Bayesian hierarchical modeling (BHM) represents hallucination rates hierarchically with model-specific and prompt-specific parameters drawn from higher-level distributions, defined as: Hij = αi + βj + γij, where Hij is the hallucination rate for model i under prompt j, αi and βj represent model-specific and prompt-specific effects, and γij represents interaction effects (Gelman et al., 2013).

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 (1)

• hallucination rate concept