ACL Findings2026
XLSR-MamBo: Scaling the Hybrid Mamba-Attention Backbone for Audio Deepfake Detection
Kwok-Ho Ng, Tingting Song, Yongdong Wu, Zhihua Xia
Abstract
Advanced speech synthesis technologies have enabled highly realistic speech generation, posing security risks that motivate research into audio deepfake detection (ADD). While state space models (SSMs) offer linear complexity, pure causal SSMs architectures often struggle with the content-based retrieval required to capture global frequency-domain artifacts. To address this, we explore the scaling properties of hybrid architectures by proposing XLSR-MamBo, a modular framework integrating an XLSR front-end with synergistic Mamba-Attention backbones. We systematically evaluate four topological designs using advanced SSM variants, Mamba, Mamba2, Hydra, and Gated DeltaNet. Experimental results demonstrate that the MamBo-3-Hydra-N3 configuration achieves competitive performance compared to other state-of-the-art systems on the ASVspoof 2021 LA, DF, and In-the-Wild benchmarks. This performance benefits from Hydra's native bidirectional modeling, which captures holistic temporal dependencies more efficiently than the heuristic dual-branch strategies employed in prior works. Furthermore, evaluations on the DFADD dataset demonstrate robust generalization to unseen diffusion-and flow-matching-based synthesis methods. Crucially, our analysis reveals that scaling backbone depth effectively mitigates the performance variance and instability observed in shallower models. These results demonstrate the hybrid framework's ability to capture artifacts in spoofed speech signals, providing an effective method for ADD. Codes are publicly available at https://github.com/saki-ciallo/ XLSR-MamBo.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
- FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness cites Efficiently Modeling Long Sequences with Structured State Spaces
- Robust Speech Recognition via Large-Scale Weak Supervision cites wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations
- Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality cites Efficiently Modeling Long Sequences with Structured State Spaces
Builds on14
- wav2vec 2.0: A Framework for Self-Supervised Learning of Speech RepresentationsAlexei Baevski, Yuhao Zhou, Abdelrahman Mohamed, Michael AuliNeurIPS 2020 · 9,451 citations
- Robust Speech Recognition via Large-Scale Weak SupervisionAlec Radford, Jong Wook Kim, Tao Xu, Greg Brockman et al.ICML 2023 · 6,966 citations
- FlashAttention: Fast and Memory-Efficient Exact Attention with IO-AwarenessTri Dao, Daniel Y. Fu, Stefano Ermon, Atri Rudra et al.NeurIPS 2022 · 5,493 citations
- Efficiently Modeling Long Sequences with Structured State SpacesAlbert Gu, Karan Goel, Christopher RéICLR 2022 · 3,482 citations
- Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space DualityTri Dao, Albert GuICML 2024 · 1,407 citations
Related papers
- SONAR: Spectral‑Contrastive Audio Residuals for Generalizable Deepfake DetectionIdo Nitzan Hidekel, Gal Lifshitz, Khen Cohen, Dan RavivICML 2026 · 1 citation
- Audio Deepfake Detection with Self-Supervised XLS-R and SLS ClassifierQishan Zhang, Shuangbing Wen, Tao HuACM MM 2024 · 54 citations
- X-AVDT: Audio-Visual Cross-Attention for Robust Deepfake DetectionYoungseo Kim, Kwan Yun, Seokhyeon Hong, Sihun Cha et al.CVPR 2026 · 2 citations
- WMamba: Wavelet-based Mamba for Face Forgery DetectionSiran Peng, Tianshuo Zhang, Li Gao, Xiangyu Zhu et al.ACM MM 2025 · 17 citations
- Rethinking Fake Speech Detection: A Generalized Framework Leveraging Spectrogram MagnitudeZihao Liu, Aobo Chen, Yan Zhang, Wensheng Zhang et al.NDSS 2026
