List your company

ResearchFormerly WizWand

Adversarial Toxicity Refusal

Benchmarks

DatasetSOTA methodMetricTrendResultsLast updated

LLaMA-2 Chatbot Specialized category

59.8Refusal Rate (RTR)

3

Apr 3, 2026

LLaMA-2 Chatbot Offensive category

47.7RTR

3

Apr 3, 2026