Improving RAG/CAG Based Additional Context Retrieval from Datasets Implementations via Pokémon-themed AI Chatbot

Abstract

Retrieval-Augmented Generation (RAG) is a commonly used, cost-effective solution for supplementing domain-focused knowledge for Large Language Models (LLMs), but contemporary RAG implementations often suffer from inconsistent accuracy and performance due to retrieval quality and context integration. In this study, a Pokémon dataset is used as a benchmark to evaluate the performance and factual accuracy of responses across a variety of model types, with the aim of identifying the most effective solution for information retrieval from a structured dataset.