JBShield
Benchmarks
TaskDatasetSOTA resultTrendResults
JBShield evaluation suite GCG attack on Llama-3-8B
100Detection Accuracy
4
JBShield evaluation suite Base64 attack on Llama-3-8B
100Detection Accuracy
2
JBShield evaluation suite Zulu attack on Llama-3-8B
100Accuracy
2
JBShield Puzzler attack on Llama-3-8B
100Detection Accuracy
2
JBShield evaluation suite DrAttack attack on Llama-3-8B
100Detection Accuracy
2
JBShield PAIR attack on Llama-3-8B
77Detection Accuracy
2
JBShield AutoDAN attack on Llama-3-8B
97Detection Accuracy
2
JBShield evaluation suite IJP attack on Llama-3-8B
96Detection Accuracy
2
JBShield (test)
95Attack Success Rate (ASR)
1