RAISED: Self-Distillation for Robustness to Prompt Injection in LLM Agents

arXiv:2610.06401 · cs.CR, cs.AI, cs.CL, cs.LG · Submitted 2026-10-05 · Read on arXiv

cs.CR, cs.AI, cs.CL, cs.LG

Submitted: 2026-10-05

Updated: 2026-10-07

Code: https://github.com/orailix/RAISED

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Distillation — 1× in this paper · explained in 2 episodes
• Robustness — 1× in this paper · explained in 5 episodes
• Self-distillation — 1× in this paper · explained in 6 episodes

Sources

• Aligning Language Models from User Interactions
• Evaluating Large Language Models Trained on Code
• StruQ: Defending Against Prompt Injection with Structured Queries
• Meta-SecAlign: Training LLMs against Prompt Injection for Robust Agents
• LlamaFirewall: An open source guardrail system for building secure AI agents
• Bias-Augmented Consistency Training Reduces Biased Reasoning in Chain-of-Thought
• AgentDojo: A Dynamic Environment to Evaluate Prompt Injection Attacks and Defenses for LLM Agents
• Defeating Prompt Injections by Design
• ToolTalk: Evaluating Tool-Usage in a Conversational Setting
• Reducing the Safety Tax in LLM Safety Alignment with On-Policy Self-Distillation
• Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection
• Bypassing LLM Guardrails: An Empirical Analysis of Evasion Attacks against Prompt Injection and Jailbreak Detection Systems
• Measuring Massive Multitask Language Understanding
• Defending Against Indirect Prompt Injection Attacks With Spotlighting
• Consistency Training Helps Stop Sycophancy and Jailbreaks
• Efficient Knowledge Injection in LLMs via Self-Distillation
• Prompt Infection: LLM-to-LLM Prompt Injection within Multi-Agent Systems
• AgentDyn: Are Your Agent Security Defenses Deployable in Real-World Dynamic Environments?
• Program Synthesis with Large Language Models
• Ignore Previous Prompt: Attack Techniques For Language Models

Related papers

• SoK: AI-Augmented Binary Reversing
• Relaxed Sender Anonymity for CBDC Interbank Settlement: A Zero-Knowledge Approach on Permissioned EVM
• Calibration-Family Overfit: Why Trusted Sabotage Monitors Don't Transfer Across Lineages
• Efficient Fuzzy PSI under One-Sided Assumptions
• Sealing the Audit-Runtime Gap for LLM Skills
• Token Composition: A Graph Based on EVM Logs