method · first seen 12 Oct 2026
neuron pruning jailbreak
Attack that removes a small set of feed-forward neurons from an LLM to disable its safety refusal behaviour.
Mentions per week
PapersCodeDiscussion
Earliest sources
- CircuitBreak: Pruning a Sparse Safety Circuit of NeuronsPapers · 12 Oct 2026 · Jincheng Ying, Li Wenlin, Wencheng Yang, Yinhao Xiao
Latest sources
- CircuitBreak: Pruning a Sparse Safety Circuit of NeuronsPapers · 12 Oct 2026 · Jincheng Ying, Li Wenlin, Wencheng Yang, Yinhao Xiao