Files
Abstract
The increasing adoption of language models in technical domains has raised curiosity about utilizing them to examine system design and specifications effectively. This research assesses how two popular LLMs, RoBERTa and GPT, can identify similarities among system technologies through their design and requirement details. Through a comparison of these model's performance, the study aims to gauge their capacity to categorize systems with traits and distinguish system attributes.
The study showed that GPT tends to make unique groups that can effectively capture subtle differences in system needs and details. This is in contrast to Roberta's tendency to combine system features into groups, which suggests that she has trouble accurately figuring out complex systems. Through methods like t-SNE projections and UMAP embedding, the research visually presents how each model excels in identifying patterns and distinguishing between system types.
The aim of this paper is to investigate how GPT and Roberta could evaluate systems in many different ways. GPT is adaptable enough to uncover variations between technological systems. The outcomes reveal how to choose models depending on the degree of complexity in the system design and requirement analysis.
The study showed that GPT tends to make unique groups that can effectively capture subtle differences in system needs and details. This is in contrast to Roberta's tendency to combine system features into groups, which suggests that she has trouble accurately figuring out complex systems. Through methods like t-SNE projections and UMAP embedding, the research visually presents how each model excels in identifying patterns and distinguishing between system types.
The aim of this paper is to investigate how GPT and Roberta could evaluate systems in many different ways. GPT is adaptable enough to uncover variations between technological systems. The outcomes reveal how to choose models depending on the degree of complexity in the system design and requirement analysis.