Abstract
Our research explores diversity in Large Language Model-generated product reviews. Leveraging Generative Artificial Intelligence, we examine both lexical and semantic diversity. Using Self-BLEU scores and cosine similarities, we investigate the impact of prompts and temperature on output diversity. Preliminary findings suggest higher diversity with multiple-output prompting. Future work includes deepening our investigation of multiple-output prompting, employing few-shot prompting, and analyzing time complexity.
Recommended Citation
Thomsen, Lars and Peters, Ralf, "Exploring Diversity in Large Language Model Outputs: A Study on Product Reviews" (2024). AMCIS 2024 TREOs. 7.
https://aisel.aisnet.org/treos_amcis2024/7
COinS