mlsanityData to 13 Oct 2026

Rebellion: Noise-Robust Reasoning Training for Audio Reasoning Models

arXiv 2511.09682 ↗First posted 12 Nov 2025cs.AI cs.SDcited by 1

On the map: LLM reasoning → Audio-language model reasoning

also in 1 more
Abstract (arXiv)

Instilling reasoning capabilities in large models (LMs) using reasoning training (RT) significantly improves LMs' performances. Thus Audio Reasoning Models (ARMs), i.e., audio LMs that can reason, are becoming increasingly popular. However, no work has studied the safety of ARMs against jailbreak attacks that aim to elicit harmful responses from target models. To this end, first, we show that standard RT with appropriate safety reasoning data can protect ARMs from vanilla audio jailbreaks, but cannot protect them against our proposed simple yet effective jailbreaks. We show that this is because of the significant representation drift between vanilla and advanced jailbreaks which forces the target ARMs to emit harmful responses. Based on this observation, we propose Rebellion, a robust RT that trains ARMs to be robust to the worst-case representation drift. All our results are on Qwen2-Audio; they demonstrate that Rebellion: 1) can protect against advanced audio jailbreaks without compromising performance on benign tasks, and 2) significantly improves accuracy-safety trade-off over standard RT method.

Key facts. Cited by 1 paper in the map (data through 13 Oct 2026). Placed in Audio-language model reasoning, one of 17 directions in LLM reasoning. Builds on 29 of its 33 references that are in the map. First posted on arXiv 12 Nov 2025.

Builds on

29 of 33 references.

Closest ancestors · 5

Same direction, newest first

  1. 2025Audio Flamingo 3: Advancing Audio Intelligence with Fully Open Large Audio Language Models10 Jul
  2. 2025SoundMind: RL-Incentivized Logic Reasoning for Audio-Language Models15 Jun
  3. 2025SARI: Structured Audio Reasoning via Curriculum-Guided Reinforcement Learning22 Apr
  4. 2025Reinforcement Learning Outperforms Supervised Fine-Tuning: A Case Study on Audio Question Answering14 Mar
  5. 2025Audio-Reasoner: Improving Reasoning Capability in Large Audio Language Models4 Mar

Other references · 24

  1. 2022Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedbackcited by 3,691
  2. 2023Universal and Transferable Adversarial Attacks on Aligned Language Modelscited by 3,064
  3. 2020Sharpness-Aware Minimization for Efficiently Improving Generalizationcited by 1,411
  4. 2023Safe RLHF: Safe Reinforcement Learning from Human Feedbackcited by 648
  5. 2024Qwen2-Audio Technical Reportcited by 551
  6. 2024Deliberative Alignment: Reasoning Enables Safer Language Modelscited by 302
  7. 2025SafeChain: Safety of Language Models with Long Chain-of-Thought Reasoning Capabilitiescited by 135
  8. 2025STAR-1: Safer Alignment of Reasoning LLMs with 1K Datacited by 67
  9. 2025Omni-R1: Do You Really Need Audio to Fine-Tune Your Audio LLM?cited by 44
  10. 2024Voice Jailbreak Attacks Against GPT-4ocited by 40
  11. 2024Unveiling the Safety of GPT-4o: An Empirical Study using Jailbreak Attackscited by 38
  12. 2024AdvWave: Stealthy Adversarial Jailbreak Attack against Large Audio-Language Modelscited by 37
12 more
  1. 2024Audio Is the Achilles' Heel: Red Teaming Audio Large Multimodal Modelscited by 36
  2. 2025Multilingual and Multi-Accent Jailbreaking of Audio LLMscited by 28
  3. 2025Jailbreak-AudioBench: In-Depth Evaluation and Analysis of Jailbreak Threats for Large Audio Language Modelscited by 21
  4. 2025Audio Jailbreak: An Open Comprehensive Benchmark for Jailbreaking Large Audio-Language Modelscited by 20
  5. 2024SpeechGuard: Exploring the Adversarial Robustness of Multimodal Large Language Modelscited by 18
  6. 2025AudioJailbreak: Jailbreak Attacks against End-to-End Large Audio-Language Modelscited by 17
  7. 2025JALMBench: Benchmarking Jailbreak Vulnerabilities in Audio Language Modelscited by 14
  8. 2025Evaluating Robustness of Large Audio Language Models to Audio Injection: An Empirical Studycited by 11
  9. 2025"I am bad": Interpreting Stealthy, Universal and Robust Audio Jailbreaks in Audio-Language Modelscited by 8
  10. 2025SPIRIT: Patching Speech Language Models against Jailbreak Attackscited by 8
  11. 2025Audio Jailbreak Attacks: Exposing Vulnerabilities in SpeechGPT in a White-Box Frameworkcited by 4
  12. 2025When Good Sounds Go Adversarial: Jailbreaking Audio-Language Models with Benign Inputscited by 3

Built on by

Cited by year (2026 to 13 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Mega-ASR: Towards In-the-wild^2 Speech Recognition via Scaling up Real-world Acoustic Simulationcited by 4

Similar, not cited

Close in wording, no citation either way

  1. 2026Audio Jailbreaks in Large Audio-Language Models: Taxonomy, Attack-Defense Analysis, and Cost-Aware Evaluation
  2. 2026AEGIS: Audio Endogenous Guarding via Internal Signals Against Large Audio-Language Model Jailbreaks
  3. 2025Reasoning-to-Defend: Safety-Aware Reasoning Can Defend Large Language Models from Jailbreaking
  4. 2026GRM: Utility-Aware Jailbreak Attacks on Audio LLMs via Gradient-Ratio Masking
  5. 2025ALMGuard: Safety Shortcuts and Where to Find Them as Guardrails for Audio-Language Models
  6. 2025`From Prompt to Perturbation': An Adaptive Framework for Voice-Based Jailbreaks on Audio LLMs
  7. 2026On Optimizing Multimodal Jailbreaks for Spoken Language Models
  8. 2025Incentivizing Consistent, Effective and Scalable Reasoning Capability in Audio LLMs via Reasoning Process Rewards