Towards Architectural-Level Safety Control: Relocating Hazardous Knowledge into Removable Experts in Mixture-of-Experts Models
Paper • 2610.25791 • Published • 4
👋 Open to Work
Xiaoyu Wen
XiaoyuWen
AI & ML interests
None yet
Recent Activity
upvoted a collection 20 days ago
REX authored a paper 24 days ago
JailbreakSkill: Scaling Automated Red-Teaming with Reusable and Ever-Evolving SkillsOrganizations
upvoted a paper 20 days ago
upvoted a collection 20 days ago
authored 7
papers 24 days ago
JailbreakSkill: Scaling Automated Red-Teaming with Reusable and Ever-Evolving Skills
Paper • 2608.16465 • Published • 1
AgentSafeEvo: Adversarial Co-Evolution for LLM Agent Safety
Paper • 2610.19422 • Published • 5
Towards Robust Offline-to-Online Reinforcement Learning via Uncertainty and Smoothness
Paper • 2309.16973 • Published
Disentangling Intent from Role: Adversarial Self-Play for Persona-Invariant Safety Alignment
Paper • 2605.01899 • Published
MAGIC: A Co-Evolving Attacker-Defender Adversarial Game for Robust LLM Safety
Paper • 2602.01539 • Published • 1
Not All Turns Matter: Credit Assignment for Multi-Turn Jailbreaking
Paper • 2605.08778 • Published
SafeWork-R1: Coevolving Safety and Intelligence under the AI-45$^{\circ}$ Law
Paper • 2507.18576 • Published • 10
upvoted a paper 24 days ago
updated a
collection 24 days ago
upvoted a paper 24 days ago
upvoted a paper 25 days ago
updated a
collection 2 months ago
updated 2
models 2 months ago