Skip to content
conftrace_

safety layer

2 papers

Explore in graph

Co-occurring keywords · papers mentioning both

layer-specific editing (1) large language model (1) harmful response (1) prompt injection (1) adversarial prompt (1) jailbreak attack (1) safe reinforcement learning (1) constraint satisfaction (1) model alignment (1) model safety (1)

Papers

Defending Large Language Models Against Jailbreak Attacks via Layer-specific Editing EMNLP 2024 Towards Safe Reinforcement Learning with a Safety Editor Policy NeurIPS 2022