Insights & Innovation

Blogs

LLM Model Comparison for Connecting Different Topics

[Author]: Aurelius Zhongyuan Huang

[Last updated]: Feb 2nd, 2026

[Abstract] Large language models are increasingly used to explain new ideas by relating them to familiar ones, yet their ability to produce clear, consistent, and meaningful conceptual connections remains uneven and under-examined. This blog evaluates how different ChatGPT models perform on this specific task by testing a fixed prompt across multiple model variants and scoring the resulting explanations with a calibrated, rubric-based judgement system. Through comparative evaluation and statistical analysis, the study finds clear performance differences: advanced reasoning models generate more stable and higher-quality connections, while smaller or lighter models show weaker consistency. These results clarify which models are best suited for analogy-driven learning systems and point toward future improvements in prompt design and cost-efficient model selection.

Insights & Innovation

Blogs

Patents (U.S. Patent Pending)