DL4SECAboutArchive / Independent research
Technical walkthroughs, experiments, and notes from applied ML/DL for security.
Post-training safety looks robust until you find the one direction in activation space that carries refusal. Ablate it and the model stops saying no, which tells us alignment is thinner than it appears.
While agents dominate AI headlines, embeddings quietly power most production systems. A deep dive into what they are, why they matter and how they enable scalable AI beyond context window limits.
Transform basic models into production-ready systems through proper dataset splitting, evaluation metrics, optimization techniques and reproducibility practices
Understand the mathematical foundations of neural network training by implementing gradient descent from scratch, then apply these concepts to build and train a working classifier
Explores essential feature extraction techniques for translating complex information, like URLs or binary files, into numerical formats that deep learning models can understand and learn from
Understanding AI's reasoning and learning approaches, how machine learning power intelligent agents and the data scientist's mindset for building effective models
No posts match this filter.