Text Classification
Safetensors
English
deberta-v2
prompt-injection
security
deberta-v3
aegis
Eval Results (legacy)
yatavent's picture
v1: Initial English injection detector (deberta-v3-base, 74K data, 99.53% acc, 99.42% F1, 0% FPR on boundary tests)
5ee85f3 verified