Effective Rank (eRank) is used as a proxy for the diversity of final-layer hidden representations, where a collapse to fewer dimensions (low eRank) may indicate that a model is ignoring crucial input signals or relying on less context, potentially manifesting as hallucinations.

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 (1)

• hallucination concept