Model Variability (MV) is a metric that measures the difference in hallucination rates across different models for a fixed prompt, where high MV indicates that hallucinations are primarily model-intrinsic.

Authors

• Person: Not available Organization: Frontiers
  Survey and analysis of hallucinations in large language models

Sources

• Survey and analysis of hallucinations in large language models www.frontiersin.org Frontiers via serper

Referenced by nodes (2)

• hallucination concept
• Model Variability concept