GROW fine-tunes flow-matching TTS through signed, group-normalized advantage weighting of the standard regression loss, improving WER and speaker similarity without degrading naturalness.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
GROW: Group-Relative Advantage-Weighted On-Policy Reinforcement Learning of Autoregressive-Diffusion Text-to-Speech model
GROW fine-tunes flow-matching TTS through signed, group-normalized advantage weighting of the standard regression loss, improving WER and speaker similarity without degrading naturalness.