arXiv Computation and Language
NOPE-HYPE: A Structured Simulation Workflow for Robust Speech-to-Text Across Diverse Acoustic Environments
Read the original on arXiv Computation and Language →NOPE-HYPE is a structured training workflow that integrates a controllable environment simulator, a coverage‑optimal reduction of Power Spectral Density templates, and a concise hyperparameter search over simulator settings. The authors demonstrate that noise generated by the simulator can match the performance of balanced real‑noise training for Whisper and SeamlessM4T models. They also provide principled environment prototype sets and practical default simulator configurations derived from a 27‑run hyperparameter sweep.
Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Computation and Language.