Theory of Hallucinations based on Equivariance

Hisaichi Shibata

This study aims to acquire knowledge for creating very large language models that are immune to hallucinations. Hallucinations in contemporary large language models are often attri…