Knowledge Cutoff
Domain: LLM Concepts
🎯 Motivation
An LLM’s training data has a cutoff date. Events after that date are unknown to the model unless provided in the context.
📋 Concrete Example
If a model’s training data ends in January 2025, it doesn’t know about events after that. Ask ‘Who won the 2025 World Series?’ and it may guess or say it doesn’t know.
🔗 Analogy
An encyclopedia printed in January 2025 - it has everything up to that date, but nothing after. If you want current information, you need to supplement it with a live news feed (RAG).