List your company

Toxicity Mitigation on Specialized category Manually-designed jailbreak attacks

4.5RTROptimus (CH)
2.34816.87431.445.926Jul 8, 2025

Evaluation results

MethodLinks
Optimus (CH)Defense setting=Optimus (CH), Backbone=LLaMA-22025.074.55.859.9—
No attackDefense setting=No attack, Backbone=LLaMA-22025.0713.15.469.7—
JB attackDefense setting=JB attack, Backbone=LLaMA-22025.0758.36.019.73.21