Probing the Creativity of Large Language Models: Can models produce divergent semantic association?

Large language models possess remarkable capacity for processing language,but it remains unclear whether these models can further generate creativecontent. The present study aims to investigate the creative thinking of largelanguage models through a cognitive perspective. We utilize the divergentassociation task (DAT), an objective measurement of creativity that asks modelsto generate unrelated words and calculates the semantic distance between them.We compare the results across different models and decoding strategies. Ourfindings indicate that: (1) When using the greedy search strategy, GPT-4outperforms 96

Further reading#

• Access Paper in arXiv.org