arxiv:2606.07473

Whisper Hallucination Detection and Mitigation via Hidden Representation Steering and Sparse AutoEncoders

Published on Jun 5

· Submitted by

Georgii Aparin on Jun 9

HUAWEI Noah's Ark Lab

Upvote

Authors:

Georgii Aparin ,

Abstract

Research demonstrates that hallucinations in Whisper ASR can be detected and reduced using internal representations from audio encoder activations and Sparse AutoEncoder latents, achieving significant hallucination rate reduction with minimal speech transcription degradation.

Generated by Qwen/Qwen2.5-Coder-32B-Instruct

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.

View arXiv page View PDF Add to collection

Community

Egorgij21

Paper author Paper submitter about 19 hours ago

•

edited about 18 hours ago

Whisper Hallucination Detection and Mitigation via Hidden Representation Steering and Sparse AutoEncoders

Whisper has a well-known failure mode: feed it silence, noise, or music, and it will often respond with confidently fabricated transcripts. This paper shows you can detect and mitigate these hallucinations purely from the model's internal activations without fine-tuning.

We probe two representation spaces in Whisper's audio encoder: raw activations and Sparse AutoEncoder (SAE) latents. Both turn out to encode linearly separable hallucination signals, concentrated in a sparse subset of features that strengthen in deeper layers. Steering activations away from these directions at inference yields large drops in hallucination rate on non-speech samples from different datasets:

Whisper small: 72.63% → 14.11% hallucination rate on non-speech samples
Whisper large-v3: 86.88% → 27.33%

WER on regular speech data barely budges, and the method reaches numbers competitive with fine-tuning approaches like Calm-Whisper, without touching any model weights. A finding worth highlighting: since steering only a handful of encoder-side SAE features is enough to suppress hallucinations, the hallucination signal is not purely a decoder-side generation issue, it is already encoded in Whisper's encoder representations of non-speech audio.

Paper: https://arxiv.org/abs/2606.07473