Claude-3.5 and o1 exhibited the lowest hallucination rates across all tasks and risk categories, including achieving a 0% hallucination rate in the Diagnosis Prediction task.

Authors

• Person: Not available Organization: medRxiv
  Medical Hallucination in Foundation Models and Their Impact on ...

Sources

• Medical Hallucination in Foundation Models and Their Impact on ... www.medrxiv.org medRxiv via serper

Referenced by nodes (2)

• hallucination rate concept
• Claude concept