Local Explanations for Large Language Models: a Brief Review of Methods

Large Language Models (LLMs) have demonstrated outstanding performance in natural language processing over the past few years. The adoption of LLMs across digital applications in various walks of life is growing exponentially. However, the mechanisms by which LLMs, like all types of artificial neural networks, yield their results remain opaque. This lack of transparency in decision-making poses risks for the further use of the technology. This paper presents a concise review of approaches to obtaining explanations of LLM outcomes. It examines the principal methods of explainable artificial intelligence that are applied to generate explanations for LLM predictions. An approach for classifying local explanation methods is proposed.