FLEURS
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Fleurs
0.106WER
56
FLEURS en
2.72WER
49
Fleurs zh
0.83WER
34
FLEURS (test)
0.175Average Error Rate
30
Fleurs En
2.4WER
24
FLEURS zh-trad (test)
91Accuracy
21
FLEURS zh-sim (test)
93Normalized Edit Similarity Accuracy
21
FLEURS sr-cyrl (test)
0.94Normalized Edit Similarity
21
FLEURS sr-latn (test)
96Accuracy (Normalized Edit Similarity)
21
FLEURS S2TT (test)
89.66Hallucination Rate (%)
18
FLEURS Kazakh (test)
15.4WER (Oracle)
17
fleurs Tamil
28.68WER
17
Fleurs en (test)
0.336WER
17
FLEURS X→En (test)
37.4BLEU (Average)
16
FLEURS (test)
84.46Robustness Score (Gaussian Noise)
15
FLEURS (test)
86.3Avg. Score
15
FLEURS (test)
0.853COMET (Low)
15
FLEURS
2.47DE WER
13
FLEURS
2.71WER (en)
13
FLEURS
12.77AMH
13
Fleurs Retrieval-Eval (test)
57.89nDCG@5
13
FLEURS Vi-En wav2vec2-base ASR transcripts (test)
23.65BLEU
10
FLEURS Vi-En PhoWhisper-large ASR transcripts (test)
28.29BLEU
10
FLEURS Vi-En (test)
33.72BLEU
10
FLEURS 9×27 directions (test)
78.7ara
10