YZ Sözlüğü
Yapay Zekanın tam sözlüğü
RAG (Retrieval-Augmented Generation)
Hybrid architecture combining information retrieval and text generation to improve the accuracy and relevance of LLM responses by leveraging external knowledge.
Embedding
Dense vector representation of text or other data in a high-dimensional space, capturing semantic and contextual relationships between elements.
Multi-Modal RAG
RAG system capable of retrieving and generating from multiple data types (text, images, audio) for richer and more contextual responses.
Query Transformation
Technique for modifying and optimizing user queries before retrieval to improve the relevance of retrieved documents.
Contextual Compression
Technique for filtering and condensing retrieved context to retain only the most relevant information before generation.
Agentic RAG
RAG architecture where autonomous agents dynamically decide on retrieval and synthesis strategies based on context and query.
Self-Query Retriever
System that automatically decomposes a natural query into a structured query and metadata filters for more precise retrieval.
Multi-hop RAG
RAG architecture performing multiple sequential retrieval cycles to build complex responses requiring information from multiple sources.
RAG Pipeline
Orchestrated sequence of retrieval, filtering, contextualization, and generation steps in a RAG system to produce coherent responses.
Similarity Search
Search algorithm identifying the closest vectors in the embedding space according to metrics like cosine similarity or Euclidean distance.
Retrieval Augmentation
Process of enriching an LLM's prompt with relevant external information retrieved dynamically to improve factuality and accuracy.