Intermediate · NLP
Groundedness
Visual diagram · (in preparation) · Math · (in preparation) · Worked example · 3 difficulty levels.
TL;DR. A model's output is grounded if it directly reflects information from its provided source material.
Technical Definition
A model's output is grounded if it directly reflects information from its provided source material.
How it works
Groundedness describes how well a model's responses are tied to specific input data, like a document or set of facts. For instance, if a language model answers a question based solely on information within a provided physics textbook, it is considered grounded. This doesn't guarantee factual accuracy, as the source material itself might contain errors.
Related Concepts
- Large Language Model (LLM) — A massive neural network trained on vast text corpora to understand and generate human language with remarkable fluency.
- Retrieval augmented generation (RAG) — LLMs that can access and use external information to improve their responses.