1 paper
Shota Fujikawa, Issei Sato
Hallucination detection has become increasingly important for improving the reliability of large language models (LLMs). Recently, hybrid approaches such as HaMI, which combine sem…