Metraipolis › archive › dc_333195bbdef5

Training-Free Instruction TTS Gender Bias Calibration Using Model-Adaptive Steering — arXiv:2610.09831

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.09831
License (as archived)
CC0-1.0
Archived
2026-10-08T14:17:28.080Z
SHA-256 of the archived text
5b28ee4d439c87744b28971607db4826bd0b1511e4e96d48cfe6b47fdeb45fe9
Stable URL of this page
https://metraipolis.com/documento/dc_333195bbdef5

How to cite

Training-Free Instruction TTS Gender Bias Calibration Using Model-Adaptive Steering — arXiv:2610.09831 — arXiv (en), https://arxiv.org/abs/2610.09831 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_333195bbdef5 (Metraipolis archive dc_333195bbdef5, sha256 5b28ee4d439c87744b28971607db4826bd0b1511e4e96d48cfe6b47fdeb45fe9)
BibTeX
@misc{dc333195bbdef5,
  title = {Training-Free Instruction TTS Gender Bias Calibration Using Model-Adaptive Steering — arXiv:2610.09831},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.09831},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_333195bbdef5; sha256 5b28ee4d439c87744b28971607db4826bd0b1511e4e96d48cfe6b47fdeb45fe9},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.09831. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Instruction text-to-speech (ITTS) systems systematically encode acoustic gender skews from descriptive style prompts, such as occupations or personas, even when demographic attributes are left unspecified. Calibrating the implicit gender distribution of the synthesized voices, across heterogeneous architectures and without model retraining or overriding explicit user prompts, is an open problem. In this work, we propose \textit{model-adaptive steering}, a training-free bias calibration method that steers post-encoder conditioning representations using a group deviation vector paired with a coarse-to-fine strength search on a development set. A deterministic lexical gate bypasses intervention whenever explicit gender keywords are detected, preserving intended prompt semantics. Evaluated on a 12,800-prompt held-out benchmark across four ITTS models (three architectures), the proposed method reduces aggregate calibration error from 11.5--28.1 to 0.8--5.9 percentage points (e.g., shifting Parler-TTS Mini from 78.1\% and PromptTTS++ from 27.3\% female to 50.8--52.4\%). Output rates stay within 0.9 pp across anchor set sizes at fixed operating points, while UTMOS decreases by at most 0.08 and WER degrades by at most 3.3 pp.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv eess.AS, nuovi lavori del giorno).
Categorie arXiv: Audio and Speech Processing (eess.AS) ; Sound (cs.SD).
Pagina del lavoro: https://arxiv.org/abs/2610.09831 — 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 2026-10-08T13:29:00.766Z dalla pagina https://arxiv.org/list/eess.AS/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).