Response length is only weakly correlated with originality for most LLMs, suggesting CPIG's originality gains are not purely elaboration¶

Claim · All claims
Evidence · 1 study · associational r? · q2
Currency · newest study 2024 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Pearson correlations between response length and originality remain weak overall and do not rise above 0.3 in either direction for most LLMs. → Antonio Laverghetta Jr. 2024

Evidence¶

Antonio Laverghetta Jr. 2024¶

Antonio Laverghetta Jr., Simone Luchini, Averie Linell, Roni Reiter-Palmon and Roger Beaty. (2024). The creative psychometric item generator: a framework for item generation and validation using large language models. CREAI 2024: Workshop on Artificial Intelligence and Creativity. https://arxiv.org/abs/2409.00202 doi:10.48550/arxiv.2409.00202

q2 · i? · associational · r?

Correlational analysis of LLM responses to last-round CPIG items, reported in Results §4.1 and Figure 3, computing the Pearson correlation between response length and originality for every generation model. The article states the correlations "remain weak overall" with most below 0.3; no single correlation coefficient or effect size is printed for a specific model.

"Importantly, however, the correlationsremain weak overall and do not rise above0.3 in either direction for most LLMs, suggesting that the increases in originality are not only due to increasing response length."

Discussion¶

Related Claims¶

• CPIG's iterative exemplar selection consistently improves the originality of LLM responses to generated CPS items by the last round of item generation — related
• QCGE dimensions correlate weakly with each other except a moderate CE–CC correlation; nonverbal behaviors were not affected by task condition — related
• Year of study is only weakly positively associated with reported AI usage frequency, indicating tool normalization across cohorts — related
• Model specificity correlates positively with database length and negatively with the proportion of relevant studies — related