Generating Predictions with Augmented Input in RAG
In the final stage of the Retrieval-Augmented Generation (RAG) process, the augmented input, denoted as , serves as the context. The Large Language Model then uses this combined information to predict the subsequent text, an operation that is mathematically represented by the conditional probability model .
0
1
Tags
Ch.3 Prompting - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Ch.2 Generative Models - Foundations of Large Language Models
Related
Knowledge Source Preparation in RAG
Text Retrieval in RAG
Generating Predictions with Augmented Input in RAG
A system is designed to answer questions by first finding relevant information in a private document library and then using that information to create a more accurate answer. Arrange the following actions into the correct operational sequence that this system would follow for each incoming question.
An automated question-answering system is designed to first search a large database of documents for relevant information and then use that information to construct a final answer. Users report that while the system's answers are well-written and factually accurate based on the documents, they often fail to address the specific question asked. For example, when asked 'What are the key features of the latest smartphone model?', the system provides a detailed history of the company that makes the phone. Which component of the system's process is the most likely point of failure?
Troubleshooting a Knowledge-Base Chatbot
Learn After
Prompting LLMs with Retrieved Texts in RAG
A system is designed to answer questions using a two-step process: first, it finds relevant documents from a database, and second, it uses a large language model to generate a final answer. A user asks, 'What is the battery life of the new "Innovate X" phone?' The system retrieves the following text: 'The Innovate X phone features a 5000mAh battery, providing up to 48 hours of talk time.' Which of the following inputs to the language model is structured to produce the most accurate and relevant final answer?
Diagnosing a Generation Failure in an Information System
Evaluating Generation Strategies in a Q&A System