"BadLlama: cheaply removing safety fine-tuning from Llama 2-Chat 13B."

Pranav Gade et al. (2023)

Details and statistics

DOI: 10.48550/ARXIV.2311.00117

access: open

type: Informal or Other Publication

metadata version: 2023-11-07

a service of  Schloss Dagstuhl - Leibniz Center for Informatics