Reference-based metrics like ROUGE show a clear misalignment with human judgments when identifying hallucinations in Question Answering tasks, as they consistently reward fluent yet factually incorrect responses.

Authors

• Person: Not available Organization: arXiv
  Re-evaluating Hallucination Detection in LLMs - arXiv

Sources

• Re-evaluating Hallucination Detection in LLMs - arXiv arxiv.org arXiv via serper

Referenced by nodes (3)

• hallucination detection concept
• ROUGE concept
• Question Answering concept