-001",
369
  "model_type": "text-to-text",
370
+ "mt_bleu": 0.34270244014502865,
371
+ "mt_chrf": 0.5241272552859138,
372
  "cls_acc": 0.8666666666666667,
373
+ "mlm_chrf": 0.9648559476590244,
374
+ "t2t_score": 0.7852166232038682
375
  },
376
  {
377
  "model": "microsoft/phi-4",
378
  "model_type": "text-to-text",
379
+ "mt_bleu": 0.27124364337229007,
380
+ "mt_chrf": 0.43849820754378105,
381
  "cls_acc": 0.8,
382
+ "mlm_chrf": 0.9387109966783314,
383
+ "t2t_score": 0.7257364014073708
384
  },
385
  {
386
  "model": "elevenlabs/scribe_v1",
387
  "model_type": "speech-to-text",
388
+ "asr_wer": 0.28107684996172205,
389
+ "asr_chrf": 0.7778871091355449,
390
+ "s2t_score": 0.5294819795486335
391
  },
392
  {
393
  "model": "openai/whisper-large-v3",
394
  "model_type": "speech-to-text",
395
  "asr_wer": 0.43522263872986894,
396
+ "asr_chrf": 0.6304928664226873,
397
+ "s2t_score": 0.5328577525762781
398
  }
399
  ],
400
  "commonvoice_hours": 16.0,
 
408
  "ZA": 1129272
409
  },
410
  "language_family": "Indo-European",
411
+ "mt_bleu": 0.2883532753161887,
412
+ "mt_chrf": 0.4523345971484474,
413
+ "cls_acc": 0.6466666666666667,
414
+ "mlm_chrf": 0.9329724602243828,
415
+ "asr_wer": 0.3581497443457955,
416
+ "asr_chrf": 0.7041899877791161,
417
+ "t2t_score": 0.6773245746798323,
418
+ "s2t_score": 0.5311698660624558
419
  },
420
  {
421
  "language_name": "Spanish",
 
425
  {
426
  "model": "openai/gpt-4o-mini",
427
  "model_type": "text-to-text",
428
+ "mt_bleu": 0.3047910872340331,
429
+ "mt_chrf": 0.4664322514490113,
430
  "cls_acc": 0.6,
431
+ "mlm_chrf": 0.9773594806492754,
432
+ "t2t_score": 0.6812639106994288
433
  },
434
  {
435
  "model": "meta-llama/llama-3.3-70b-instruct",
436
  "model_type": "text-to-text",
437
+ "mt_bleu": 0.2681893620935732,
438
+ "mt_chrf": 0.4507327709042758,
439
+ "cls_acc": 0.5333333333333333,
440
+ "mlm_chrf": 0.9387079449855046,
441
+ "t2t_score": 0.6409246830743712
442
  },
443
  {
444
  "model": "mistralai/mistral-small-24b-instruct-2501",
445
  "model_type": "text-to-text",
446
+ "mt_bleu": 0.2794753489249633,
447
+ "mt_chrf": 0.4300653056934316,
448
  "cls_acc": 0.6333333333333333,
449
+ "mlm_chrf": 0.9521615389464084,
450
+ "t2t_score": 0.6718533926577245
451
  },
452
  {
453
  "model": "google/gemini-2.0-flash-001",
454
  "model_type": "text-to-text",
455
+ "mt_bleu": 0.34138272852201396,
456
+ "mt_chrf": 0.5025141961861047,
457
  "cls_acc": 0.8666666666666667,
458
+ "mlm_chrf": 0.9816202595213187,
459
+ "t2t_score": 0.7836003741246967
460
  },
461
  {
462
  "model": "microsoft/phi-4",
463
  "model_type": "text-to-text",
464
+ "mt_bleu": 0.2422835449771352,
465
+ "mt_chrf": 0.4339464294262581,
466
  "cls_acc": 0.7666666666666667,
467
+ "mlm_chrf": 0.9650899128314744,
468
+ "t2t_score": 0.7219010029747998
469
  },
470
  {
471
  "model": "elevenlabs/scribe_v1",
472
  "model_type": "speech-to-text",
473
+ "asr_wer": 0.1927295314766095,
474
+ "asr_chrf": 0.8941604003451139,
475
+ "s2t_score": 0.5434449659108617
476
  },
477
  {
478
  "model": "openai/whisper-large-v3",
479
  "model_type": "speech-to-text",
480
  "asr_wer": 0.17561491933862197,
481
+ "asr_chrf": 0.9185915393819567,
482
+ "s2t_score": 0.5471032293602893
483
  }
484
  ],
485
  "commonvoice_hours": 446.0,
 
526
  "VE": 23488572
527
  },
528
  "language_family": "Indo-European",
529
+ "mt_bleu": 0.28722441435034374,
530
+ "mt_chrf": 0.4567381907318164,
531
+ "cls_acc": 0.6799999999999999,
532
+ "mlm_chrf": 0.9629878273867962,
533
+ "asr_wer": 0.18417222540761574,
534
+ "asr_chrf": 0.9063759698635353,
535
+ "t2t_score": 0.6999086727062043,
536
+ "s2t_score": 0.5452740976355754
537
  },
538
  {
539
  "language_name": "Arabic",
 
