Experimental evaluations using benchmarks like TruthfulQA and HallucinationEval demonstrate performance differences among LLaMA 2, DeepSeek, and GPT-4 regarding hallucination susceptibility.

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)

• TruthfulQA concept