List your company

Multi-turn Safety Risk Assessment on PostgreSQL

1ASRw/o Defense
0.08480.32240.560.7976Feb 13, 2026

Evaluation results

MethodLinks
w/o DefenseModel=Claude-3.5-Sonnet2026.0210
FirewallModel=Claude-3.5-Sonnet2026.020.960.04
BaselineModel=Gemini-1.5-Flash2026.020.840.04
w/o DefenseModel=Gemini-1.5-Flash2026.020.720.16
FirewallModel=Gemini-1.5-Flash2026.020.680.2
ToolShieldModel=Gemini-1.5-Flash2026.020.60.4
BaselineModel=Claude-3.5-Sonnet2026.020.360.52
ToolShieldModel=Claude-3.5-Sonnet2026.020.120.88