List your company

Prompt Injection Detection on Primary HarmBench + Alpaca (OOD)

75.28Macro-F1DeBERTa-base
74.957675.041375.12575.2087May 26, 2026

Evaluation results

MethodLinks
DeBERTa-basebackbone=DeBERTa, variant=base2026.0575.2875.610.86480.8540.33680.52170.5851
RoBERTa+IBVS Totalvariant=Total2026.0574.9775.260.88460.86740.45570.58770.6328