
Inference scaling for LLM safety assurance

The code was tested under the following dependencies:
typing_extensions==4.14.0numpy==2.2.6torch==2.5.1huggingface_hub==0.30.2accelerate==1.1.1datasets==3.1.0evaluate==0.4.3transformers==4.45.2peft==0.15.0deepspeed==0.16.7See saffron.py. More descriptions coming soon...
