Whiteboard Benchmark Evaluates LLM Imagination versus Hallucination
September 22, 2026
The Whiteboard benchmark introduces a cross-taxonomy evaluation of seven imagination subtypes against ten hallucination boundaries. It aims to test whether creative content generation stems from the same mechanism as hallucination by measuring prompt-licensed departure from common answers.
HOW THIS AFFECTS YOU
●
researcherYou can use this to distinguish between controlled creative generation and uncontrolled factual errors in your models.