Mechanistic Steering of LLMs Reveals Layer-wise Feature Vulnerabilities in Adversarial Settings

Documenti analoghi

• Human-Readable Adversarial Prompts: An Investigation into LLM Vulnerabilities Using Situational Context
  di: Das, Nilanjana, et al.
  Pubblicazione: (2024)
• Human-Interpretable Adversarial Prompt Attack on Large Language Models with Situational Context
  di: Das, Nilanjana, et al.
  Pubblicazione: (2024)
• Mechanistic Knobs in LLMs: Retrieving and Steering High-Order Semantic Features via Sparse Autoencoders
  di: Zhang, Ruikang, et al.
  Pubblicazione: (2026)
• Mental Health Equity in LLMs: Leveraging Multi-Hop Question Answering to Detect Amplified and Silenced Perspectives
  di: Haider, Batool, et al.
  Pubblicazione: (2025)
• Experiments or Outcomes? Probing Scientific Feasibility in Large Language Models
  di: Mohammadi, Seyedali, et al.
  Pubblicazione: (2026)