From Text Metrics to Model Internals: A Study of Whisper ASR Hallucination Detection

• URL: http://arxiv.org/abs/2606.23060v1
• Date: Mon, 22 Jun 2026 09:13:04 GMT
• Title: From Text Metrics to Model Internals: A Study of Whisper ASR Hallucination Detection
• Abstract summary: Hallucinations of ASR models pose risks in downstream applications.<n>This paper studies Whisper large v3 hallucination detection on real-speech human-annotated data.
• Score: 13.55239135877789
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Hallucinations of ASR models - fluent transcriptions with no basis in audio - degrade system performance and pose risks in downstream applications. Robust detection of such errors remains a challenge. This paper studies Whisper large v3 hallucination detection on real-speech human-annotated data across three paradigms: text-based, LLM-based, and internal decoder state probing. Text classifiers utilizing metrics for text evaluation achieve high recall but degrade without reference transcripts. LLM-based detection improves precision with domain-specific prompt conditioning, yet remains less competitive than the lightweight text-based methods. Probing Whisper's decoder representations, without a ground-truth reference, yields the strongest performance, revealing that hallucination traits are encoded across intermediate decoding layers. A late-fusion meta-classifier combining text and internal-state outputs achieves the best overall detection performance.

Related papers

• Beyond Clean Text: Evaluating Encoder and Decoder Robustness for Bangla Event Detection in Noisy Text [0.0]
  Event detection systems are typically evaluated on clean, curated text, leaving their robustness to real-world noise largely unexplored.<n>We introduce a generalized Bangla news event and a benchmark comprising 9,979 annotated sentences across 40 event subtypes.<n>We show that encoder models achieve higher performance on clean text but degrade substantially under noise, whereas decoder-only LLMs are markedly more robust.
  arXiv  Detail & Related papers  (2026-06-29T21:03:32Z)
• Reward-Guided Semantic Evolution for Test-time Adaptive Object Detection [82.2968697030677]
  Open-vocabulary object detection with vision-language models (VLMs) such as Grounding DINO suffers from performance degradation under test-time distribution shifts.<n>We propose Reward-Guided Semantic Evolution (RGSE), a training-free framework that directly refines the text embeddings at test time.
  arXiv  Detail & Related papers  (2026-05-06T06:17:41Z)
• Toward Faithful Retrieval-Augmented Generation with Sparse Autoencoders [39.5490415037017]
  Retrieval-Augmented Generation (RAG) improves the factuality of large language models (LLMs) by grounding outputs in retrieved evidence.<n>Existing hallucination detection methods for RAG often rely on large-scale detector training.<n>We introduce RAGLens, a lightweight hallucination detector that accurately flags unfaithful RAG outputs.
  arXiv  Detail & Related papers  (2025-12-09T18:33:22Z)
• Diversity Boosts AI-Generated Text Detection [51.56484100374058]
  DivEye is a novel framework that captures how unpredictability fluctuates across a text using surprisal-based features.<n>Our method outperforms existing zero-shot detectors by up to 33.2% and achieves competitive performance with fine-tuned baselines.
  arXiv  Detail & Related papers  (2025-09-23T10:21:22Z)
• RepreGuard: Detecting LLM-Generated Text by Revealing Hidden Representation Patterns [50.401907401444404]
  Large language models (LLMs) are crucial for preventing misuse and building trustworthy AI systems.<n>We propose RepreGuard, an efficient statistics-based detection method.<n> Experimental results show that RepreGuard outperforms all baselines with average 94.92% AUROC on both in-distribution (ID) and OOD scenarios.
  arXiv  Detail & Related papers  (2025-08-18T17:59:15Z)
• Text-ADBench: Text Anomaly Detection Benchmark based on LLMs Embedding [27.02879006439693]
  This work performs a comprehensive empirical study and introduces a benchmark for text anomaly detection.<n>Our work systematically evaluates the effectiveness of embedding-based text anomaly detection.<n>By open-sourcing our benchmark toolkit, this work provides a foundation for future research in robust and scalable text anomaly detection systems.
  arXiv  Detail & Related papers  (2025-07-16T14:47:41Z)
• REFIND at SemEval-2025 Task 3: Retrieval-Augmented Factuality Hallucination Detection in Large Language Models [15.380441563675243]
  REFIND (Retrieval-augmented Factuality hallucINation Detection) is a novel framework that detects hallucinated spans within large language model (LLM) outputs.<n>We propose the Context Sensitivity Ratio (CSR), a novel metric that quantifies the sensitivity of LLM outputs to retrieved evidence.<n> REFIND demonstrated robustness across nine languages, including low-resource settings, and significantly outperformed baseline models.
  arXiv  Detail & Related papers  (2025-02-19T10:59:05Z)
• Who Wrote This? The Key to Zero-Shot LLM-Generated Text Detection Is GECScore [51.65730053591696]
  We propose a simple yet effective black-box zero-shot detection approach based on the observation that human-written texts typically contain more grammatical errors than LLM-generated texts.<n> Experimental results show that our method outperforms current state-of-the-art (SOTA) zero-shot and supervised methods.
  arXiv  Detail & Related papers  (2024-05-07T12:57:01Z)
• Enhancing Uncertainty-Based Hallucination Detection with Stronger Focus [99.33091772494751]
  Large Language Models (LLMs) have gained significant popularity for their impressive performance across diverse fields. LLMs are prone to hallucinate untruthful or nonsensical outputs that fail to meet user expectations. We propose a novel reference-free, uncertainty-based method for detecting hallucinations in LLMs.
  arXiv  Detail & Related papers  (2023-11-22T08:39:17Z)
• HyPoradise: An Open Baseline for Generative Speech Recognition with Large Language Models [81.56455625624041]
  We introduce the first open-source benchmark to utilize external large language models (LLMs) for ASR error correction. The proposed benchmark contains a novel dataset, HyPoradise (HP), encompassing more than 334,000 pairs of N-best hypotheses. LLMs with reasonable prompt and its generative capability can even correct those tokens that are missing in N-best list.
  arXiv  Detail & Related papers  (2023-09-27T14:44:10Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.