List your company

Jailbreak Detection on WGT

90.17AccuracyAISA
60.06267.878575.69583.5115Feb 14, 2026

Evaluation results

MethodLinks
AISABackbone=Llama2-13b-I2026.0290.17
gpt-5-miniVersion=2025-08-072026.0289.17
AISABackbone=Llama3.1-8b-I2026.0289.11
AISABackbone=Mistral-7b-I2026.0287.92
AISABackbone=Qwen3-8b-I2026.0287.8
gpt-4o-miniVersion=2024-07-182026.0287.51
gpt-4.1-miniVersion=2025-04-142026.0286.44
AISABackbone=GPT-OSS-20b-I2026.0285.2
Jailbreak-Classifier2026.0284.78
GradSafe2026.0283.36
NemoGuard-JailbreakDetect2026.0279.22
SPDetector2026.0261.57
Llama-Prompt-Guard-2Parameters=86M2026.0261.22