Glosarium AI
Kamus lengkap Kecerdasan Buatan
Knowledge Base
Structured collection of documents, articles, and external information indexed and vectorized to be retrieved by the RAG system during response generation.
Retrieval Component
Module of the RAG system responsible for identifying and extracting the most relevant passages from the knowledge base based on the user query.
Generation Component
Part of the RAG system that synthesizes a coherent response using the retrieved information as additional context for the language model.
Document Store
Persistent management system for source documents and their metadata, integrated with vector indexes to facilitate RAG retrieval.
Fine-tuning vs RAG
Distinction between adapting model weights (fine-tuning) and dynamic contextual enrichment (RAG) to improve performance without modifying the base model.
Reranking Models
Specialized models (such as Cross-Encoders) that reassess the relevance of retrieved documents by simultaneously considering the query and each candidate document.