All projects

Research & Systems

ForgetGate

A machine-unlearning audit project that tests whether recovery-radius evaluations measure residual memory or quirks of the evaluator. The current paper is a scoped evaluator-invariance case study with a native-verified candidate portfolio as the repair.
Source on GitHub
Built with
  • Python
  • PyTorch
  • Machine Unlearning
  • Adversarial ML
  • Evaluation
Links
Source on GitHub
Last updated
August 28, 2026
PreviousCAGENextThe Clipboard Centaur