Fact-checking is an essential aspect of the verification process for Retrieval Augmented Generation (RAG) systems. FaaF: Facts as a Function for the evaluation of RAG systems introduces a fresh perspective to this process with the Facts as a Function (FaaF) approach. By utilizing the function calling capabilities of LMs, FaaF enables more reliable and cost-efficient identification of unsupported facts. Key insights include:

• Improving fact verification in RAG systems
• Utilizing function calls in LMs for better efficiency
• Significantly reducing costs compared to prompt-based methods

The introduction of FaaF propounds a new and effective method for ensuring the accuracy and factual integrity of RAG systems, pivotal for maintaining trust in AI-generated content.