Reshaping Representation Space to Balance the Safety and Over-rejection in Large Audio Language Models
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866916759243587584 |
|---|---|
| author | Yang, Hao Qu, Lizhen Shareghi, Ehsan Haffari, Gholamreza |
| author_facet | Yang, Hao Qu, Lizhen Shareghi, Ehsan Haffari, Gholamreza |
| contents | Large Audio Language Models (LALMs) have extended the capabilities of Large Language Models (LLMs) by enabling audio-based human interactions. However, recent research has revealed that LALMs remain vulnerable to harmful queries due to insufficient safety-alignment. Despite advances in defence measures for text and vision LLMs, effective safety-alignment strategies and audio-safety dataset specifically targeting LALMs are notably absent. Meanwhile defence measures based on Supervised Fine-tuning (SFT) struggle to address safety improvement while avoiding over-rejection issues, significantly compromising helpfulness. In this work, we propose an unsupervised safety-fine-tuning strategy as remedy that reshapes model's representation space to enhance existing LALMs safety-alignment while balancing the risk of over-rejection. Our experiments, conducted across three generations of Qwen LALMs, demonstrate that our approach significantly improves LALMs safety under three modality input conditions (audio-text, text-only, and audio-only) while increasing over-rejection rate by only 0.88% on average. Warning: this paper contains harmful examples. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_19670 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Reshaping Representation Space to Balance the Safety and Over-rejection in Large Audio Language Models Yang, Hao Qu, Lizhen Shareghi, Ehsan Haffari, Gholamreza Computation and Language Multimedia Sound Audio and Speech Processing Large Audio Language Models (LALMs) have extended the capabilities of Large Language Models (LLMs) by enabling audio-based human interactions. However, recent research has revealed that LALMs remain vulnerable to harmful queries due to insufficient safety-alignment. Despite advances in defence measures for text and vision LLMs, effective safety-alignment strategies and audio-safety dataset specifically targeting LALMs are notably absent. Meanwhile defence measures based on Supervised Fine-tuning (SFT) struggle to address safety improvement while avoiding over-rejection issues, significantly compromising helpfulness. In this work, we propose an unsupervised safety-fine-tuning strategy as remedy that reshapes model's representation space to enhance existing LALMs safety-alignment while balancing the risk of over-rejection. Our experiments, conducted across three generations of Qwen LALMs, demonstrate that our approach significantly improves LALMs safety under three modality input conditions (audio-text, text-only, and audio-only) while increasing over-rejection rate by only 0.88% on average. Warning: this paper contains harmful examples. |
| title | Reshaping Representation Space to Balance the Safety and Over-rejection in Large Audio Language Models |
| topic | Computation and Language Multimedia Sound Audio and Speech Processing |
| url | https://arxiv.org/abs/2505.19670 |