REDDIT: Correcting Model-Generated Timestamp Drift in ASR without Forgetting via Replay-Based Distribution Editing
By Cheng-Kang Chou · Paper · cs.CL
Modern autoregressive ASR systems can emit timestamps as decoded tokens, enabling timestamped transcription without frame-level aligners or inference-time post-processing. We show that these generated timestamps can drift across long non-speech spans: the transcript may remain pl