claim
Semantic similarity and token similarity detectors for hallucination detection show very low accuracy and recall but perform well with regards to precision, indicating they are primarily useful for identifying the most evident hallucinations.

Authors

Sources

Referenced by nodes (2)