Metraipolis › archive › dc_370b3cd7136e
SPEAR-Gen: Generation-Aware Pre-training for Unified Speech Representations — arXiv:2609.34147
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.34147
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T01:37:17.754Z
- SHA-256 of the archived text
d533a22657e86f0df094fca48b6f4e7c1792503e4d399337f24e679acdef06ff- Stable URL of this page
https://metraipolis.com/documento/dc_370b3cd7136e
How to cite
SPEAR-Gen: Generation-Aware Pre-training for Unified Speech Representations — arXiv:2609.34147 — arXiv (en), https://arxiv.org/abs/2609.34147 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_370b3cd7136e (Metraipolis archive dc_370b3cd7136e, sha256 d533a22657e86f0df094fca48b6f4e7c1792503e4d399337f24e679acdef06ff)
BibTeX
@misc{dc370b3cd7136e,
title = {SPEAR-Gen: Generation-Aware Pre-training for Unified Speech Representations — arXiv:2609.34147},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.34147},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_370b3cd7136e; sha256 d533a22657e86f0df094fca48b6f4e7c1792503e4d399337f24e679acdef06ff},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34147. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Speech understanding and generation place different demands on speech representations, and existing models are typically optimised towards one capability or the other. To reduce this gap, we introduce SPEAR-Gen, a speech representation model that learns a single representation for both capabilities. Task-aligned feature aggregation consolidates complementary linguistic and paralinguistic information across a frozen encoder into discrete targets for masked prediction, while a coarse-to-fine objective combines log-Mel reconstruction with residual flow matching to preserve spectral structure and fine-grained acoustic variation. Experiments on SUPERB and speech resynthesis show that SPEAR-Gen maintains strong understanding performance while substantially improving resynthesis quality and speaker preservation. These results demonstrate that a single speech representation can effectively support both understanding and generation. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in eess.AS). Categorie arXiv: Audio and Speech Processing (eess.AS) ; Sound (cs.SD). Pagina del lavoro: https://arxiv.org/abs/2609.34147 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-08T01:11:22.065Z dalla pagina https://arxiv.org/catchup/eess.AS/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).