Zero-resource speech modeling on sBLIMP in-vocabulary
56.8sBLIMP IV ScoreLAST
Evaluation results
| Method | Links | |
|---|---|---|
| LASTTWIST (Fine-tuning)=true, Adaptation layers=false, Model=OPT-350M, Codebook size=500, Backbone=HuBERT-base2024.09 | 56.8 | |
| k-meansTWIST (Fine-tuning)=true, Model=OPT-350M, Codebook size=500, Backbone=HuBERT-base2024.09 | 55.73 | |
| LASTTWIST (Fine-tuning)=false, Adaptation layers=true, Model=OPT-350M, Codebook size=500, Backbone=HuBERT-base2024.09 | 55.25 | |
| LASTTWIST (Fine-tuning)=false, Adaptation layers=false, Model=OPT-350M, Codebook size=500, Backbone=HuBERT-base2024.09 | 54.76 | |
| k-meansTWIST (Fine-tuning)=false, Model=OPT-350M, Codebook size=500, Backbone=HuBERT-base2024.09 | 53.3 |