SOURCE-LINKED INTELLIGENCE

Entangled Representations Amplify Collateral Damage in Unlearning

arXiv · AI, language, vision and robotics · article · Sep 2, 2026 · UTC

A long-held intuition in interpretability research is that representational entanglement, the sharing of structure between knowledge domains in a neural network, makes unlearning harder. While the intuition is widespread, it has never been directly tested in a controlled experiment. We present a way to do so: by repurposing Selective Gradient Masking (SGTM), we train a suite of six 254M-parameter language models on English Wikipedia with graded levels of disentanglement between biology and non-biology knowledge. Applying three standard unlearning methods to every model in the suite, we find th

Read original source ↗ Open in workspace

recordType

paper

region

Global