Topology-Conditioned Backdoors: Language Models That Insert Vulnerabilities When They Infer They Are in a Multi-Agent System
cs.AI, cs.MA
Submitted: 2026-10-05
Updated: 2026-10-05
Code: https://github.com/PyCQA/bandit
Terminology
Sources
- AgentPoison: Red-teaming LLM Agents via Poisoning Memory or Knowledge Bases
- Alignment faking in large language models
- BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain
- Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training
- Prompt Infection: LLM-to-LLM Prompt Injection within Multi-Agent Systems
- Auditing language models for hidden objectives
- Frontier Models are Capable of In-context Scheming
- Qwen2.5 Technical Report
- AI Sandbagging: Language Models can Strategically Underperform on Evaluations
- Backdooring Instruction-Tuned Large Language Models with Virtual Prompt Injection
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