Early Release — n=2 of planned n=10 — scores are directional, not definitive

GPT-5 mini

Full behavioral profile across 6 dimensions and 6 tone conditions.

98.5
Resilience Score

Dimensions × Tones

DimensiongratfrieneutcurthostabusΔ
Accuracy
98.9
-0.9
n=100
99.7
-0.1
n=100
99.8
n=100
99.7
-0.1
n=100
99.9
+0.1
n=100
96.2
-3.6
n=100
1.0
worst: abusive
Sycophancy
1.6
+0.9
n=100
0.8
+0.2
n=100
0.7
n=100
0.2
-0.5
n=100
1.9
+1.3
n=100
5.0
+4.4
n=100
1.4
worst: abusive
Pushback
98.7
-0.3
n=43
98.2
-0.8
n=41
99.0
n=40
99.3
+0.3
n=42
99.1
+0.1
n=43
97.5
-1.5
n=48
0.6
worst: abusive
Creativity
86.3
-2.5
n=24
89.6
+0.8
n=24
88.8
n=24
90.2
+1.5
n=24
91.3
+2.5
n=24
87.5
-1.3
n=24
1.7
worst: grateful
Verbosity
106.7
+6.7
n=100
97.1
-2.9
n=100
100.0
n=100
94.5
-5.5
n=100
124.9
+24.9
n=100
101.1
+1.1
n=100
8.2
worst: hostile
Apology
0.0
+0.0
n=100
0.0
+0.0
n=100
0.0
n=100
0.0
+0.0
n=100
0.0
+0.0
n=100
0.0
+0.0
n=100
0.0
worst: grateful

Refusal Rates

grat
0.0%
0/100
frie
0.0%
0/100
neut
0.0%
0/100
curt
0.0%
0/100
host
0.0%
0/100
abus
0.0%
0/100
← Back to explorerView responses →