1 paper
Adrian Theuma, Ehsan Shareghi
Large language models (LLMs) have exhibited an array of reasoning capabilities but face challenges like error propagation and hallucination, particularly in specialised areas like…