ConceptAI & ML
Retrieval-augmented generation
Retrieval-augmented generation is a technique that augments a language model by retrieving specific or domain knowledge documents to use as context for generating outputs.
Know first
Named as a prerequisite by the lectures that define this. The number is how many say so. A grey one has no page yet.
Language model3Large language models2LLMs2Context window1Custom instructions1Embedding1Embedding models1Encoders and decoders1
Opens up
Concepts that name this one as a prerequisite.
Lecturers
7Each in their own terms, from the easiest level.
Steven Feng
A technique that retrieves relevant information from an external data store to augment the context and guide an LLM's output.Andrej Karpathy
A technique where ChatGPT references chunks of text from uploaded files to use when creating responses.Jeremy Howard
Retrieval augmented generation uses search to find external documents that help a language model answer a specific question.Bryan Wilder
A system that enhances language model responses by retrieving relevant information from an external knowledge base.Cheng Chen
RAG is a cost-efficient approach to generative AI that combines document retrieval with text generation using in-context learning.Douwe Kiela
Coupling a language model to an external memory and retriever to provide retrieved context alongside the input and prompt.Ravi Vats
RAG powered LLM applications are a subset of LLM applications where you embed or store information about historical user interaction or other data to use as context for better serving users or solving problems.
Lectures that assume it
1Connections
18The lectures that define it or assume it, and the concepts next to it.
Concepts9
- Language model
- Large language models
- Context window
- Custom instructions
- FLARE
- Parametric vs semi-parametric models