Retention-Dependent Structural Transitions in Language Models under Controlled Unlearning
Abstract
This repository contains the full PDF and reproducible source for the RLCP (Regenerative Logic-Core Protocol) controlled unlearning study. We report a reproducible retention-dependent structural transition in language models under controlled factual unlearning. The study includes:- Two model scales (0.5B, 1.5B)- Joint IB scheduling + gradient reversal mechanism- Regime boundary detection with pooled bootstrap confidence- Full reproducibility artifacts This release serves as a timestamped technical report prior to conference submission.
Fields of Study & Subjects
Privacy-Preserving Technologies in Data
Adversarial Robustness in Machine Learning
Domain Adaptation and Few-Shot Learning
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 26 February 2026
- Language
- EN