543
  {
544
  "model": "openai/gpt-4o-mini",
545
  "model_type": "text-to-text",
546
+ "mt_bleu": 0.26593690487060095,
547
+ "mt_chrf": 0.46048583874582116,
548
  "cls_acc": 0.6,
549
+ "mlm_chrf": 0.9386512881662338,
550
+ "t2t_score": 0.6663790423040182
551
  },
552
  {
553
  "model": "meta-llama/llama-3.3-70b-instruct",
554
  "model_type": "text-to-text",
555
+ "mt_bleu": 0.2811587891139184,
556
+ "mt_chrf": 0.4610903378292814,
557
  "cls_acc": 0.4666666666666667,
558
+ "mlm_chrf": 0.9425014853511972,
559
+ "t2t_score": 0.623419496615715
560
  },
561
  {
562
  "model": "mistralai/mistral-small-24b-instruct-2501",
563
  "model_type": "text-to-text",
564
+ "mt_bleu": 0.20683544337550522,
565
+ "mt_chrf": 0.3857194391808861,
566
  "cls_acc": 0.7,
567
+ "mlm_chrf": 0.9148307958510079,
568
+ "t2t_score": 0.6668500783439647
569
  },
570
  {
571
  "model": "google/gemini-2.0-flash-001",
572
  "model_type": "text-to-text",
573
+ "mt_bleu": 0.3368489944570755,
574
+ "mt_chrf": 0.5132383488065384,
575
  "cls_acc": 0.9,
576
+ "mlm_chrf": 0.97921999148367,
577
+ "t2t_score": 0.7974861134300695
578
  },
579
  {
580
  "model": "microsoft/phi-4",
581
  "model_type": "text-to-text",
582
+ "mt_bleu": 0.2419401543819311,
583
+ "mt_chrf": 0.4175700747114916,
584
  "cls_acc": 0.7,
585
+ "mlm_chrf": 0.9332394461837199,
586
+ "t2t_score": 0.6836031736317372
587
  },
588
  {
589
  "model": "elevenlabs/scribe_v1",
590
  "model_type": "speech-to-text",
591
+ "asr_wer": 0.2976757980818083,
592
+ "asr_chrf": 0.8082975812173028,
593
+ "s2t_score": 0.5529866896495556
594
  },
595
  {
596
  "model": "openai/whisper-large-v3",
597
  "model_type": "speech-to-text",
598
  "asr_wer": 0.17370718156523782,
599
+ "asr_chrf": 0.8868973505122795,
600
+ "s2t_score": 0.5303022660387586
601
  }
602
  ],
603
  "commonvoice_hours": 91.0,
 
643
  "YE": 22114456
644
  },
645
  "language_family": "Afro-Asiatic",
646
+ "mt_bleu": 0.26654405723980623,
647
+ "mt_chrf": 0.44762080785480374,
648
  "cls_acc": 0.6733333333333332,
649
+ "mlm_chrf": 0.9416886014071657,
650
+ "asr_wer": 0.23569148982352306,
651
+ "asr_chrf": 0.8475974658647911,
652
+ "t2t_score": 0.6875475808651009,
653
+ "s2t_score": 0.5416444778441571
654
  },
655
  {
656
  "language_name": "Urdu",
 
660
  {
661
  "model": "meta-llama/llama-3.3-70b-instruct",
662
  "model_type": "text-to-text",
663
+ "mt_bleu": 0.21847309277555946,
664
+ "mt_chrf": 0.4036061618297173,
665
+ "cls_acc": 0.4666666666666667,
666
+ "mlm_chrf": 0.9330700955297437,
667
+ "t2t_score": 0.6011143080087092
668
  },
669
  {
670
  "model": "elevenlabs/scribe_v1",
671
  "model_type": "speech-to-text",
672
+ "asr_wer": 0.33586677704198,
673
+ "asr_chrf": 0.7430938264813908,
674
+ "s2t_score": 0.5394803017616854
675
  }
676
  ],
677
  "commonvoice_hours": 77.0,
 
684
  "PK": 221825950
685
  },
686
  "language_family": "Indo-European",
687
+ "mt_bleu": 0.21847309277555946,
688
+ "mt_chrf": 0.4036061618297173,
689
+ "cls_acc": 0.4666666666666667,
690
+ "mlm_chrf": 0.9330700955297437,
691
+ "asr_wer": 0.33586677704198,
692
+ "asr_chrf": 0.7430938264813908,
693
+ "t2t_score": 0.6011143080087092,
694
+ "s2t_score": 0.5394803017616854
695
  },
696
  {
697
  "language_name": "French",
 
