The ICR Score (Information Contribution to Residual Stream) and the ICR Probe are metrics used for reference-free hallucination detection that aggregate layer-wise residual updates, outperforming prior hidden-state baselines with a lightweight MLP.

Authors

• Person: Not available Organization: GitHub
  EdinburghNLP/awesome-hallucination-detection - GitHub

Sources

• EdinburghNLP/awesome-hallucination-detection - GitHub github.com GitHub via serper

Referenced by nodes (1)

• hallucination detection concept