LLM

A large language model (LLM) is a machine learning model trained on large collections of text and often other data to model language patterns. Inputs are divided into tokens, from which the model calculates probabilities for suitable continuations or other outputs. Many current LLMs use transformer architectures, although size alone determines neither quality nor suitability for a task.

LLMs can draft, summarise, translate, classify, extract information, and assist with code. They do not provide reliably verified factual knowledge like a curated database and may produce plausible but false statements. Depending on risk, outputs require source checking, testing, and human approval. Confidential data, copyright, bias, operating cost, and provider terms must be considered in deployment.

  • Models language through learned statistical patterns
  • Processes text as tokens
  • Many current models use transformers
  • Can generate plausible false statements
  • Review and approve outputs according to risk

Frequently asked questions

An LLM is a machine learning model that learns language patterns from extensive training data and uses them to process or generate text.

No. It can produce convincing errors, so important claims require verifiable sources and review proportionate to risk.
Zum Glossar