English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

REDDIT: Correcting Model-Generated Timestamp Drift in ASR without Forgetting

Forum topic · 小凯 · 2026-07-08

Summary

Modern autogressive ASR systems can emit timestamps as decoded tokens, enabling timestamped transcription without frame-level aligners or post-processing. This paper identifies a key failure: generated timestamps drift over long non-speech spans, producing plausible text whose decoded timeline diverges from the audio. The authors propose REDDIT (REplay-based Distributional EDITing), a lightweight two-stage post-training framework that corrects timestamps while avoiding catastrophic forgetting. It first edits timestamp targets conditioned on the model's own replayed decoder context while matching the frozen base distribution for non-timestamp tokens, then applies a short edit-prefix refinement stage. On Whisper-tiny, using only 34.9 hours of correction audio and updating just 1.6% of model parameters, REDDIT raises long-interval mIoU from 38.7% to 95.0%, reduces mixed-interval out-of-domain AAS from 2752ms to 223ms, and preserves CV-en MER at 41.3%, versus 524.2% for naive SFT decoder tuning. arXiv: 2607.05364.

Paper Overview

Research Area: NLP

Authors: Cheng-Kang Chou, Ming-To Chuang, Ke-Han Lu, Chan-Jan Hsu, Hung-yi Lee

Published: 2026-07-06

arXiv: 2607.05364

Summary

Modern autoregressive ASR systems can emit timestamps directly as decoded tokens, enabling timestamped transcription without frame-level aligners or inference-time post-processing. However, this paper finds that such generated timestamps drift over long non-speech spans: the transcribed text may still look reasonable, but the decoded timeline diverges from the actual audio.

The authors propose REDDIT (REplay-based Distributional EDITing), a lightweight two-stage post-training framework that corrects timestamps while avoiding catastrophic forgetting:

1. Stage 1: Edit the timestamp targets conditioned on the model's own replayed decoder context, while matching the frozen base distribution for non-timestamp tokens. 2. Stage 2: Apply a short edit-prefix refinement phase.

Results

Experiments on Whisper-tiny, using only 34.9 hours of target correction audio and updating just 1.6% of model parameters, show:

  • Long-interval mIoU improved from 38.7% to 95.0%
  • Mixed-interval out-of-domain AAS reduced from 2752ms to 223ms
  • CV-en MER preserved at 41.3%, compared to 524.2% for naive SFT decoder tuning
---

*Auto-collected on 2026-07-06*

Tags

#asr#timestamps#whisper#post-training#catastrophic-forgetting#nlp#speech-recognition#arxiv

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178346211