701
  {
702
  "model": "meta-llama/llama-3.3-70b-instruct",
703
  "model_type": "text-to-text",
704
+ "mt_bleu": 0.32618133837885355,
705
+ "mt_chrf": 0.4952644559037655,
706
  "cls_acc": 0.5666666666666667,
707
+ "mlm_chrf": 0.9639910578331403,
708
+ "t2t_score": 0.6753073934678575
709
  },
710
  {
711
  "model": "elevenlabs/scribe_v1",
712
  "model_type": "speech-to-text",
713
  "asr_wer": 0.2610754929736961,
714
+ "asr_chrf": 0.8775590287945104,
715
+ "s2t_score": 0.5693172608841033
716
  }
717
  ],
718
  "commonvoice_hours": 1052.0,
 
782
  "YT": 110580
783
  },
784
  "language_family": "Indo-European",
785
+ "mt_bleu": 0.32618133837885355,
786
+ "mt_chrf": 0.4952644559037655,
787
  "cls_acc": 0.5666666666666667,
788
+ "mlm_chrf": 0.9639910578331403,
789
  "asr_wer": 0.2610754929736961,
790
+ "asr_chrf": 0.8775590287945104,
791
+ "t2t_score": 0.6753073934678575,
792
+ "s2t_score": 0.5693172608841033
793
  },
794
  {
795
  "language_name": "Bangla",
 
799
  {
800
  "model": "meta-llama/llama-3.3-70b-instruct",
801
  "model_type": "text-to-text",
802
+ "mt_bleu": 0.23230854865261916,
803
+ "mt_chrf": 0.42063939652271853,
804
+ "cls_acc": 0.36666666666666664,
805
+ "mlm_chrf": 0.8995877938471141,
806
+ "t2t_score": 0.5622979523454997
807
  },
808
  {
809
  "model": "elevenlabs/scribe_v1",
810
  "model_type": "speech-to-text",
811
+ "asr_wer": 0.3066054718228631,
812
+ "asr_chrf": 0.8006938100379883,
813
+ "s2t_score": 0.5536496409304257
814
  }
815
  ],
816
  "commonvoice_hours": 49.0,
 
823
  "NP": 28508
824
  },
825
  "language_family": "Indo-European",
826
+ "mt_bleu": 0.23230854865261916,
827
+ "mt_chrf": 0.42063939652271853,
828
+ "cls_acc": 0.36666666666666664,
829
+ "mlm_chrf": 0.8995877938471141,
830
+ "asr_wer": 0.3066054718228631,
831
+ "asr_chrf": 0.8006938100379883,
832
+ "t2t_score": 0.5622979523454997,
833
+ "s2t_score": 0.5536496409304257
834
  },
835
  {
836
  "language_name": "Portuguese",
 
840
  {
841
  "model": "meta-llama/llama-3.3-70b-instruct",
842
  "model_type": "text-to-text",
843
+ "mt_bleu": 0.2806919135940658,
844
+ "mt_chrf": 0.4531472771818437,
845
  "cls_acc": 0.5666666666666667,
846
+ "mlm_chrf": 0.9630716853128435,
847
+ "t2t_score": 0.6609618763871179
848
  },
849
  {
850
  "model": "elevenlabs/scribe_v1",
851
  "model_type": "speech-to-text",
852
+ "asr_wer": 0.22800492332171055,
853
+ "asr_chrf": 0.8922038015648965,
854
+ "s2t_score": 0.5601043624433035
855
  }
856
  ],
857
  "commonvoice_hours": 177.0,
 
875
  "TL": 816395
876
  },
877
  "language_family": "Indo-European",
878
+ "mt_bleu": 0.2806919135940658,
879
+ "mt_chrf": 0.4531472771818437,
880
  "cls_acc": 0.5666666666666667,
881
+ "mlm_chrf": 0.9630716853128435,
882
+ "asr_wer": 0.22800492332171055,
883
+ "asr_chrf": 0.8922038015648965,
884
+ "t2t_score": 0.6609618763871179,
885
+ "s2t_score": 0.5601043624433035
886
  },
887
  {
888
  "language_name": "Punjabi",
 
892
  {
893
  "model": "meta-llama/llama-3.3-70b-instruct",
894
  "model_type": "text-to-text",
895
+ "mt_bleu": 0.3260979917168239,
896
+ "mt_chrf": 0.48620285493698817,
897
+ "cls_acc": 0.4,
898
+ "mlm_chrf": 0.8966325892385384,
899
+ "t2t_score": 0.5942784813918421
900
  },
901
  {
902
  "model": "elevenlabs/scribe_v1",
903
  "model_type": "speech-to-text",
904
+ "asr_wer": 0.2622994225519276,
905
+ "asr_chrf": 0.7896064299629604,
906
+ "s2t_score": 0.525952926257444
907
  }
908
  ],
909
  "commonvoice_hours": 2.3,
 
917
  "SG": 9314
918
  },
919
  "language_family": "Indo-European",
920
+ "mt_bleu": 0.3260979917168239,
921
+ "mt_chrf": 0.48620285493698817,
922
+ "cls_acc": 0.4,
923
+ "mlm_chrf": 0.8966325892385384,
924
+ "asr_wer": 0.2622994225519276,
925
+ "asr_chrf": 0.7896064299629604,
926
+ "t2t_score": 0.5942784813918421,
927
+ "s2t_score": 0.525952926257444
928
  },
929
  {
930
  "language_name": "Russian",