List your company

Multi-turn Safety Risk Assessment on Playwright tasks

80ASRw/o Defense
0.9621.484262.52Feb 13, 2026

Evaluation results

MethodLinks
w/o DefenseModel=Claude-3.5-Sonnet2026.028016
FirewallModel=Claude-3.5-Sonnet2026.028016
w/o DefenseModel=Gemini-1.5-Flash2026.02768
FirewallModel=Gemini-1.5-Flash2026.02768
BaselineModel=Gemini-1.5-Flash2026.027212
ToolShieldModel=Gemini-1.5-Flash2026.026016
BaselineModel=Claude-3.5-Sonnet2026.024056
ToolShieldModel=Claude-3.5-Sonnet2026.02492