Knowing without Acting: The Disentangled Geometry of Safety Mechanisms in Large Language Models
Knowing without Acting: The Disentangled Geometry of Safety Mechanisms in Large Language Models explores Surgical attacks on LLM safety mechanisms enable novel jailbreaking and reveal architectural vulnerabilities, paving the way for robust security tools.. Commercial viability score: 8/10 in LLM Security.
BUILDER'S SANDBOX
Build This Paper
Use an AI coding agent to implement this research.
OpenAI CodexAI Agent
Lightweight coding agent in your terminal.
Claude CodeAI Agent
Agentic coding tool for terminal workflows.
AntiGravity IDEScaffolding
AI agent mindset installer and workflow scaffolder.
CursorIDE
AI-first code editor built on VS Code.
VS CodeIDE
Free, open-source editor by Microsoft.
Recommended Stack
Startup Essentials
Estimated $11K - $15K over 6-10 weeks.
See exactly what it costs to build this -- with 3 comparable funded startups.
7-day free trial. Cancel anytime.
Discover the researchers behind this paper and find similar experts.
7-day free trial. Cancel anytime.
Related Papers
- Builds On ThisEfficient Refusal Ablation in LLM through Optimal TransportViability: 6.0
- Builds On ThisHiding in Plain Text: Detecting Concealed Jailbreaks via Activation DisentanglementViability: 2.0
- Builds On ThisOOD-MMSafe: Advancing MLLM Safety from Harmful Intent to Hidden ConsequencesViability: 7.0
- Builds On ThisDeactivating Refusal Triggers: Understanding and Mitigating Overrefusal in Safety AlignmentViability: 7.0
- Builds On ThisBeyond Content Safety: Real-Time Monitoring for Reasoning Vulnerabilities in Large Language Models