Erase the Injected Region: Multimodal Prompt-Injection Defense with On-Policy Self-Distillation

arXiv:2610.41307 · cs.CR, cs.AI · Submitted 2026-10-30 · Read on arXiv

cs.CR, cs.AI

Submitted: 2026-10-30

Updated: 2026-10-30

Terminology

• Distillation — 1× in this paper · explained in 2 episodes
• Self-Distillation — 1× in this paper · explained in 7 episodes

Related papers

• WADBERT: Dual-Channel Web Attack Detection via Permutation-Invariant Parameter Modeling
• Your Token Is Gone: A Token-Consumption Countermeasure for Disrupting LLM-Based Scam Systems
• Membership Inference Attack on Private Evolution
• Rank Ceiling for Twiddle-Perturbation Faults on the Forward NTT
• WebAgentGuard: A Reasoning-Driven Guard Model for Detecting Prompt Injection Attacks in Web Agents
• SoK: AI-Augmented Binary Reversing