Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
Jailbreaks and attacks · Papers
Found 8 papers提出 ARENA 音频红队框架，自动构造文本安全且音频接地的越狱输入
- Published
- Evaluated models
- Audio Flamingo 3, Qwen2-Audio-7B-Instruct, MiMo-Audio-7B-Instruct and others4
提出 CFD 跨会话分解攻击，利用工件来源缺口绕过工具型 Agent 的本地检查
- Published
- Evaluated models
- GPT-4.1, GPT-5-nano, Claude-3.5-Sonnet and others6
AbstractCFD 用无指令 artifact 把越狱拆到跨会话。
CFD针对工具型 LLM 智能体：利用 artifact 来源不被跟踪这一部署失败模式，把有害目标拆开后延迟组合。
提出 CodecAttack，在编解码器隐空间构造可穿透有损压缩的对抗音频
- Published
- Evaluated models
- Qwen2-Audio-7B-Instruct, Audio Flamingo 3, Qwen2.5-Omni
提出声学干扰攻击 AIA，用良性音频潜语义越狱音频语言模型
- Published
- Evaluated models
- Qwen3-Omni, Qwen2.5-Omni, Qwen2-Audio and others10
提出 SWhisper，用近超声波向语音驱动 LLM 隐蔽注入越狱提示
- Published
- Evaluated models
- Meta-Llama-3.1-8B-Instruct, Gemma-3-4B, Qwen2.5-7B-Instruct and others9
提出音频叙事越狱，以嗓音投递风格绕过端到端语音模型拒答
- Published
- Evaluated models
- GPT-4o Realtime, Gemini 2.0 Flash, Qwen2.5-Omni-7B
AttacksarXiv 2511.10692· 提出 StyleBreak 风格化音频越狱，利用语音属性绕过音频语言模型对齐
- Published
- Evaluated models
- Qwen2-Audio-7B-Instruct, MERaLiON-AudioLLM-Whisper-SEA-LION, Ultravox-v0.4.1-Llama-3.1-8B and others6
AttacksarXiv 2504.01094· 提出 MULTI-AUDIOJAIL，以多语多口音和声学扰动越狱音频模型
- Published
- Evaluated models
- Qwen2-Audio, DiVA-llama-3-v0-8b, MERaLiON-AudioLLM-Whisper-SEA-LION and others5
Abstract黑盒多语多口音音频加声学扰动可抬高 LALM 的 JSR，文本示范防御并非在每个模型上都下降。
You have reached the end