# Lost in Speech: Trilingual Spoken Hallucination Detection Across Audio and Transcripts

arXiv: https://arxiv.org/abs/2608.24707 · HTML: https://mlsanity.com/paper/2608.24707/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 25 of its 44 references that are in the map.
- First posted on arXiv 25 Aug 2026.

## Builds on (most cited first, 25 of 25)

- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Robust Speech Recognition via Large-Scale Weak Supervision](https://mlsanity.com/paper/2212.04356/) (2022)
- [A Survey on LLM-as-a-Judge](https://mlsanity.com/paper/2411.15594/) (2024)
- [FEVER: a large-scale dataset for Fact Extraction and VERification](https://mlsanity.com/paper/1803.05355/) (2018)
- [AIR-Bench: Benchmarking Large Audio-Language Models via Generative Comprehension](https://mlsanity.com/paper/2402.07729/) (2024)
- [Fine-grained Hallucination Detection and Editing for Language Models](https://mlsanity.com/paper/2401.06855/) (2024)
- [AVeriTeC: A Dataset for Real-world Claim Verification with Evidence from the Web](https://mlsanity.com/paper/2305.13117/) (2023)
- [X-FACT: A New Benchmark Dataset for Multilingual Fact Checking](https://mlsanity.com/paper/2106.09248/) (2021)
- [ASVspoof 5: Crowdsourced Speech Data, Deepfakes, and Adversarial Attacks at Scale](https://mlsanity.com/paper/2408.08739/) (2024)
- [Careless Whisper: Speech-to-Text Hallucination Harms](https://mlsanity.com/paper/2402.08021/) (2024)
- [Fighting Fire with Fire: The Dual Role of LLMs in Crafting and Detecting Elusive Disinformation](https://mlsanity.com/paper/2310.15515/) (2023)
- [Sparks of Large Audio Models: A Survey and Outlook](https://mlsanity.com/paper/2308.12792/) (2023)
- [Hallucinations in Neural Automatic Speech Recognition: Identifying Errors and Hallucinatory Models](https://mlsanity.com/paper/2401.01572/) (2024)
- [SemEval-2025 Task 3: Mu-SHROOM, the Multilingual Shared Task on Hallucinations and Related Observable Overgeneration Mistakes](https://mlsanity.com/paper/2504.11975/) (2025)
- [KnowHalu: Hallucination Detection via Multi-Form Knowledge Based Factual Checking](https://mlsanity.com/paper/2404.02935/) (2024)
- [Lost in Transcription, Found in Distribution Shift: Demystifying Hallucination in Speech Foundation Models](https://mlsanity.com/paper/2502.12414/) (2025)
- [A Crowdsourced Open-Source Kazakh Speech Corpus and Initial Speech Recognition Baseline](https://mlsanity.com/paper/2009.10334/) (2020)
- [HalluVerse25: Fine-grained Multilingual Benchmark Dataset for LLM Hallucinations](https://mlsanity.com/paper/2503.07833/) (2025)
- [KazakhTTS: An Open-Source Kazakh Text-to-Speech Synthesis Dataset](https://mlsanity.com/paper/2104.08459/) (2021)
- [Fakes of Varying Shades: How Warning Affects Human Perception and Engagement Regarding LLM Hallucinations](https://mlsanity.com/paper/2404.03745/) (2024)
- [KazakhTTS2: Extending the Open-Source Kazakh TTS Corpus With More Data, Speakers, and Topics](https://mlsanity.com/paper/2201.05771/) (2022)
- [KazMMLU: Evaluating Language Models on Kazakh, Russian, and Regional Knowledge of Kazakhstan](https://mlsanity.com/paper/2502.12829/) (2025)
- [How Much Do LLMs Hallucinate across Languages? On Realistic Multilingual Estimation of LLM Hallucination](https://mlsanity.com/paper/2502.12769/) (2025)
- [MAD: A Benchmark for Multi-Turn Audio Dialogue Fact-Checking](https://mlsanity.com/paper/2508.12186/) (2025)
- [Controlled Automatic Task-Specific Synthetic Data Generation for Hallucination Detection](https://mlsanity.com/paper/2410.12278/) (2024)

## Built on by (most cited first, 1 of 1)

- [TRILOGUE: A Trilingual Spoken Dialogue Fact-Checking Benchmark with Evidence and Paired Audio](https://mlsanity.com/paper/2609.04452/) (2026)

## Abstract (arXiv)

While text-based hallucination detection is well studied, reference-free detection of factual alterations in speech remains underexplored, especially for low-resource languages. Our spoken benchmark comprises 12,013 English, Russian, and Kazakh news samples with three synthetic alteration types and three severity levels, pairing source articles with rewrites as text, synthesized audio, and ASR transcripts. We add 290 fact-checked misinformation items collected in Russian (225) and Kazakh (65), translated into the other language and rendered through the same TTS-ASR pipeline. We evaluate fine-tuned multilingual encoders and zero-shot multimodal decoders on text, transcripts, and audio. Detectors receive only target inputs without source articles or external evidence; the task evaluates reference-free classification rather than evidence-grounded verification. Encoder degradation from source text to transcripts generally tracks per-language ASR error on the binary task. Among decoders, only Gemma-3n exceeds the binary majority-class baseline in macro-F1, on transcripts only; the other four fall below their respective baselines. Comparisons between audio and transcripts are confounded by differences in evaluation coverage and class balance. Synthetic-trained detectors achieve 0.82--0.88 macro-F1 on real-world misinformation source text; Russian provenance analysis reveals veracity-related and model-dependent machine-style signals, a key confound in synthetic hallucination benchmarks.