Optimizing the Performance of Generative Artificial Intelligence ,Recent Approaches to Engineering Large Language Models

In the ever-growing field of Generative Artificial Intelligence, quick engineering has become a revolutionary method in the field of NLP for big language versions. This technique entails the manipulation of input queries to improve the quality and relation of the textual outputs of these models. New findings point to the fact that through the enhancement of the aspect of prompt engineering it could indeed be a way of significantly boosting LLM performance through a change of the input query format. This paper highlights the practical use of the prompt engineering concept to Tamil-based LLMs with an explicit purpose of developing a well-structured process capable of generating accurate and contextually relevant conversational responses. Thus, with subtle changes to how prompts are offered, this research aims to enhance the efficacy of Tamil LLMs that demand trivial data entry. The research uses the Query Transformation Module (QTM), which is an elaborate method developed to systematically change the input prompts into three forms of inquiries. Therefore, each format is built with objectives and key points in mind to enhance the understanding of the models as well as the results. To measure the effectiveness of this strategy, the QTM was employed with leading Tamil LLMs SKT GPT-2 and Tamil ChatGPT. We employed four different query methods in our experiments: all the initially employed unaltered request and three modified requests according to the QTM. The evaluation was made using Google SSA to fit the evaluation criteria of naturalness and specificity for the sentences generated by the models. Our experimental findings show that overall, there is a reasonable improvement in the performance by an average of 11 percent. Identified that there is an increase of 46% in the quality of generated sentences when using the transformed queries as compared to the unmodified prompts. This improvement also depicts how the QTM has an effectiveness of enhancing the performance of Tamil LLMs, resulting into its usability as a tool in the corresponding field of prompt engineering.