From Tokens to Semantics: Leveraging Complementary Signals for Hallucination Detection in Black-Box LLMs
In the evaluation across models and datasets, Stacked gave the best performance in nearly half of the cases, while TopK and CoCoA remain competitive without supervised training labels, although their thresholds require careful calibration.
Urja Pawar, R. Ramanayake, O. O'Neill et al.
· arXiv.org · 0 citations