Research library
Curate the field by topic, not just recency.
This section is the long-lived knowledge layer for important papers, organizing the multi-agent security literature into durable subareas that can support reading lists, annotated references, and future topic pages.
Stored article buckets
These groups come from the categorized article database and show a preview of the latest papers in each bucket.
2280 papers
Governance and Policy
- The Pushback Paradox: A Two-Probe Diagnostic for Language Model Compliance
- HERA: Harness-Environment Co-Evolution for Reliable Agentic Abstention
- ANT: A Multi-Granularity Network Traffic Dataset and Benchmark for Agents Behavior Auditing
1492 papers
Benchmarks and Evaluation
- BazaarBench: Delegation Safety in Decentralized C2C Marketplaces Run by LLM Agents
- The Pushback Paradox: A Two-Probe Diagnostic for Language Model Compliance
- HERA: Harness-Environment Co-Evolution for Reliable Agentic Abstention
1300 papers
Orchestration Risk
- BazaarBench: Delegation Safety in Decentralized C2C Marketplaces Run by LLM Agents
- HERA: Harness-Environment Co-Evolution for Reliable Agentic Abstention
- AECP: Artifact-Exclusive Communication Protocol for Multi-Agent Code Generation
1244 papers
Trust and Identity
- BazaarBench: Delegation Safety in Decentralized C2C Marketplaces Run by LLM Agents
- AgentPrivArena: Evaluating and Auditing Real-world AI Agent Privacy
- Let the Agent Do It? How Software Practitioners Understand and Make Permission Decisions in Agentic AI Assistants
1162 papers
Agent-to-Agent Communication
- AECP: Artifact-Exclusive Communication Protocol for Multi-Agent Code Generation
- Do Small Language Models Learn to Negotiate? A Controlled Scaling Study of RL-Trained Sellers
- Can CaMeLs Talk? Securing Multi-Agent Systems Against Indirect Prompt Injection Attacks
688 papers
Prompt Injection
- AECP: Artifact-Exclusive Communication Protocol for Multi-Agent Code Generation
- RAISED: Self-Distillation for Robustness to Prompt Injection in LLM Agents
- Compromise Is Not Consequence: Evaluating Task-Scoped Authorization in LLM Agents with Paired Replay
358 papers
Memory Poisoning
- StegoMemory: Agentic Memory Acts as Covert Steganographic Channel
- MemLeak: Cross-User Semantic Leakage in Multi-Tenant AI Agent Memory
- AgentGuardBench: A Multilingual Benchmark for Privacy, Security and Responsible Behaviour in AI Agents