When Synthetic Speech Is All You Have: Better Call GRPO

Researchers at an undisclosed institution applied Group Relative Policy Optimization (GRPO) to adapt an LLM-based automatic speech recognition (ASR) system to synthetic speech, achieving a 40% relative word error rate (WER) reduction over supervised fine-tuning (SFT) and a 45% reduction with combined SFT and GRPO. The method improves stopping calibration and speech-to-text alignment without altering early-layer representations, offering a privacy-preserving alternative for regulated domains like banking.

arXiv:2607.08409v1 Announce Type: new Abstract: LLM-based ASR adapted to regulated domains such as banking is bottlenecked by privacy: real speech is costly and legally constrained to collect, making synthetic text-to-speech TTS an attractive substitute. Yet synthetic speech stays acoustically mismatched with real recordings, and work on this gap has stayed within supervised fine-tuning SFT . We instead turn to reinforcement learning, and show that Group Relative Policy Optimization GRPO extracts far more from the same synthetic speech than SFT. Synthetic-only adaptation of the model with GRPO, a critic-free method rewarding low-WER hypotheses, reduces WER by 40\% relative to SFT 36.71\%$\to$22.09\% , and an SFT-then-GRPO combination pushes this further to 45\%. We trace the gain to behavior rather than representation: GRPO reduces insertion errors by improving stopping calibration and speech-to-text alignment by better anchoring attention to audio, leaving early-layer representations intact. When synthetic speech is the main resource, reinforcement learning should be preferred over supervised fine-tuning.