Verif.ai - A Trustworthy Scientific Generative Search Engine
Explore a 27-minute conference talk from DSC Adria 2024 where Milos Kosprdic introduces an innovative open-source scientific question-answering system that delivers referenced and verifiable answers. Discover how this pioneering system utilizes a dual-stage process, combining robust information retrieval techniques with fine-tuned generative models and advanced verification engines to ensure answer accuracy and reliability. Learn about the integration of semantic and lexical search techniques over scientific papers and the implementation of a Retrieval-Augmented Generation (RAG) module that generates claims with references to source materials. Understand how the Verification Engine cross-checks generated claims against original articles to identify potential inaccuracies, ultimately enhancing scientists' productivity while building trust in generative language models within scientific domains. Presented at DSC Adria 2024 in Zagreb on May 23rd, this talk demonstrates a revolutionary approach to scientific question-answering and knowledge dissemination. Read more