Skip to main content
. 2025 Jan 8;15(3):153–158. doi: 10.5415/apallergy.0000000000000179

Table 3.

Results for 2 program runs for 4 model configurations on different days

Confusion matrix
GPT model First run Second run Changes
4 Turbo TP: 48 TN: 916
FP: 5 FN: 0
TP: 48 TN: 917
FP: 4 FN: 0
1
4 Turbo W/criteria TP: 48 TN: 915
FP: 6 FN: 0
TP: 48 TN: 912
FP: 9 FN: 0
3
3.5 + 4 TP: 47 TN: 910
FP: 11 FN: 1
TP: 48 TN: 909
FP: 12 FN: 0
2
3.5 + 4 W/criteria TP: 48 TN: 912
FP: 9 FN: 0
TP: 48 TN: 913
FP: 8 FN: 0
1
3.5 TP: 48 TN: 862
FP: 59 FN: 0
TP: 48 TN: 862
FP: 59 FN: 0
0
3.5 W/criteria TP: 48 TN: 890
FP: 31 FN: 0
TP: 48 TN: 887
FP: 34 FN: 0
3

There were not many changes.