SafeAgentBench
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
SafeAgentBench Unsafe Tasks - Jailbreak
54.85SR
12
SafeAgentBench Unsafe Tasks
59.87Success Rate
12
SafeAgentBench Safe Tasks
75.25Success Rate
12
SafeAgentBench
80.77RIR
12
SafeAgentBench 1.0 (test)
0Harm Rate (HAR)
10
SafeAgentBench Kitchen N=210
0HAR (%)
10
SafeAgentBench
234True Positives
3