deepset reposted this
🚀 Benchmarking RAG Pipelines for Optimal Performance! The latest article by David S. Batista, PhD, breaks down step-by-step instructions to evaluate and optimize your RAG pipeline's performance using Haystack, complete with detailed benchmarking against the ARAGOG dataset. 🌟 Key Highlights: - Explore three different Haystack pipelines: indexing, RAG, and evaluation. - Gain insights from the “ARAGOG: Advanced RAG Output Grading” dataset. - Learn how to fine-tune parameters like top_k, embedding_model, and chunk_size. - Discover the best-performing combinations for context relevance, faithfulness, and semantic answer similarity - Detailed analysis and lessons learned to help you achieve the best performance for your use case Read the full article 👉 https://lnkd.in/d-2NmawF #rag #opensource #evaluation #haystack
Sound and thorough work, but also pragmatic. Well done, David!
🤗 Sentence Transformers, SetFit & NLTK maintainer, MLE @ Hugging Face
3wThis is extremely useful, great job David S. Batista, PhD