Speaker Identification on LibriSpeech MF
98.6ScoreBEATs (300M) Speech
Evaluation results
| Method | Links | |
|---|---|---|
| BEATs (300M) SpeechParameter Count=300M, Pre-training Mixture=Speech-heavy (70:15:15)2026.01 | 98.6 | |
| Challenge BaselineDescription=Best among Dasheng-base, data2vec, and Whisper2026.01 | 98.5 | |
| Ensemble (Submission)Ensemble Strategy=Embedding Concatenation2026.01 | 98.5 | |
| UniWhisperEvaluation Protocol=MLP2026.02 | 98.1 | |
| WavLMEvaluation Protocol=MLP2026.02 | 97.9 | |
| HuBERTEvaluation Protocol=MLP2026.02 | 97.8 | |
| Dasheng 1.2BParameter Count=1.2B2026.01 | 97.5 | |
| BEATsEvaluation Protocol=MLP2026.02 | 97.5 | |
| BEATs (300M) BalancedParameter Count=300M, Pre-training Mixture=Balanced (40:30:30)2026.01 | 97.3 | |
| Whisper-UEvaluation Protocol=MLP2026.02 | 97.2 | |
| CLAP-UEvaluation Protocol=MLP2026.02 | 97.1 | |
| BEATs (90M) iter3Parameter Count=90M, Iteration=32026.01 | 97 | |
| CLAP-UEvaluation Protocol=kNN2026.02 | 96.4 | |
| wav2vec 2.0Evaluation Protocol=MLP2026.02 | 95.3 | |
| BEATsEvaluation Protocol=kNN2026.02 | 92.3 | |
| UniWhisperEvaluation Protocol=kNN2026.02 | 91.3 | |
| WhisperEvaluation Protocol=MLP2026.02 | 90.8 | |
| CLAPEvaluation Protocol=MLP2026.02 | 90.2 | |
| CLAPEvaluation Protocol=kNN2026.02 | 86.8 | |
| Whisper-U-3Evaluation Protocol=MLP2026.02 | 82.7 | |
| HuBERTEvaluation Protocol=kNN2026.02 | 79.9 | |
| WavLMEvaluation Protocol=kNN2026.02 | 77.2 | |
| wav2vec 2.0Evaluation Protocol=kNN2026.02 | 70.3 | |
| WhisperEvaluation Protocol=kNN2026.02 | 58 | |
| Whisper-U-3Evaluation Protocol=kNN2026.02 | 55.9 | |
| Whisper-UEvaluation Protocol=kNN2026.02 | 53.9 |