An imitation learning approach with two-stage on-policy reward learning enhances TTS for elderly listeners and outperforms standard GRPO and supervised baselines.
Preference alignment improves language model- based TTS,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SD 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Imitation Learning for Elder-Facing Speech Synthesis
An imitation learning approach with two-stage on-policy reward learning enhances TTS for elderly listeners and outperforms standard GRPO and supervised baselines.