<ul><li>This paper introduces a new approach, RouteNator, for fine-tuning Large Language Models (LLMs) for function calling tasks using synthetic data generation.</li><li>RouteNator addresses the challenge of limited real user interaction data by leveraging domain resources, content metadata, knowledge graphs, and language models to create diverse and high-quality synthetic training data.</li><li>The flexible routing mechanism in RouteNator ensures that the synthetic data generated matches real-world distributions, leading to improved function classification accuracy and API parameter selection.</li><li>Evaluation shows that models fine-tuned with RouteNator's synthetic data outperform traditional approaches, setting new benchmarks for function calling tasks.</li></ul>

RouteNator: A Router-Based Multi-Modal Architecture for Generating Synthetic Training Data for Function Calling LLMs

Discover more