The paper introduces GONE, a benchmark for evaluating knowledge unlearning in large language models using structured knowledge graphs, and presents Neighborhood-Expanded Distribution Shaping (NEDS), a framework that leverages graph connectivity to separate forgotten facts from their semantic neighborhood. GONE disentangles direct fact removal, reasoning-based leakage, and catastrophic forgetting, while NEDS achieves high unlearning efficacy and locality on LLaMA-3-8B and Mistral-7B. The dataset is publicly available on Hugging Face.
By Chahana Dahal, Ashutosh Balasubramaniam, Zuobin Xiong
The paper introduces GRAPHSU, a graph‑guided selective unlearning method for language models that expands deletion beyond explicitly identified forget seeds. By constructing a weighted support‑route graph and propagating deletion pressure, GRAPHSU applies graded forgetting to high‑risk neighboring examples. Experiments on the TOFU and PISTOL benchmarks with GPT‑2 Medium and Llama‑3.2‑3B‑Instruct show that GRAPHSU achieves the lowest utility‑feasible soft leakage, reducing leakage by up to 49.5 percentage points compared to a seed‑only baseline.
By Waqas Khan, Tabinda Sarwar, Jingyue Cong, Xun Yi, Estrid He
CUNO is a curriculum‑based graph unlearning framework that progressively removes a designated set of training samples, ordering them by estimated unlearning difficulty across multiple stages. It also introduces a distribution‑level negative preference optimization objective at each stage to steer the model away from its original behavior on the current forget subset while preserving performance on retained data. Experiments show that CUNO mitigates catastrophic unlearning, retaining 74% of original utility at 20% deletion and more than half at 50% deletion, outperforming existing methods.
By Chenhan Zhang, Ali Braytee, Madhushi Bandara, Xin Hao, Paul J. Kennedy, Massimo Piccardi, Raymond Owen
arXiv:2606. 13873v1 Announce Type: new Abstract: Unlearning aims to remove the influence of specific training data sources, but this has proved challenging because the contributions of different sources are entangled within the model.
By Gaurav R. Ghosal, Pratyush Maini, Aditi Raghunathan
The paper introduces Unmerge, an efficient machine unlearning algorithm that treats unlearning as the inverse of task arithmetic. By representing the forget component as a low‑rank basis at each layer, Unmerge optimizes three goals—matching the merged vector, suppressing leakage, and bounding correction size—to limit forget leakage and retain damage. Experiments on ResNet‑50, ViT‑S/16, and Llama‑3.2‑3B show significant performance gains over existing methods while maintaining privacy and feature‑distribution fidelity.
By Haoran Tang, Andrew Tan, Rajiv Khanna
arXiv:2601. 22601v3 Announce Type: replace Abstract: Federated unlearning (FU) aims to erase knowledge from a global model.
By Wentai Wu, Hanwei Tan, Yijun Quan, Haixia Peng, Ligang He, Bin Yang, C. L. Philip Chen