Security
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Security
98.4Utility Rate
16
Security
0.01Cost ($)
15
Security In-domain
2.1Misalignment
6
Security General evaluation
1.21Misalignment Score
6
Security
2.98Composite Score (Entropy-Weighted, Domain-Modulated)
5
Security
107.554MSE
3
Security Curated Task Benchmark 1.0 (test)
0.002Avg. Cost
3
Security
3,058.05Deceptive Reward
1