Category:
Category:
Hallucinations
Category:
Evaluation & Quality
Definition
Situations where the LLM produces incorrect or fabricated information.
Explanation
Hallucinations happen when an LLM guesses instead of recalling facts. Causes include missing grounding, non-deterministic reasoning, poor prompts, or flawed retrieval. Enterprises mitigate hallucinations through RAG, structured prompting, verification layers, self-reflection, or tool calling. Reducing hallucinations is essential for compliance-heavy sectors like healthcare, finance, and law.
Technical Architecture
LLM Output → Verification Layer → (Correct / Reject / Revise) → Final Output
Core Component
RAG, rerankers, validators, guardrails, fact-checking tools
Use Cases
Legal summaries, medical QA, analytics, enterprise copilots
Pitfalls
High risk in regulated industries; can undermine trust
LLM Keywords
Hallucinations, Hallucination Mitigation, Grounded AI
Related Concepts
Related Frameworks
• RAG
• Guardrails
• Self-Verification
• Hallucination Reduction Framework