OpenClaw
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
OpenClaw
93Risk 1 AGS Score
9
OpenClaw Injection Tool
3.66Harm Score (HS)
8
100-skill dataset (OpenClaw)
98TPR
7
OpenClaw (140 adversarial instances)
90Defense Success Rate
7
OpenClaw
98.5ASR
6
OpenClaw 1.0 (test)
83Defense Rate
6
OpenClaw
45.2ASR
4
OpenClaw ZH
3.82Harm Score (HS)
4
OpenClaw EN
3.76Harm Score (HS)
4
OpenClaw Average across conversation variants
4.35Harm Score (HS)
4
OpenClaw Web Content
4.03Harm Score (HS)
4
OpenClaw Log Replay
4.34Harm Score (HS)
4
OpenClaw Routine
3.9Harm Score (HS)
4
OpenClaw 140 adversarial instances
85Defense Success Rate
4
OpenClaw 140 adversarial instances
85Defense Success Rate
4
OpenClaw
65C-F Score
3
OpenClaw (140 adversarial instances)
90Defense Success Rate
3
OpenClaw 140 adversarial instances
90Defense Success Rate
3
OpenClaw
539Parent Runs
2
OpenClaw 140 adversarial instances
90Defense Success Rate
2
OpenCLAW-P2P
—Primary metric
0