LLMs use "safety" specific neuron layers to identify vulnerabilities in code 5 points by summarity 5 hours ago 2 comments story
Hacker News