Create a RAG Hallucination Policy
Select a Moderation Behavior
After selecting Hallucination - RAG Haallucination, first determine which moderation behavior your policy should follow. Hallucination policies support: Flag, Block.
Select Policy Usage
Hallucination policies can only be applied to model outputs.
Provide Policy Information
Provide a policy name and description, these will only be used for identification in the future. In this step, you can also set a threshold score for each RAG hallucination metric: retrieval relevance, response faithfulness, and response relevance. Model responses with a score greater than this threshold will be considered violations.
Create your Policy
After the policy is created, it will appear in the policy dashboard.