Determining Generative Search Resault Document for queries using Generative Artificial Intelligens Models and Arbitration Models
Topics: LLM Readability, LLMO / GEO, Microsoft, Retrieval Augmented Generation (RAG), Scoring, Search Query Processing
This Microsoft patent describes a system called the “generative document system” that uses multiple generative AI models to create different types of search results documents in response to a user’s search query. Instead of just showing a list of links, the system generates several versions of AI-written answer documents at the same time and then uses a separate “arbitration model” to automatically pick the best one to show the user. The goal is to deliver more accurate, relevant, and well-formatted answers to search queries faster and with less effort from the user.
