Steering in the Shadows: Causal Amplification for Activation Space Attacks in Large Language Models

Ähnliche Einträge

• The dark deep side of DeepSeek: Fine-tuning attacks against the safety alignment of CoT-enabled models
  von: Xu, Zhiyuan, et al.
  Veröffentlicht: (2025)
• Steering Externalities: Benign Activation Steering Unintentionally Increases Jailbreak Risk for Large Language Models
  von: Xiong, Chen, et al.
  Veröffentlicht: (2026)
• Trojan Activation Attack: Red-Teaming Large Language Models using Activation Steering for Safety-Alignment
  von: Wang, Haoran, et al.
  Veröffentlicht: (2023)
• Blockchain Amplification Attack
  von: Tsuchiya, Taro, et al.
  Veröffentlicht: (2024)
• Toward Efficient Inference Attacks: Shadow Model Sharing via Mixture-of-Experts
  von: Bai, Li, et al.
  Veröffentlicht: (2025)