Unlocking Strong Supervision: A Data-Centric Study of General-Purpose Audio Pre-Training Methods
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866910076752625664 |
|---|---|
| author | Zhou, Xuanru Shao, Yiwen Tseng, Wei-Cheng Yu, Dong |
| author_facet | Zhou, Xuanru Shao, Yiwen Tseng, Wei-Cheng Yu, Dong |
| contents | Current audio pre-training seeks to learn unified representations for broad audio understanding tasks, but it remains fragmented and is fundamentally bottlenecked by its reliance on weak, noisy, and scale-limited labels. Drawing lessons from vision's foundational pre-training blueprint, we argue that the audio field must first establish its own large-scale, strong supervision framework. We introduce a new data-centric pipeline that leverages a high-fidelity captioner to create SOTA-quality captions and the first Unified Tag System (UTS) that bridges speech, music, and environmental sounds. We then conduct a systematic comparative study of different pre-training objectives on these strong source data. Our experiments suggest that data quality and coverage are the primary drivers of performance, while the choice of objective dictates downstream task specialization. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_25767 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Unlocking Strong Supervision: A Data-Centric Study of General-Purpose Audio Pre-Training Methods Zhou, Xuanru Shao, Yiwen Tseng, Wei-Cheng Yu, Dong Sound Artificial Intelligence Audio and Speech Processing Current audio pre-training seeks to learn unified representations for broad audio understanding tasks, but it remains fragmented and is fundamentally bottlenecked by its reliance on weak, noisy, and scale-limited labels. Drawing lessons from vision's foundational pre-training blueprint, we argue that the audio field must first establish its own large-scale, strong supervision framework. We introduce a new data-centric pipeline that leverages a high-fidelity captioner to create SOTA-quality captions and the first Unified Tag System (UTS) that bridges speech, music, and environmental sounds. We then conduct a systematic comparative study of different pre-training objectives on these strong source data. Our experiments suggest that data quality and coverage are the primary drivers of performance, while the choice of objective dictates downstream task specialization. |
| title | Unlocking Strong Supervision: A Data-Centric Study of General-Purpose Audio Pre-Training Methods |
| topic | Sound Artificial Intelligence Audio and Speech Processing |
| url | https://arxiv.org/abs/2603.25767 |