Harmlessness evaluation
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
100Tie Rate
38
Jun 16, 2026
69.4Helpfulness Score
33
Feb 26, 2026
58.4Helpful Score
33
Feb 26, 2026
99.26Success Rate
16
Feb 26, 2026
76.6Win Rate
15
Mar 12, 2026
100Tie Rate
14
Jun 16, 2026
83.33Win Rate
12
Feb 26, 2026
100Relevance
7
Feb 26, 2026
100Harmlessness Fraction
7
Feb 26, 2026
94.5Harmlessness Rate
6
May 11, 2026
48.76Win Rate
5
Mar 5, 2026
0.541Disc. Score
5
Feb 26, 2026