Taking RAG Pipelines in Haystack to the Next Level With Document Ranking, Tuana Celik, Lead Developer Advocate, Deepset
Retrieval Augmented Generation is one of the most promising techniques to make use of LLMs, while avoiding hallucinations. In this 10 minute lightning talk, we will discuss new ranking techniques for the retrieval step: Diversity Ranking, Recentness Ranking and “Lost in the Middle” Ranking. Each technique helps us improve different use cases of RAG pipelines. For example, we will see how diversity ranking can help achieve better results for long-form question answering, while “Lost in the Middle” aims to minimize valuable context being overlooked.
Video Recorded at The AI Conference. Copyright, The AI Conference, All Rights Reserved
Add comment