Large Language Models

Large Language Models (LLMs) are AI systems trained on vast text corpora to understand and generate human language. Examples include GPT, Claude, Llama, and Gemini.

How LLMs Work

LLMs predict the next token (word or sub-word) in a sequence based on context. By repeating this billions of times across diverse text, they develop surprisingly broad capabilities: answering questions, writing code, summarizing documents, and more.

Strengths

• Fluent, context-aware text generation
• Broad general knowledge from training data
• Adaptability to many tasks via prompting
• Multilingual capabilities in many models

Limitations

• Hallucinations: Generating false information confidently
• No real-time knowledge: Training data has a cutoff date
• No true understanding: Pattern matching, not comprehension
• Bias: Reflects biases in training text
• Inconsistency: Same prompt can yield different answers

Human Evaluation Focus

When testing LLMs, evaluate: factual accuracy, relevance to the prompt, safety, tone appropriateness, and whether the response actually helps the user. Rate specific failures with quoted evidence.