Entity · paper

From Self-Supervised Speech Models to Mixture-of-Experts for Robust Anti-Spoofing

paperactivefrom-self-supervised-speech-models-to-mixture-of-experts-for-robust-anti-spoofing-817d69b8·1 events·first seen Jun 15, 2026

Aliases: From Self-Supervised Speech Models to Mixture-of-Experts for Robust Anti-Spoofing

Co-occurring entities

Mixture of Experts

More like this (12)

Sparse Mixture-of-Experts Speaker Group Encoding in Self-supervised Speech Recognition Models RAT: Reference-Augmented Training for ASV Anti-Spoofing Tying the Loop -- Tied Expert Layers in Mixture-of-Experts Language Models Leveraging Audio-LLMs to Filter Speech-to-Speech Training Data Cross-Modal Masking for Robust Silent Speech Synthesis Using sEMG and Lipreading Toward Calibrated Mixture-of-Experts Under Distribution Shift Evaluation of Adversarial Robustness in Arabic Language Models Mixture of Efficient Experts Beyond Fully Random Masking: Attention-Guided Denoising and Optimization for Diffusion Language Models Spend Experts Where You Are Unsure: Confidence-Adaptive Routing for Mixture-of-Experts LoRA Mixture of Experts

Recent events (1)

3arXiv · cs.AI·Jun 15, 2026·source ↗

MoE architecture improves self-supervised speech model robustness for anti-spoofing

Researchers propose converting a self-supervised speech representation model into a Mixture-of-Experts (MoE) architecture to improve generalization in synthetic speech detection. Feed-forward blocks in selected encoder layers are replaced by expert networks with a layer-wise gating mechanism, allowing complementary acoustic pattern capture while preserving pretrained representations. Evaluated across 14 spoofing datasets, the approach reduces macro Equal Error Rate from 5.46% to 4.81%, an 11.9% relative improvement over the baseline.

Evaluation and Benchmarking Mixture of Experts From Self-Supervised Speech Models to Mixture-of-Experts for Robust Anti-Spoofing