Скачать или смотреть Neural Search Talks: Finetuning Embeddings for RAG

Welcome to Zeta Alpha's first RAG Meetup in San Francisco! In this session, Jakub Zavrel, Founder & CEO of Zeta Alpha, discusses the details of fine-tuning embedding models for Retrieval-Augmented Generation (RAG). You will find answers to why semantic search using neural embeddings outperforms traditional keyword search, and how combining BM25 with fine-tuned embeddings and cross-encoders can yield highly relevant search results. This talk also features an overview of the InPars methodology for generating domain-specific question-answer pairs to fine-tune models and a discussion of the complexities that arise when evaluating search results and LLM-generated outputs in a production environment.

Check out the second part of this meetup here:    • Neural Search Talks: Using LLMs for Evalua...  

Zeta Alpha is focused on transforming internal expertise into generative AI capabilities for enterprises. Learn more by visiting the platform and trying our public index for AI research at https://zeta-alpha.com