Jailbreak Context Lingers: Divergent Safety Routing and Its Cross-Task Predictability in Tool Agents
cs.AI
Submitted: 2026-09-28
Updated: 2026-09-28
Comments: 28 pages, 9 figures, 17 tabels, ICLR 2027 Under Review
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Sources
- Mind the GAP: Text Safety Does Not Transfer to Tool-Call Safety in LLM Agents
- The Struggle Between Continuation and Refusal: A Mechanistic Analysis of the Continuation-Triggered Jailbreak in LLMs
- The Llama 3 Herd of Models
- Feedback That Backfires: Why Small Language Model Agents Repeat the Call They Just Watched Fail
- How to use and interpret activation patching
- STAC: When Innocent Tools Form Dangerous Chains for LLM Agents
- Context-Fractured Decomposition Attacks on Tool-Using LLM Agents: Exploiting Artifact Provenance Gaps
- Agent MechSuits: Mechanistic Subspace Safety Steering for Multi-Turn CLI Agents
- History Anchors: How Prior Behavior Steers LLM Decisions Toward Unsafe Actions
- From Risk Classification to Action Plan Remediation: A Guardrail Feedback Driven Framework for LLM Agents
- Gemma 3 Technical Report
- Gemma 4 Technical Report
- Qwen3 Technical Report
- Don't Blindly Trust It: How Unreliable Feedback Breaks Tool-Using LLM Agents
- Universal and Transferable Adversarial Attacks on Aligned Language Models
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Agents' Last Exam
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems