aiqbench analysis
This research investigates the limitations of large audio-language models like Voxtral in detecting speech spoofing, revealing that standard LLM layers often prioritize semantic meaning over critical acoustic cues. The authors propose a lightweight adaptation technique using weight-decomposed low-rank adaptation (DoRA) to enhance the model's ability to distinguish between genuine and spoofed audio without heavy task-specific fine-tuning. Achieving an equal error rate of 4.25% on ASVspoof5, this work demonstrates a viable path toward integrating robust anti-spoofing capabilities directly into next-generation multimodal AI frameworks.
Spooftral: Can Voxtral Audio-Language Model Detect Speech Spoofing?
arXiv:2609.28713v1 Announce Type: cross Abstract: Self-supervised learning (SSL) countermeasures (CMs) have shown strong performance in recent years. However, they often show degraded performance while facing unseen spoofing attacks and mismatched conditions. This study examines the Voxtral audio-language model (ALM) framework for spoofing detection, as a step toward combining CM capabilities within the ALM framework. We analyze how Voxtral captures spoofing cues through audio-text processing and propose an instruction-guided approach that uses label-sequence likelihoods to evaluate bonafide and spoofed speech. Experiments on the ASVspoof databases show that without task-specific adaptation, the LLM layers emphasize semantic representations, reducing the separability of spoof-discriminative acoustic cues compared to the Whisper-based audio encoder. Consequently, spoofing-related information becomes less separable after language-model processing. We also applied lightweight adaptation using weight-decomposed low-rank adaptation (DoRA) to the Voxtral model and propose the Spooftral model, achieving an equal error rate (EER) of 4.25% on the ASVspoof5 evaluation set.
Title:Spooftral: Can Voxtral Audio-Language Model Detect Speech Spoofing?
Abstract:Self-supervised learning (SSL) countermeasures (CMs) have shown strong performance in recent years. However, they often show degraded performance while facing unseen spoofing attacks and mismatched conditions. This study examines the Voxtral audio-language model (ALM) framework for spoofing detection, as a step toward combining CM capabilities within the ALM framework. We analyze how Voxtral captures spoofing cues through audio-text processing and propose an instruction-guided approach that uses label-sequence likelihoods to evaluate bonafide and spoofed speech. Experiments on the ASVspoof databases show that without task-specific adaptation, the LLM layers emphasize semantic representations, reducing the separability of spoof-discriminative acoustic cues compared to the Whisper-based audio encoder. Consequently, spoofing-related information becomes less separable after language-model processing. We also applied lightweight adaptation using weight-decomposed low-rank adaptation (DoRA) to the Voxtral model and propose the Spooftral model, achieving an equal error rate (EER) of 4.25% on the ASVspoof5 evaluation set.
8 pages, 3 figures, 5 tables. Accepted to the Spoken Language Technology (SLT) 2026
Audio and Speech Processing (eess.AS); Computation and Language (cs.CL)
arXiv:2609.28713 [eess.AS]
(or arXiv:2609.28713v1 [eess.AS] for this version)
https://doi.org/10.48550/arXiv.2609.28713 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 266 words
Related coverage
Best Open Speech Recognition (ASR) Models in 2026: WER, Languages, Latency, and License Compared
MarkTechPost · Jul 23, 2026
Linguistically Augmented Audio Speech Data (LinguAS)
arXiv · Jun 10, 2026
Can Voice Agents Handle Bilingual Customers? Benchmarking Frontier ASR on Code-Switched Speech
Hugging Face · Jun 9, 2026
ArFake: A Robust Framework for Multi-Dialect Arabic Speech Spoofing Detection Benchmark
arXiv · Jun 17, 2026