# Wav2DF-TSL: Two-stage Learning with Efficient Pre-training and Hierarchical Experts Fusion for Robust Audio Deepfake Detection

arXiv: https://arxiv.org/abs/2509.04161 · HTML: https://mlsanity.com/paper/2509.04161/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Multimodal MoE models, one of 12 directions in Mixture-of-experts.
- First posted on arXiv 4 Sep 2025.

## Placed in

- Mixture-of-experts / [Multimodal MoE models](https://mlsanity.com/moe/multimodal-moe-models/)

## Built on by (most cited first, 4 of 4)

- [Addressing Gradient Misalignment in Data-Augmented Training for Robust Speech Deepfake Detection](https://mlsanity.com/paper/2509.20682/) (2025)
- [Alethia: A Foundational Encoder for Voice Deepfakes](https://mlsanity.com/paper/2605.00251/) (2026)
- [From Self-Supervised Speech Models to Mixture-of-Experts for Robust Anti-Spoofing](https://mlsanity.com/paper/2606.14639/) (2026)
- [Domain-Adaptive Dual-Gating Mixture of Experts for Generalizable Speech Deepfake Detection](https://mlsanity.com/paper/2609.33709/) (2026)

## Abstract (arXiv)

In recent years, self-supervised learning (SSL) models have made significant progress in audio deepfake detection (ADD) tasks. However, existing SSL models mainly rely on large-scale real speech for pre-training and lack the learning of spoofed samples, which leads to susceptibility to domain bias during the fine-tuning process of the ADD task. To this end, we propose a two-stage learning strategy (Wav2DF-TSL) based on pre-training and hierarchical expert fusion for robust audio deepfake detection. In the pre-training stage, we use adapters to efficiently learn artifacts from 3000 hours of unlabelled spoofed speech, improving the adaptability of front-end features while mitigating catastrophic forgetting. In the fine-tuning stage, we propose the hierarchical adaptive mixture of experts (HA-MoE) method to dynamically fuse multi-level spoofing cues through multi-expert collaboration with gated routing. Experimental results show that the proposed method significantly outperforms the baseline system on all four benchmark datasets, especially on the cross-domain In-the-wild dataset, achieving a 27.5% relative improvement in equal error rate (EER), outperforming the existing state-of-the-art systems. Index Terms: audio deepfake detection, self-supervised learning, parameter-efficient fine-tuning, mixture of experts