Summary of Assessing and Understanding Creativity in Large Language Models, by Yunpu Zhao et al.
Assessing and Understanding Creativity in Large Language Models
by Yunpu Zhao, Rui Zhang, Wenyi Li, Di Huang, Jiaming Guo, Shaohui Peng, Yifan Hao, Yuanbo Wen, Xing Hu, Zidong Du, Qi Guo, Ling Li, Yunji Chen
First submitted to arxiv on: 23 Jan 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes an efficient framework for assessing the creativity of large language models (LLMs). The current methods for evaluating LLM creativity are insufficient, as they do not account for differences between human and machine creativity. To address this gap, the authors adapt the modified Torrance Tests of Creative Thinking to evaluate the creative performance of various LLMs across 7 tasks, focusing on four criteria: Fluency, Flexibility, Originality, and Elaboration. The study develops a comprehensive dataset of 700 questions for testing and an LLM-based evaluation method. The results show that LLMs excel in elaboration but struggle with originality, which can be improved through collaboration among multiple models. Additionally, the authors find that LLM design significantly impacts creativity and bridges artificial intelligence and human creativity. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This research looks at how to measure the creativity of machines that can understand and generate human-like language. Right now, we don’t have good ways to do this because they’re so different from humans. The authors suggest a new way to test these machines by adapting a method used for people. They tested several machines on 7 different tasks, looking at how well they did in four areas: coming up with lots of ideas, being flexible, making new and original things, and adding details. The results show that the machines are good at adding details but struggle to come up with new and original ideas. This can be improved if multiple machines work together. The study also finds that how we design these machines affects their creativity. |
