https://mahmoud-a.com/ 2026-03-01 weekly 1.0 https://mahmoud-a.com/blog 2026-03-01 weekly 0.9 https://mahmoud-a.com/#about 2026-03-01 monthly 0.8 https://mahmoud-a.com/#projects 2026-03-01 monthly 0.8 https://mahmoud-a.com/#skills 2026-03-01 monthly 0.7 https://mahmoud-a.com/#education 2026-03-01 monthly 0.7 https://mahmoud-a.com/#contact 2026-03-01 monthly 0.7 https://mahmoud-a.com/blog/activation-guardrails-part-1 2026-03-02 monthly 0.8 https://mahmoud-a.com/blog/ActBlog.png Activation-Space Safety Guardrails for LLMs Building safety rails using a model's own activations: kNN and direction-based classification in activation space https://mahmoud-a.com/blog/activation-guardrails-part-2 2026-03-02 monthly 0.8 https://mahmoud-a.com/blog/ActBlogPt2.png Topic Control and Dual-Use Risks in Activation Steering Extending activation-space guardrails to topic control, and examining the dual-use risks of steering directions