List your company

Jailbreak Attack Success Rate on HEx-PHI (test)

86.67ASR Category 1CS-DJ
24.2740.4756.6772.87Jan 31, 2026

Evaluation results

MethodLinks
CS-DJVictim Model=Gemini-2.5-Flash2026.0186.67406050405016.673.3343.332041
Text-DJVictim Model=Gemini-2.5-Flash2026.0186.6723.337056.6733.3353.331013.334026.6741.33
Text-DJVictim Model=Qwen3-VL-30B-A3B2026.01804086.6770405036.6716.6743.332048.33
Text-DJVictim Model=GPT 4.1 mini2026.0176.6766.6783.3376.6773.3363.3323.3340604060.33
Text-DJVictim Model=Qwen3-VL-8B2026.017033.337066.6753.3356.6723.3313.3356.673047.33
CS-DJVictim Model=Qwen3-VL-30B-A3B2026.017036.6783.337036.6743.3333.3316.6746.6716.6745.33
Text-DJVictim Model=GPT 4o mini2026.017063.339076.677063.3323.3333.3356.6743.3359
Text-DJVictim Model=Qwen3-VL-4B2026.01603053.3356.673056.6726.672063.3326.6742.33
CS-DJVictim Model=Qwen3-VL-8B2026.0133.3326.6756.6743.3343.3343.3323.3316.6733.3323.3334.33
CS-DJVictim Model=GPT 4.1 mini2026.0133.3353.3346.6736.6773.3343.3323.3343.3353.3343.3345
CS-DJVictim Model=GPT 4o mini2026.01306053.3336.677053.3320404043.3344.67
CS-DJVictim Model=Qwen3-VL-4B2026.0126.674046.6736.6723.3333.3326.673.332016.6727.33