通过内部表示控制，显著降低Whisper语音识别的幻觉问题。
Whisper Hallucination Detection and Mitigation via Hidden Representation Steering and Sparse AutoEncoders
- 利用编码器激活和稀疏自编码器特征检测幻觉信号。
- Whisper小模型幻觉率从72.63%降至14.11%，大模型从86.88%降至27.33%。
- 无需微调即可实现接近微调效果，适合实时部署场景。
Whisper是一种广泛使用的语音识别模型，但其在非语音音频上会产生与输入无关的连贯转录（即幻觉）。本文研究是否可通过Whisper的内部表示检测并缓解幻觉。我们提取音频编码器激活，并评估两种表示空间：原始Whisper激活和稀疏自编码器（SAE）潜在表示。结果表明，两者均编码线性可分的幻觉相关特征，且判别能力集中在稀疏特征子集，并随编码层加深而增强。我们提出两种调控策略：激活空间调控与SAE潜在空间调控。基于SAE的调控在全非语音测试集上将Whisper small的幻觉率从72.63%降至14.11%，Whisper large-v3从86.88%降至27.33%，同时对语音数据的词错误率（WER）影响极小，性能接近微调方法。
原文摘要 · Abstract (English)
Whisper, a widely adopted ASR model, is known to suffer from hallucinations - coherent transcriptions generated for non-speech audio entirely disconnected from the input. We investigate whether hallucinations can be detected and mitigated through Whisper's internal representations. We extract audio encoder activations and evaluate two representation spaces: raw Whisper activations and Sparse AutoEncoder (SAE) latents. We show that both spaces encode linearly separable hallucination-related information, with discriminative power concentrated in a sparse feature subset and increasing toward deeper encoder layers. We propose two steering strategies: activation-space steering and SAE latent-space steering. SAE-based steering reduces hallucination rate from 72.63% to 14.11% for Whisper small and from 86.88% to 27.33% for Whisper large-v3 on the full non-speech test set, with small WER degradation on speech data, approaching the performance of fine-tuning-based methods.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。