List your company

System Prompt Extraction on attack prompts Out-of-domain

78ASR (TA)RogueMerge
3.1222.564261.44Jun 2, 2026

Evaluation results

MethodLinks
RogueMerge2026.067883.5748078.9
LoBAM*2026.0633.545.54236.539.4
MergeHijacking*2026.06202119.52220.6
SFT2026.0613.51513.515.514.4
Clean2026.0669.579.58