AI / LLM
Hallucination.
EN
When a model confidently makes up information β nonexistent libraries, fake APIs, imaginary function signatures.
TR
Modelin kendinden emin bir tonla uydurma bilgi ΓΌretmesi. Var olmayan kΓΌtΓΌphane, sahte API, hayali fonksiyon imzasΔ±.
Related terms.
03001
Slop
CultureLow-quality, generic, soulless content produced by AI. Often loaded with typical LLM tics: "It's important to note...", emoji overload, needless padding.
002
RAG
AI / LLMRetrieval-Augmented Generation. Fetching relevant docs into the context before the model answers. Reduces hallucinations.
003
Guardrails
SecuritySafety layer defining what the model can't do or say. Lives in the prompt, the system prompt, or pre/post filters.