Persona vectors: monitoring and controlling character traits in language models
"模型安全" 相关文章
Exploration hacking: can reasoning models subvert RL?
MCP Server工具参数设计与AI约束指南
Running in CIRCLE? A Simple Benchmark for LLM Code Interpreter Security
Jailbreak迎来“最后一卷”？港科大用“内容评分”重塑大模型越狱评估范式
Building and evaluating alignment auditing agents
Depth Gives a False Sense of Privacy: LLM Internal States Inversion
多模态大模型存在「内心预警」，无需训练，就能识别越狱攻击
Why it's hard to make settings for high-stakes control research
[奇思妙想] 关于公私钥技术在 AI 降智中的取证作用
Defense-as-a-Service: Black-box Shielding against Backdoored Graph Models
Efficiently Detecting Hidden Reasoning with a Small Predictor Model
Evaluating and monitoring for AI scheming
Beyond Weaponization: NLP Security for Medium and Lower-Resourced Languages in Their Own Right
AI应用安全挑战与测评实践指南
Early Signs of Steganographic Capabilities in Frontier LLMs
Layered AI Defenses Have Holes: Vulnerabilities and Key Recommendations
推理AI致命弱点，大模型变「杠精」，被带偏后死不悔改
SLT for AI Safety
Backdoor awareness and misaligned personas in reasoning models
