What an LLM Actually Does

1. What is the primary function of an LLM when it generates a response?

Retrieving information from a database Generating the most plausible next token based on the input tokens Searching its knowledge base for the closest match Using predefined answers to respond to queries

2. Why do LLMs produce hallucinations or incorrect answers?

They do not have access to factual information The model's objective is plausibility, not truth They lack a proper training dataset The training corpus is too small

3. What does 'context window' refer to in the functioning of an LLM?

The entire dataset it was trained on The limit on how long the output can be The model's working memory containing the input tokens The source of training data for the model

4. How does temperature affect the output generated by an LLM?

It determines the response length It sets the accuracy of fact-based answers It influences the randomness in token selection It affects training efficiency

Continue: Embeddings and Vector Search → Review this lesson Retake quiz