DiceHuBERT: Distilling HuBERT with a Self-Supervised Learning Objective
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866909675537039360 |
|---|---|
| author | Chi, Hyung Gun Aldeneh, Zakaria Likhomanenko, Tatiana Rudovic, Oggi Higuchi, Takuya Chen, Li-Wei Watanabe, Shinji Abdelaziz, Ahmed Hussen |
| author_facet | Chi, Hyung Gun Aldeneh, Zakaria Likhomanenko, Tatiana Rudovic, Oggi Higuchi, Takuya Chen, Li-Wei Watanabe, Shinji Abdelaziz, Ahmed Hussen |
| contents | We introduce DiceHuBERT, a knowledge distillation framework for compressing HuBERT, a widely used self-supervised learning (SSL)-based speech foundation model. Unlike existing distillation methods that rely on layer-wise and feature-wise mapping between teacher and student models, DiceHuBERT leverages HuBERT's iterative self-distillation mechanism by directly replacing the original model with a student model. This replacement allows the student to be trained using the same SSL objective used when pre-training HuBERT, eliminating the need for additional modules or architectural constraints. Experimental results on SUPERB show that DiceHuBERT consistently outperforms existing distillation methods, improving phoneme recognition performance by over 21% and ASR performance by more than 14%. Furthermore, DiceHuBERT demonstrates competitive performance across multiple tasks, highlighting its clear advantage. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_02911 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | DiceHuBERT: Distilling HuBERT with a Self-Supervised Learning Objective Chi, Hyung Gun Aldeneh, Zakaria Likhomanenko, Tatiana Rudovic, Oggi Higuchi, Takuya Chen, Li-Wei Watanabe, Shinji Abdelaziz, Ahmed Hussen Machine Learning Artificial Intelligence Sound Audio and Speech Processing We introduce DiceHuBERT, a knowledge distillation framework for compressing HuBERT, a widely used self-supervised learning (SSL)-based speech foundation model. Unlike existing distillation methods that rely on layer-wise and feature-wise mapping between teacher and student models, DiceHuBERT leverages HuBERT's iterative self-distillation mechanism by directly replacing the original model with a student model. This replacement allows the student to be trained using the same SSL objective used when pre-training HuBERT, eliminating the need for additional modules or architectural constraints. Experimental results on SUPERB show that DiceHuBERT consistently outperforms existing distillation methods, improving phoneme recognition performance by over 21% and ASR performance by more than 14%. Furthermore, DiceHuBERT demonstrates competitive performance across multiple tasks, highlighting its clear advantage. |
| title | DiceHuBERT: Distilling HuBERT with a Self-Supervised Learning Objective |
| topic | Machine Learning Artificial Intelligence Sound Audio and Speech Processing |
| url | https://arxiv.org/abs/2507.02911 |