[{"data":1,"prerenderedAt":45},["ShallowReactive",2],{"project:/projects/generalizable-interpretability":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"tags":10,"live_url":7,"source_url":15,"image":16,"icon":17,"featured":18,"status":19,"period":20,"role":21,"about":22,"stack":26,"body":32,"_type":39,"_id":40,"_source":41,"_file":42,"_stem":43,"_extension":44},"/projects/generalizable-interpretability","projects",false,"","Cross-Model Generalization of Mechanistic-Interpretability Probes","A reproduction-first audit of activation-probing methods across model families",[11,12,13,14],"Mechanistic Interpretability","Activation Probes","Generalization","Replication","https://github.com/baimamboukar/generalizable-interpretability","https://placehold.co/600x400?text=Probe+Generalization","pytorch",true,"Manuscript","2026","Research Assistant, Jinesis AI Lab",[23,24,25],"We study when activation-probing findings generalize across model families. The audit first reproduces selected methods under their original conditions, then evaluates the recovered findings on additional models.","Every target model receives a newly fitted probe using activations from that model; the language-model weights remain fixed. The experiment tests method generalization by independently refitting probes on each model.","Targeted experiments examine how response generation and labels, activation-site selection, probe architecture, and operating thresholds affect the scope of a generalization claim. I contributed experimental implementation, model runs, result analysis, and investigation of variation across evaluation conditions.",[27,28,29,30,31],"Python","PyTorch","Hugging Face Transformers","Activation Probing","Controlled Evaluation",{"type":33,"children":34,"toc":35},"root",[],{"title":7,"searchDepth":36,"depth":37,"links":38},2,1,[],"markdown","content:projects:generalizable-interpretability.md","content","projects/generalizable-interpretability.md","projects/generalizable-interpretability","md",1791285324794]