a short list, kept honest.

reading

topics i want to dig into

• using model weights to reconstruct training data — how adversarial attacks (model inversion, membership inference) reveal what a model was trained on.