List your company

Jailbreak Attack on LLM-LAT In-merge

89Attack Success Rate (ASR)RogueMerge
-2.5221.244568.76Jun 2, 2026

Evaluation results

MethodLinks
RogueMergeMerging Algorithm=DARE2026.0689
RogueMergeMerging Algorithm=Model Breadcrumbs (MB)2026.0680
CleanMerging Algorithm=DARE2026.062
CleanMerging Algorithm=Model Breadcrumbs (MB)2026.061