Traditional n-gram metrics like ROUGE and BLEU are insufficient for capturing the clinical validity of generated text in medical LLMs.

Authors

• Person: Not available Organization: arXiv
  A Comprehensive Benchmark and Evaluation Framework for Multi ...

Sources

• A Comprehensive Benchmark and Evaluation Framework for Multi ... arxiv.org arXiv via serper

Referenced by nodes (2)

• ROUGE concept
• BLEU concept