Metraipolis › archive › dc_3a877dc4629b
GROW: Group-Relative Advantage-Weighted On-Policy Reinforcement Learning of Autoregressive-Diffusion Text-to-Speech model — arXiv:2608.03215
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.03215
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-31T12:18:39.640Z
- SHA-256 of the archived text
fc05faf6d079a9e07bd5392972839d1f70505fc2553d99df6c8e213bc941ea4b- Stable URL of this page
https://metraipolis.com/documento/dc_3a877dc4629b
How to cite
GROW: Group-Relative Advantage-Weighted On-Policy Reinforcement Learning of Autoregressive-Diffusion Text-to-Speech model — arXiv:2608.03215 — arXiv (en), https://arxiv.org/abs/2608.03215 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_3a877dc4629b (Metraipolis archive dc_3a877dc4629b, sha256 fc05faf6d079a9e07bd5392972839d1f70505fc2553d99df6c8e213bc941ea4b)
BibTeX
@misc{dc3a877dc4629b,
title = {GROW: Group-Relative Advantage-Weighted On-Policy Reinforcement Learning of Autoregressive-Diffusion Text-to-Speech model — arXiv:2608.03215},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.03215},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_3a877dc4629b; sha256 fc05faf6d079a9e07bd5392972839d1f70505fc2553d99df6c8e213bc941ea4b},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.03215. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Reinforcement learning for flow-matching text-to-speech is complicated by deterministic ODE sampling: trajectory-level policy-gradient methods typically convert the ODE into an SDE and track per-step likelihood ratios, introducing stochastic perturbations and substantial overhead. We propose GROW, a group-relative advantage-weighted on-policy RL method that acts directly on the standard flow-matching objective. For each prompt, GROW samples a group of on-policy utterances, separately standardizes intelligibility and speaker-similarity rewards within the group, and combines them to reweight flow-matching regression. A Wasserstein-2 velocity penalty anchors the updated model to a frozen pretrained reference. A group-mean reward baseline is introduced to convert reward weighting into advantage weighting. For strong pretrained TTS models with concentrated rewards, positive exponential weighting is dominated by reward-agnostic self-imitation, whereas a zero-mean signed advantage preserves effective within-group credit assignment. Instantiated on DiTAR and evaluated on LibriSpeech and Seed-TTS EN/ZH, GROW reduces average WER from 2.016 to 1.558 and raises speaker similarity from 0.676 to 0.715 while keeping UTMOS. With 10-NFE training rollouts and 32-NFE evaluation, GROW retains comparable performance while training 2.9x faster than 32-NFE DiTAR-GRPO. We will open-source complete GROW codes, faithful DiTAR reproduction, and all model checkpoints. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-05 in cs.AI). Categorie arXiv: Audio and Speech Processing (eess.AS) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2608.03215 — 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 2026-10-31T12:05:53.014Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-05?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).