Using Evaluations to Optimize a RAG Pipeline: from Chunkings and Embeddings to LLMs | by Christy Bergman | Jul, 2024

Retrieval Augmented Technology (RAG) is a helpful approach for utilizing your personal information in an AI-powered Chatbot. On this weblog submit, I’ll stroll via three key methods to get essentially the most out of RAG and consider every technique to search out one of the best combos.

For readers who simply need to know the TL;DR conclusion: essentially the most RAG accuracy enchancment got here from exploring totally different chunking methods.

• 89% Enchancment by altering the Chunking Technique 📦
• 20% Enchancment by altering the Embedding Mannequin 🤖
• 6% Enchancment by altering the LLM Mannequin 🧪

Let’s dive into every technique and discover the best-performers for a real-world RAG utility utilizing RAG element evaluations! 🚀📚

I’ll use Milvus documentation public net pages because the docs information and Ragas because the analysis methodology. See my earlier blog about how to use RAGAS. The remainder of this weblog is organized as follows:

1. Textual content Chunking Methods
2. Embedding Fashions
3. LLM (Generative) Fashions