Whisper-CD: Accurate Long-Form Speech Recognition using Multi-Negative Contrastive Decoding
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866911493137629184 |
|---|---|
| author | Ahn, Hoseong Chae, Jeongyun Park, Yoonji Shim, Kyuhong |
| author_facet | Ahn, Hoseong Chae, Jeongyun Park, Yoonji Shim, Kyuhong |
| contents | Long-form speech recognition with large encoder-decoder models such as Whisper often exhibit hallucinations, repetition loops, and content omissions. These errors can accumulate and be further amplified when the previous segment's transcription is used as decoding context. We propose Whisper-CD, a training-free contrastive decoding framework that contrasts clean-audio logits against negative logits computed from three acoustically motivated perturbations: Gaussian noise injection, silence signal, and audio temporal shift. We aggregate these negatives via the log-sum-exp operator, building a unified multi-negative objective for token-by-token decoding. Across five English long-form benchmarks, Whisper-CD reduces WER by up to 24.3pp on CORAAL and shows 48% faster token generation throughput than beam search. Because Whisper-CD operates purely at inference time, it can be applied as a drop-in replacement to already-deployed Whisper systems without retraining. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_06193 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Whisper-CD: Accurate Long-Form Speech Recognition using Multi-Negative Contrastive Decoding Ahn, Hoseong Chae, Jeongyun Park, Yoonji Shim, Kyuhong Sound Artificial Intelligence Audio and Speech Processing Long-form speech recognition with large encoder-decoder models such as Whisper often exhibit hallucinations, repetition loops, and content omissions. These errors can accumulate and be further amplified when the previous segment's transcription is used as decoding context. We propose Whisper-CD, a training-free contrastive decoding framework that contrasts clean-audio logits against negative logits computed from three acoustically motivated perturbations: Gaussian noise injection, silence signal, and audio temporal shift. We aggregate these negatives via the log-sum-exp operator, building a unified multi-negative objective for token-by-token decoding. Across five English long-form benchmarks, Whisper-CD reduces WER by up to 24.3pp on CORAAL and shows 48% faster token generation throughput than beam search. Because Whisper-CD operates purely at inference time, it can be applied as a drop-in replacement to already-deployed Whisper systems without retraining. |
| title | Whisper-CD: Accurate Long-Form Speech Recognition using Multi-Negative Contrastive Decoding |
| topic | Sound Artificial Intelligence Audio and Speech Processing |
| url | https://arxiv.org/abs/2603.06193 |