Questions
Common questions.
What is Data Poisoning?
Data poisoning is an attack where an adversary manipulates training or retrieval data to influence a model's outputs.
How is Data Poisoning used in production?
In RAG-based agents, poisoned knowledge base entries can cause the agent to return harmful or misleading information.
Which terms are related to Data Poisoning?
Closely related concepts include Hallucination Rate, Answer Faithfulness, Continuous Integration Agent, Retrieval Agent. Each is defined in the Prefactor glossary.
Keep reading