대규모 언어 모델 (LLM)
거대 텍스트 기반 AI.
## The Hallucination Problem
LLMs do not have a 'database' of facts; they have a mathematical model of likelihood. When an LLM 'hallucinates,' it isn't lying—it is simply following a path of high statistical probability that happens to be factually wrong. Preventing hallucinations currently requires RAG (Retrieval-Augmented Generation), where the model is forced to look up external, trusted documents before answering.
graph LR
Center["대규모 언어 모델 (LLM)"]:::main
Pre_transformer["transformer"]:::pre --> Center
click Pre_transformer "/terms/transformer"
Pre_deep_learning["deep-learning"]:::pre --> Center
click Pre_deep_learning "/terms/deep-learning"
Pre_natural_language_processing["natural-language-processing"]:::pre --> Center
click Pre_natural_language_processing "/terms/natural-language-processing"
Center --> Child_context_window["context-window"]:::child
click Child_context_window "/terms/context-window"
Center --> Child_hallucination_ai["hallucination-ai"]:::child
click Child_hallucination_ai "/terms/hallucination-ai"
Rel_prompt_engineering["prompt-engineering"]:::related -.-> Center
click Rel_prompt_engineering "/terms/prompt-engineering"
Rel_model_distillation["model-distillation"]:::related -.-> Center
click Rel_model_distillation "/terms/model-distillation"
Rel_multimodal_ai["multimodal-ai"]:::related -.-> Center
click Rel_multimodal_ai "/terms/multimodal-ai"
classDef main fill:#7c3aed,stroke:#8b5cf6,stroke-width:2px,color:white,font-weight:bold,rx:5,ry:5;
classDef pre fill:#0f172a,stroke:#3b82f6,color:#94a3b8,rx:5,ry:5;
classDef child fill:#0f172a,stroke:#10b981,color:#94a3b8,rx:5,ry:5;
classDef related fill:#0f172a,stroke:#8b5cf6,stroke-dasharray: 5 5,color:#94a3b8,rx:5,ry:5;
linkStyle default stroke:#4b5563,stroke-width:2px;
🧒 5살도 이해할 수 있게 설명
📚 A powerful computer brain that can read and write like a human by predicting the next best word in a sentence.
🤓 Expert Deep Dive
## The Hallucination Problem
LLMs do not have a 'database' of facts; they have a mathematical model of likelihood. When an LLM 'hallucinates,' it isn't lying—it is simply following a path of high statistical probability that happens to be factually wrong. Preventing hallucinations currently requires RAG (Retrieval-Augmented Generation), where the model is forced to look up external, trusted documents before answering.