# Reshaping Representation Space to Balance the Safety and Over-rejection in Large Audio Language Models

arXiv: https://arxiv.org/abs/2505.19670 · HTML: https://mlsanity.com/paper/2505.19670/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Builds on 18 of its 33 references that are in the map.
- First posted on arXiv 26 May 2025.

## Builds on (most cited first, 18 of 18)

- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!](https://mlsanity.com/paper/2310.03693/) (2023)
- [BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Dataset](https://mlsanity.com/paper/2307.04657/) (2023)
- [Qwen-Audio: Advancing Universal Audio Understanding via Unified Large-Scale Audio-Language Models](https://mlsanity.com/paper/2311.07919/) (2023)
- [AIR-Bench: Benchmarking Large Audio-Language Models via Generative Comprehension](https://mlsanity.com/paper/2402.07729/) (2024)
- [VoiceBench: Benchmarking LLM-Based Voice Assistants](https://mlsanity.com/paper/2410.17196/) (2024)
- [VITA-1.5: Towards GPT-4o Level Real-Time Vision and Speech Interaction](https://mlsanity.com/paper/2501.01957/) (2025)
- [Safety Layers in Aligned Large Language Models: The Key to LLM Security](https://mlsanity.com/paper/2408.17003/) (2024)
- [Keeping LLMs Aligned After Fine-tuning: The Crucial Role of Prompt Templates](https://mlsanity.com/paper/2402.18540/) (2024)
- [Lisa: Lazy Safety Alignment for Large Language Models against Harmful Fine-tuning Attack](https://mlsanity.com/paper/2405.18641/) (2024)
- [CoVoST 2 and Massively Multilingual Speech-to-Text Translation](https://mlsanity.com/paper/2007.10310/) (2020)
- [Antidote: Post-fine-tuning Safety Alignment for Large Language Models against Harmful Fine-tuning](https://mlsanity.com/paper/2408.09600/) (2024)
- [CoVoST: A Diverse Multilingual Speech-To-Text Translation Corpus](https://mlsanity.com/paper/2002.01320/) (2020)
- [Audio Is the Achilles' Heel: Red Teaming Audio Large Multimodal Models](https://mlsanity.com/paper/2410.23861/) (2024)
- [NLSR: Neuron-Level Safety Realignment of Large Language Models Against Harmful Fine-Tuning](https://mlsanity.com/paper/2412.12497/) (2024)
- [Panacea: Mitigating Harmful Fine-tuning for Large Language Models via Post-fine-tuning Perturbation](https://mlsanity.com/paper/2501.18100/) (2025)
- [MoGU: A Framework for Enhancing Safety of Open-Sourced LLMs While Preserving Their Usability](https://mlsanity.com/paper/2405.14488/) (2024)
- [Towards Probing Speech-Specific Risks in Large Multimodal Models: A Taxonomy, Benchmark, and Insights](https://mlsanity.com/paper/2406.17430/) (2024)

## Built on by (most cited first, 7 of 7)

- [SARSteer: Safeguarding Large Audio-Language Models via Safe-Ablated Refusal Steering](https://mlsanity.com/paper/2510.17633/) (2025)
- [Omni-Safety under Cross-Modality Conflict: Vulnerabilities, Dynamics Mechanisms and Efficient Alignment](https://mlsanity.com/paper/2602.10161/) (2026)
- [Resurfacing Paralinguistic Awareness in Large Audio Language Models](https://mlsanity.com/paper/2603.11947/) (2026)
- [Audio Jailbreaks in Large Audio-Language Models: Taxonomy, Attack-Defense Analysis, and Cost-Aware Evaluation](https://mlsanity.com/paper/2605.30031/) (2026)
- [AEGIS: Audio Endogenous Guarding via Internal Signals Against Large Audio-Language Model Jailbreaks](https://mlsanity.com/paper/2609.29287/) (2026)
- [Causal Localization of the Refusal Direction in Audio Language Models](https://mlsanity.com/paper/2609.22260/) (2026)
- [From Acoustic Sensing to Physical Actions: A Comprehensive Survey of Security Risks in Edge-Deployed Audio LLM-Based Agents](https://mlsanity.com/paper/2610.15390/) (2026)

## Abstract (arXiv)

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.