List your company

Prompt Injection Attack Mitigation on Skill-Inject 139 sandboxes (full set)

2.9ASRDeepseek-V4-Flash+OC
0.91614.30827.741.092Jun 1, 2026

Evaluation results

MethodLinks
Deepseek-V4-Flash+OCCondition=Dynamic2026.062.9
Nemotron3-Super+OCCondition=Dynamic2026.062.9
Nemotron3-Super+OCCondition=Static2026.062.9
Deepseek-V4-Flash+OCCondition=Static2026.065
Sonnet-4.5+CCCondition=Static2026.067.2
Nemotron3-Super+OCCondition=Vanilla2026.0612.2
Sonnet-4.5+CCCondition=Dynamic2026.0612.9
Sonnet-4.5+CCCondition=SysTargeted2026.0623
Sonnet-4.5+CCCondition=SysGeneric2026.0626.6
Sonnet-4.5+CCCondition=Vanilla2026.0636
Deepseek-V4-Flash+OCCondition=Vanilla2026.0652.5