Bo Li
University of Illinois at Urbana-Champaign
最近一年的 AI 安全论文
本站共收录 169 篇- EvoSafeHarness：为 Agent 自动演化模型与领域专属安全护栏
EvoSafeHarness: Evolving Model- and Domain-Specific Harnesses for Securing Agents
- Twin Agent：面向权限分离智能体的上下文残差压缩
Twin Agent: Context Residual Compression for Privilege Separated Agents
- PolicyShiftGuard：政策自适应图像护栏的基准测试与改进
PolicyShiftGuard: Benchmarking and Improving Policy-Adaptive Image Guardrails
- FragFuse：通过基于记忆的查询碎片化与融合绕过大语言模型 Agent 的访问控制
FragFuse: Bypassing Access Control of Large Language Model Agents via Memory-Based Query Fragmentation and Fusion
- 谁付出代价？面向真实世界网页 Agent 的利益相关方中心提示注入基准测试
Who Pays the Price? Stakeholder-Centric Prompt Injection Benchmarking for Real-world Web Agents
- MAStrike：用 Shapley 值引导多智能体系统协同红队攻击
MAStrike: Shapley-Guided Collusive Red-Teaming on Multi-Agent Systems
- 晚了一轮：学习何时干预多轮对话中的恶意意图
One Turn Too Late: Learning When to Intervene Against Multi-Turn Malicious Intent
- DecodingTrust-Agent Platform（DTap）：面向 AI 智能体的可控交互式红队测试平台
DecodingTrust-Agent Platform (DTap): A Controllable and Interactive Red-Teaming Platform for AI Agents
- BabelSafe：以政策为依据的多语言大模型安全基准
BabelSafe: A Policy-Grounded Multilingual Safety Benchmark for LLMs
- Green Shielding：一种面向可信赖 AI 的以用户为中心的方法
Green Shielding: A User-Centric Approach Towards Trustworthy AI
- ShieldNet：针对智能体系统新兴供应链注入的网络级护栏
ShieldNet: Network-Level Guardrails against Emerging Supply-Chain Injections in Agentic Systems
- 多轮之下更不安全：使用工具的 Agent 多轮安全风险的基准测试与防御
Unsafer in Many Turns: Benchmarking and Defending Multi-Turn Safety Risks in Tool-Using Agents
- Just Ask：好奇的代码 Agent 揭示前沿 LLM 的系统提示词
Just Ask: Curious Code Agents Reveal System Prompts in Frontier LLMs
- GPT-5.2、Gemini 3 Pro、Qwen3-VL、Grok 4.1 Fast、Nano Banana Pro 与 Seedream 4.5 的安全报告
A Safety Report on GPT-5.2, Gemini 3 Pro, Qwen3-VL, Grok 4.1 Fast, Nano Banana Pro, and Seedream 4.5
- 通过指令遵循意图分析缓解间接提示注入
Mitigating Indirect Prompt Injection via Instruction-Following Intent Analysis
署名以 dblp 和本人的 Google Scholar 为准；只列本站收录的 AI 安全论文。