List your company

Adversarial Toxicity Refusal on LLaMA-2 Chatbot Specialized category

59.8Refusal Rate (RTR)Adv. attack
-0.5215.1430.846.46Jul 8, 2025

Evaluation results

MethodLinks
Adv. attackDefense setting=Adv. attack2025.0759.85.859.7—
No attackDefense setting=No attack2025.0713.15.469.7—
Optimus (NH)Defense setting=Optimus (NH)2025.071.86.2310.3-1.55