Term
Hallucination
A hallucination is a language-model response that sounds plausible but is factually wrong or fabricated — typically caused by statistical guessing without a factual anchor.
Hallucination — explained in more detail
Language models generate text token by token based on probabilities. When the model lacks reliable context for a question, it still fills the gap fluently — the output looks convincing but may invent sources, misquote figures, or attribute statements to people who never said them. Hallucinations cluster in areas where training data is thin, the knowledge cutoff has passed, or the question asks for internal details the model simply does not have.
Example / Practical context
Classic cases: fabricated book titles with plausible-looking ISBNs, invented court rulings with realistic case numbers, made-up API parameters in code samples. Countermeasures include retrieval-augmented generation (RAG), requiring sources in the prompt, lowering temperature, and verifying through external tools or tests.
Distinction from related concepts
A hallucination is neither a deliberate lie nor a classical bug — it is a property of generative systems. It differs from prompt injection (external manipulation of the input) and from a knowledge-cutoff gap (where the model simply does not know something rather than inventing it).
Entdecke mehr
AI Workflows by Keyword: How We Make Recurring Routines Enforceable
A typed keyword triggers a fixed AI routine — and every single step must be committed before the next one appears. Why that's the actual trick.
GlossarReasoning Effort
Reasoning Effort is a control parameter on modern reasoning models that sets how much internal step-by-step thinking (thinking tokens) a model spends before answering — higher levels raise quality but also latency and cost.
LexikonLLM Hallucinations — Causes and Remedies
Why LLMs confidently invent falsehoods, what types of hallucinations exist, and which remedies actually help — RAG, source enforcement, verification.