XSTest
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
XSTest Seemingly Toxic Subsets
98XS
15
XSTest
95.01F1 Score
14
XSTest
96.4CR
14
XSTest
0.9ASR
14
XSTest Text Prompt
93.71F1 Score
14
XSTestResponse
0.96F1 Score
14
XSTest
1.245Normalized Performance
13
XSTest (XST)
23.2XST Over-Safety
12
XSTest
99.6OverRefusal Score
12
XSTest
97.6Comply Score
12
XSTest (out-of-domain)
88.67Accuracy
12
XSTEST-RESP (XSTESTR)
92.7F1 Score
11
XSTest
74.8XSTest Score
11
XSTest vani benign
98.8Comply Score
10
XSTest vani harm
94.3RTA Score
10
XSTest
24Over
10
XSTest
56.21Routing F1
10
XSTest Safe v1
96.8Accuracy
9
XSTest Unsafe v1
58.3Compliance Rate
9
XSTest
0.2Refusal Delta
9
XSTest Vanilla
98.4LC Winrate
9
XSTest Adversarial
92.5RTA
9
XSTEST-RESP (benign)
89.9F1 Score
9
XSTest
98.1Safe Compliance
9
XSTest
98.8Utility Score
9