Back to directory
AgentPoison
Free217[NeurIPS 2024] Official implementation for "AgentPoison: Red-teaming LLM Agents via Memory or Knowledge Base Backdoor Po
About
This tool tests LLM agent vulnerabilities by injecting poisoned data into memory or knowledge bases. It's built for security researchers studying AI agent robustness. The key feature is demonstrating how backdoor attacks can compromise agent behavior through contaminated training data.
Key Features
- llm-agent
- red-team
- retrieval-augmented-generation
Pricing
FreeOpen source. You supply your own LLM API keys.
Categories
Research