Hallucination
When a model confidently states something false.
We are working on a detailed page for Hallucination - covering why it matters, how it works, related terms, and the tools that use it.
Related terms
From the glossaryFrequently asked questions
Can hallucinations be eliminated?+
Not entirely with current architectures. They can be reduced with grounding techniques like RAG, output verification, constrained decoding, and human review, but a zero-hallucination guarantee does not exist yet.
Why do LLMs hallucinate?+
Models are trained to produce plausible next tokens, not to verify factual accuracy. When the correct answer is uncertain or absent from training data, the model generates a confident-sounding guess.
How can I detect hallucinations?+
Check outputs against authoritative sources, use a second model to critique the first, or implement retrieval so claims can be traced back to a cited document.