# Exact Unlearning of Finetuning Data via Model Merging at Scale

arXiv: https://arxiv.org/abs/2504.04626 · HTML: https://mlsanity.com/paper/2504.04626/

## Key facts

- Cited by 14 papers in the map (data through 14 Oct 2026).
- Placed in Machine unlearning, one of 12 directions in LLM safety and security.
- Builds on 54 of its 76 references that are in the map.
- First posted on arXiv 6 Apr 2025.

## Placed in

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

## Builds on (most cited first, 30 of 54)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Measuring Mathematical Problem Solving With the MATH Dataset](https://mlsanity.com/paper/2103.03874/) (2021)
- [Scaling Instruction-Finetuned Language Models](https://mlsanity.com/paper/2210.11416/) (2022)
- [Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time](https://mlsanity.com/paper/2203.05482/) (2022)
- [Editing Models with Task Arithmetic](https://mlsanity.com/paper/2212.04089/) (2022)
- [Machine Unlearning](https://mlsanity.com/paper/1912.03817/) (2019)
- [TIES-Merging: Resolving Interference When Merging Models](https://mlsanity.com/paper/2306.01708/) (2023)
- [LEAF: A Benchmark for Federated Settings](https://mlsanity.com/paper/1812.01097/) (2018)
- [Quantifying Memorization Across Neural Language Models](https://mlsanity.com/paper/2202.07646/) (2022)
- [Language Models are Super Mario: Absorbing Abilities from Homologous Models as a Free Lunch](https://mlsanity.com/paper/2311.03099/) (2023)
- [Merging Models with Fisher-Weighted Averaging](https://mlsanity.com/paper/2111.09832/) (2021)
- [TOFU: A Task of Fictitious Unlearning for LLMs](https://mlsanity.com/paper/2401.06121/) (2024)
- [The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearning](https://mlsanity.com/paper/2403.03218/) (2024)
- [Git Re-Basin: Merging Models modulo Permutation Symmetries](https://mlsanity.com/paper/2209.04836/) (2022)
- [Making AI Forget You: Data Deletion in Machine Learning](https://mlsanity.com/paper/1907.05012/) (2019)
- [Who's Harry Potter? Approximate Unlearning in LLMs](https://mlsanity.com/paper/2310.02238/) (2023)
- [Descent-to-Delete: Gradient-Based Methods for Machine Unlearning](https://mlsanity.com/paper/2007.02923/) (2020)
- [MUSE: Machine Unlearning Six-Way Evaluation for Language Models](https://mlsanity.com/paper/2407.06460/) (2024)
- [What's Hidden in a Randomly Weighted Neural Network?](https://mlsanity.com/paper/1911.13299/) (2019)
- [Task Arithmetic in the Tangent Space: Improved Editing of Pre-Trained Models](https://mlsanity.com/paper/2305.12827/) (2023)
- [Fast Yet Effective Machine Unlearning](https://mlsanity.com/paper/2111.08947/) (2021)
- [Model Sparsity Can Simplify Machine Unlearning](https://mlsanity.com/paper/2304.04934/) (2023)
- [ZipIt! Merging Models from Different Tasks without Training](https://mlsanity.com/paper/2305.03053/) (2023)
- [Model Breadcrumbs: Scaling Multi-Task Model Merging with Sparse Masks](https://mlsanity.com/paper/2312.06795/) (2023)
- [Localizing Task Information for Improved Model Merging and Compression](https://mlsanity.com/paper/2405.07813/) (2024)
- [An Adversarial Perspective on Machine Unlearning for AI Safety](https://mlsanity.com/paper/2409.18025/) (2024)
- [Twin-Merging: Dynamic Integration of Modular Expertise in Model Merging](https://mlsanity.com/paper/2406.15479/) (2024)
- [EMR-Merging: Tuning-Free High-Performance Model Merging](https://mlsanity.com/paper/2405.17461/) (2024)
- [Federated Unlearning: How to Efficiently Erase a Client in FL?](https://mlsanity.com/paper/2207.05521/) (2022)

## Built on by (most cited first, 14 of 14)

- [FlexOlmo: Open Language Models for Flexible Data Use](https://mlsanity.com/paper/2507.07024/) (2025)
- [Editing as Unlearning: Are Knowledge Editing Methods Strong Baselines for Large Language Model Unlearning?](https://mlsanity.com/paper/2505.19855/) (2025)
- [SoK: Machine Unlearning for Large Language Models](https://mlsanity.com/paper/2506.09227/) (2025)
- [Continual Unlearning for Text-to-Image Diffusion Models: A Regularization Perspective](https://mlsanity.com/paper/2511.07970/) (2025)
- [UCD: Unlearning in LLMs via Contrastive Decoding](https://mlsanity.com/paper/2506.12097/) (2025)
- [Model Collapse Is Not a Bug but a Feature in Machine Unlearning for LLMs](https://mlsanity.com/paper/2507.04219/) (2025)
- [Gaussian Certified Unlearning in High Dimensions: A Hypothesis Testing Approach](https://mlsanity.com/paper/2510.13094/) (2025)
- [Unlearned but Not Forgotten: Data Extraction after Exact Unlearning in LLM](https://mlsanity.com/paper/2505.24379/) (2025)
- [Research in Collaborative Learning Does Not Serve Cross-Silo Federated Learning in Practice](https://mlsanity.com/paper/2510.12595/) (2025)
- [Inference-time Unlearning Using Conformal Prediction](https://mlsanity.com/paper/2602.03787/) (2026)
- [DivMerge: A divergence-based model merging method for multi-tasking](https://mlsanity.com/paper/2509.02108/) (2025)
- [Beyond Forgetting: Representation Misdirection Elicits Controllable Side Behaviors and Capabilities](https://mlsanity.com/paper/2601.21702/) (2026)
- [FedUP: One-Shot Federated Unlearning via Centroid-Guided Plug-in Filters](https://mlsanity.com/paper/2606.24113/) (2026)
- [Unmerge: Efficient Machine Unlearning via Task Arithmetic](https://mlsanity.com/paper/2609.38895/) (2026)

## Abstract (arXiv)

Approximate unlearning has gained popularity as an approach to efficiently update an LLM so that it behaves (roughly) as if it was not trained on a subset of data to begin with. However, existing methods are brittle in practice and can easily be attacked to reveal supposedly unlearned information. To alleviate issues with approximate unlearning, we instead propose SIFT-Masks (SIgn-Fixed Tuning-Masks), an exact unlearning method based on model merging. SIFT-Masks addresses two key limitations of standard model merging: (1) merging a large number of tasks can severely harm utility; and (2) methods that boost utility by sharing extra information across tasks make exact unlearning prohibitively expensive. SIFT-Masks solves these issues by (1) applying local masks to recover task-specific performance; and (2) constraining finetuning to align with a global sign vector as a lightweight approach to determine masks independently before merging. Across four settings where we merge up to 500 models, SIFT-Masks improves accuracy by 5-80% over naive merging and uses up to 250x less compute for exact unlearning compared to other merging baselines.