Skip to main content
. 2023 Nov 22;13:20512. doi: 10.1038/s41598-023-46995-z

Table 6.

Results of the correlation analysis with Pearson correlation coefficient and obtained p-value given in the brackets along with p-value obtained from the Mann–Whitney U test comparing the values of the discrimination power index for correct and incorrect answers for temperature parameter equal to 1.

S22 A22 S23
Polish
 GPT-3.5
  Pearson correlation coefficient (p-value) − 0.092 (0.200 ns) − 0.235 (< 0.001***) − 0.197 (0.006**)
  p-value from Mann–Whitney U test 0.156 ns  < 0.001*** 0.009**
  Cohen’s d 0.184 0.487 0.411
 GPT-4
  Pearson correlation coefficient (p-value) − 0.037 (0.607 ns) − 0.253 (< 0.001***) − 0.190 (0.008**)
  p-value from Mann–Whitney U test 0.294 ns  < 0.001*** 0.037*
  Cohen’s d 0.086 0.636 0.512
English
 GPT-3.5
  Pearson correlation coefficient (p-value) − 0.098 (0.173 ns) − 0.191 (0.007**) − 0.109 (0.129 ns)
  p-value from Mann–Whitney U test 0.116 ns 0.002** 0.171
  Cohen’s d 0.198 0.395 0.221
 GPT-4
  Pearson correlation coefficient (p-value) − 0.050 (0.485 ns) − 0.151 (0.034*) − 0.045 (0.555 ns)
  p-value from Mann–Whitney U test 0.346 ns 0.055 ns 0.929 ns
  Cohen’s d 0.121 0.386 0.106