The best performing model on the MedHallu benchmark achieved an F1 score as low as 0.625 for detecting 'hard' category hallucinations.

Authors

• Person: Not available Organization: arXiv
  [2502.14302] MedHallu: A Comprehensive Benchmark for Detecting ...
• Person: Not available Organization: The Moonlight
  [Literature Review] MedHallu: A Comprehensive Benchmark for ...

Sources

• [2502.14302] MedHallu: A Comprehensive Benchmark for Detecting ... arxiv.org arXiv via serper
• [Literature Review] MedHallu: A Comprehensive Benchmark for ... www.themoonlight.io The Moonlight via serper

Referenced by nodes (2)

• hallucination concept
• F1 score concept