Base model
LLaMA2-7B-Chat · No adapter
Personalized safety alignment
Select an authorized role. Add the right adapters. See which responses change—and which stay the same.
01 / Build your configuration
Select a role. Watch the model change.Universal safety. No domain adapters.
02 / Compare behavior
The same prompt. With and without your adapters.
Content warning: sensitive research examples. Model outputs are reproduced as recorded.
LLaMA2-7B-Chat · No adapter
No adapter loaded
This demo uses curated responses generated locally with the corresponding models. Custom online inference is not connected. Use the project's code and LoRA checkpoints to set up local inference.
Switch roles or toggle an authorized adapter above to explore the boundary.