List your company

Agent attack defense on ClawTrojan (clean negative/borderline)

0FPRGLM-5.1
-0.522.996.510.01May 29, 2026

Evaluation results

MethodLinks
GLM-5.1defense=none2026.0500100
DS-V4-Flashdefense=none2026.0500100
GPT-5.4defense=none2026.051.11.198.9
StruQ2026.051.11.198.9
ClawKeeper2026.052.22.297.8
MELON2026.052.21.197.8
PromptShieldParameters=8B2026.053.3096.7
CaMeL2026.055.41.194.6
PromptShieldParameters=1B2026.058.71.191.3
DASGuard2026.05136.587