Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2604.26327 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917449948987392 |
|---|---|
| author | Shangguan, Qituan Du, Junhao Peng, Kunyang Xue, Feng Zhang, Hui Wang, Xinsheng Yu, Kai Wang, Shuai |
| author_facet | Shangguan, Qituan Du, Junhao Peng, Kunyang Xue, Feng Zhang, Hui Wang, Xinsheng Yu, Kai Wang, Shuai |
| contents | Cross-lingual speaker verification suffers from severe language-speaker entanglement. This causes systematic degradation in the hardest scenario: correctly accepting utterances from the same speaker across different languages while rejecting those from different speakers sharing the same language. Standard adversarial disentanglement degrades speaker discriminability; blind discriminators inadvertently penalize speaker-discriminative traits that merely correlate with language. To address this, we propose Dual-LoRA, injecting trainable task-factorized LoRA adapters into a frozen pre-trained backbone. Our core innovation is a Language-Anchored Adversary: by grounding the discriminator with an explicit language branch, adversarial gradients target true linguistic cues rather than arbitrary correlations, preserving essential speaker characteristics. Evaluated on the TidyVoice benchmark, our system achieves a 0.91% validation EER and achieves 3rd place in the official challenge. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_26327 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Dual-LoRA: Parameter-Efficient Adversarial Disentanglement for Cross-Lingual Speaker Verification Shangguan, Qituan Du, Junhao Peng, Kunyang Xue, Feng Zhang, Hui Wang, Xinsheng Yu, Kai Wang, Shuai Audio and Speech Processing Cross-lingual speaker verification suffers from severe language-speaker entanglement. This causes systematic degradation in the hardest scenario: correctly accepting utterances from the same speaker across different languages while rejecting those from different speakers sharing the same language. Standard adversarial disentanglement degrades speaker discriminability; blind discriminators inadvertently penalize speaker-discriminative traits that merely correlate with language. To address this, we propose Dual-LoRA, injecting trainable task-factorized LoRA adapters into a frozen pre-trained backbone. Our core innovation is a Language-Anchored Adversary: by grounding the discriminator with an explicit language branch, adversarial gradients target true linguistic cues rather than arbitrary correlations, preserving essential speaker characteristics. Evaluated on the TidyVoice benchmark, our system achieves a 0.91% validation EER and achieves 3rd place in the official challenge. |
| title | Dual-LoRA: Parameter-Efficient Adversarial Disentanglement for Cross-Lingual Speaker Verification |
| topic | Audio and Speech Processing |
| url | https://arxiv.org/abs/2604.26327 |