Metraipolis › archive › dc_9ec0792fd087

Transcript-Supervised Post-Training of Generative Speech Enhancement on Real Recordings via Reinforce Adjoint Matching — arXiv:2609.29405

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.29405
License (as archived)
CC0-1.0
Archived
2026-10-08T05:26:33.235Z
SHA-256 of the archived text
c030eb4864b11af53d60bd5ea105ad09d91ca026616a4cf1a9fc6bcfa4ff2153
Stable URL of this page
https://metraipolis.com/documento/dc_9ec0792fd087

How to cite

Transcript-Supervised Post-Training of Generative Speech Enhancement on Real Recordings via Reinforce Adjoint Matching — arXiv:2609.29405 — arXiv (en), https://arxiv.org/abs/2609.29405 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_9ec0792fd087 (Metraipolis archive dc_9ec0792fd087, sha256 c030eb4864b11af53d60bd5ea105ad09d91ca026616a4cf1a9fc6bcfa4ff2153)
BibTeX
@misc{dc9ec0792fd087,
  title = {Transcript-Supervised Post-Training of Generative Speech Enhancement on Real Recordings via Reinforce Adjoint Matching — arXiv:2609.29405},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.29405},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_9ec0792fd087; sha256 c030eb4864b11af53d60bd5ea105ad09d91ca026616a4cf1a9fc6bcfa4ff2153},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.29405. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

We adapt Reinforce Adjoint Matching (RAM), a reward-based post-training method, to generative speech enhancement (SE). Starting from a pretrained SE model, RAM tilts the model's conditional distribution toward outputs with higher reward. During training, the current model generates enhanced speech on-policy, evaluates each generated endpoint with a potentially non-differentiable reward, and analytically re-noises the endpoint to construct inputs for a reward-guided regression objective. This enables post-training directly on real recordings using weak supervision, such as text transcripts, without requiring paired clean speech targets or reward gradients. We investigate word error rate (WER)-based post-training and whether recognition performance can be improved without compromising perceptual speech quality. Experiments on real CHiME-4 recordings reduce WER by 5.08 percentage points relative to pretrained FlowSE without reducing any of the reported non-intrusive speech quality metrics. A subjective listening test at the default reward scale finds no statistically significant preference between the post-trained and pretrained models.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-25 in cs.LG).
Categorie arXiv: Audio and Speech Processing (eess.AS) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.29405 — 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 2026-10-08T04:50:39.669Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-25?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).