Using Mechanistic Interpretability to Craft Adversarial Attacks against Large Language Models

Documenti analoghi

• Crafting Imperceptible On-Manifold Adversarial Attacks for Tabular Data
  di: He, Zhipeng, et al.
  Pubblicazione: (2025)
• Binary Autoencoder for Mechanistic Interpretability of Large Language Models
  di: Cho, Hakaze, et al.
  Pubblicazione: (2025)
• Towards Interpretable Adversarial Examples via Sparse Adversarial Attack
  di: Lin, Fudong, et al.
  Pubblicazione: (2025)
• Crafting Adversarial Inputs for Large Vision-Language Models Using Black-Box Optimization
  di: Guan, Jiwei, et al.
  Pubblicazione: (2026)
• Mechanistic Interpretability with SAEs: Probing Religion, Violence, and Geography in Large Language Models
  di: Simbeck, Katharina, et al.
  Pubblicazione: (2025)