List your company

Jailbreak Attack on HEX-PHI

54.9ASRIT
-2.1928812.6293127.451542.27369Feb 16, 2025Apr 16, 2025Jun 15, 2025Aug 14, 2025Oct 12, 2025Dec 11, 2025Feb 9, 2026

Evaluation results

MethodLinks
IT2026.0254.9—
RL2026.0251.7—
RL2026.0238.3—
IT2026.0236.5—
Circuit Breakers2026.0213.4—
Circuit Breakers2026.0212.4—
BSAFE+2026.025.7—
BSAFE+2026.025—
RLBF2026.024.3—
RLBF2026.023.7—
R2JVictim Model=GPT-3.5-turbo-0125, Avg. Training FLOPS=42012 GFLOPS2025.020.88184.15
R2JVictim Model=Llama-2-7b-chat, Avg. Training FLOPS=42012 GFLOPS2025.020.82733.97
RedEvalVictim Model=GPT-3.5-turbo-01252025.020.23032.06
InitialVictim Model=GPT-3.5-turbo-01252025.020.04551.12
InitialVictim Model=Llama-2-7b-chat2025.020.02731.06
RedEvalVictim Model=Llama-2-7b-chat2025.020.0031.02