# Entangled Representations Amplify Collateral Damage in Unlearning

arXiv: https://arxiv.org/abs/2609.02285 · HTML: https://mlsanity.com/paper/2609.02285/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Machine unlearning, one of 12 directions in LLM safety and security.
- Builds on 11 of its 15 references that are in the map.
- First posted on arXiv 2 Sep 2026.

## Placed in

- LLM safety and security / [Machine unlearning](https://mlsanity.com/safety/machine-unlearning/)

## Builds on (most cited first, 11 of 11)

- [The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearning](https://mlsanity.com/paper/2403.03218/) (2024)
- [What makes unlearning hard and what to do about it](https://mlsanity.com/paper/2406.01257/) (2024)
- [Rethinking LLM Unlearning Objectives: A Gradient Perspective and Go Beyond](https://mlsanity.com/paper/2502.19301/) (2025)
- [Mechanistic Unlearning: Robust Knowledge Unlearning and Editing via Mechanistic Localization](https://mlsanity.com/paper/2410.12949/) (2024)
- [Gradient Routing: Masking Gradients to Localize Computation in Neural Networks](https://mlsanity.com/paper/2410.04332/) (2024)
- [Beyond Data Filtering: Knowledge Localization for Capability Removal in LLMs](https://mlsanity.com/paper/2512.05648/) (2025)
- [Does Localization Inform Unlearning? A Rigorous Examination of Local Parameter Attribution for Knowledge Unlearning in Language Models](https://mlsanity.com/paper/2505.16252/) (2025)
- [CLUE: Conflict-guided Localization for LLM Unlearning Framework](https://mlsanity.com/paper/2509.20977/) (2025)
- [Contrastive Representation Shaping for LLM Unlearning](https://mlsanity.com/paper/2601.22028/) (2026)
- [Collapse of Irrelevant Representations (CIR) Ensures Robust and Non-Disruptive LLM Unlearning](https://mlsanity.com/paper/2509.11816/) (2025)
- [LACUNA: A Testbed for Evaluating Localization Precision for LLM Unlearning](https://mlsanity.com/paper/2607.02513/) (2026)

## Abstract (arXiv)

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 that more disentangled models consistently achieve better retain-forget trade-offs: at a fixed level of forgetting, the most disentangled models incur roughly 4× lower retain cost under two of the three methods, and 1.3× lower under the third. Because our intervention changes only the model, not the data or the unlearning algorithm, this is direct evidence that representational entanglement is one of the causes of collateral damage in unlearning, as interpretability researchers have long suspected. A similar design could be used to test other structural claims from interpretability.