Consistent hallucinations across different models suggest prompt-induced errors, while divergent hallucination patterns imply architecture-specific behaviors or training artifacts.

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
• prompt concept