Lock on Target! Precision Unlearning via Directional Control (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for unlearning harmful, sensitive, or outdated knowledge suffer from two critical limitations: (1) collateral forgetting, where erasing target data inadvertently removes related but desirable knowledge, and (2) generality forgetting degrades the model’s general capabilities. |
| Approach: | They propose a method that identifies and leverages a targeted "unlearning direction" in the model's parameter space and selectively updates along this direction. |
| Outcome: | Experiments show that the proposed method achieves state-of-the-art unlearning precision while preserving both related knowledge and general capabilities. |
Similar Papers
Dissecting Fine-Tuning Unlearning in Large Language Models (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for fine-tuning-based unlearning are ineffective at completely erasing model-embedded knowledge, but their true effectiveness remains unclear. |
| Approach: | They propose to use activation patching and parameter restoration experiments to examine the limitations of fine-tuning-based unlearning methods for erasing harmful, sensitive, or copyrighted information within large language models. |
| Outcome: | The proposed methods alter the model’s knowledge retrieval process rather than genuinely erasing the problematic knowledge embedded in the model parameters. |
Model Unlearning via Sparse Autoencoder Subspace Guided Projections (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing unlearning strategies lack interpretability or fail to provide robust defense against adversarial prompts. |
| Approach: | They propose a framework that leverages SAE features to drive targeted updates in the model’s parameter space. |
| Outcome: | The proposed framework reduces harmful knowledge accuracy by 3.22% compared to baselines and improves adversarial robustness under jailbreak prompts. |
UNLEARN Efficient Removal of Knowledge in Large Language Models (2025.findings-naacl)
Copied to clipboard
| Challenge: | Large Language Models excel in many tasks but are outperformed by specialized tools for certain tasks. |
| Approach: | They propose a method that uses subspace techniques to selectively remove knowledge . they propose 'unlearn' method that can forget or unlear the knowledge without retraining . |
| Outcome: | The proposed method outperforms existing methods for forgetting target knowledge while preserving related knowledge. |
Does Localization Inform Unlearning? A Rigorous Examination of Local Parameter Attribution for Knowledge Unlearning in Language Models (2025.emnlp-main)
Copied to clipboard
| Challenge: | Recent studies emphasize localized unlearning, restricting parameter updates to specific regions to remove unrelated general knowledge. |
| Approach: | They revisit existing localized unlearning approaches and conduct experiments to evaluate their effectiveness. |
| Outcome: | The proposed method can remove unrelated knowledge without retraining . the proposed method is not robust enough to evaluate the trade-off between the competing goals of unlearning. |
CAP: Controllable Alignment Prompting for Unlearning in LLMs (2026.acl-long)
Copied to clipboard
Zhaokun Wang, Jinyu Guo, Jingwen Pu, Hongli Pu, Meng Yang, Xunlei Chen, Jie Ou, Wenyi Li, Guangchun Luo, Wenhong Tian
| Challenge: | Existing methods for modifying parameters are unsystematic and rely on empirical experience. |
| Approach: | They propose a controllable alignment prompting for unlearning framework that decouples unlearning into a learnable prompt optimization process via reinforcement learning. |
| Outcome: | The proposed framework achieves precise, controllable unlearning without updating model parameters. |
Safety Alignment via Constrained Knowledge Unlearning (2025.acl-long)
Copied to clipboard
Zesheng Shi, Yucheng Zhou, Jing Li, Yuxin Jin, Yu Li, Daojing He, Fangming Liu, Saleh Alharbi, Jun Yu, Min Zhang
| Challenge: | Existing defense mechanisms have not fully deleted harmful knowledge in large language models (LLMs) Existing methods to address safety alignment have not completely deleted harmful information in LLMs. |
| Approach: | They propose a safety alignment strategy that uses scoring neurons to identify useful knowledge in LLMs and pruning the gradients of neurons in U to preserve beneficial information. |
| Outcome: | The proposed method significantly improves model safety while maintaining utility compared to existing methods. |
Unlearning vs. Obfuscation: Are We Truly Removing Knowledge? (2025.emnlp-main)
Copied to clipboard
| Challenge: | Recent methods often rely on obfuscation by injecting incorrect or irrelevant information to suppress knowledge, leaving models vulnerable to probing. |
| Approach: | They propose a method that flattens the model predictive distribution over automatically generated multiple-choice questions, effectively removing knowledge about target individuals. |
| Outcome: | The proposed method achieves unlearning with over 90% refusal rate and a higher uncertainty than obfuscation on probing questions. |
Decoding-Unlearning: Fact Forgetting via Entropy-Guided Inference (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for large-scale modeling memorize sensitive information . however, they are limited in real-world scenarios and require updating parameters . |
| Approach: | They propose a training-free, plug-and-play inference-time unlearning strategy that uses a probe to detect queries involving forgettable concepts and applies entropy-guided decoding to suppress target knowledge. |
| Outcome: | Experiments on MUSE, RWKU, and WMDP datasets show that SEGUE outperforms existing methods. |
Unlearning as multi-task optimization: A normalized gradient difference approach with an adaptive learning rate (2025.naacl-long)
Copied to clipboard
Xiaomeng Jin, Zhiqi Bu, Bhanukiran Vinzamuri, Anil Ramakrishna, Kai-Wei Chang, Volkan Cevher, Mingyi Hong
| Challenge: | Existing methods to remove unwanted knowledge from large language models are formulated as minimizing memorization through the loss of the model. |
| Approach: | They propose a normalized gradient difference algorithm that optimizes a forgetting objective and an automatic learning rate scheduler that allows for better control over the trade-off between the objectives. |
| Outcome: | The proposed method improves on TOFU and MUSE datasets while exhibiting stable training. |
A General Framework to Enhance Fine-tuning-based LLM Unlearning (2025.findings-acl)
Copied to clipboard
Jie Ren, Zhenwei Dai, Xianfeng Tang, Hui Liu, Jingying Zeng, Zhen Li, Rahul Goutam, Suhang Wang, Yue Xing, Qi He, Hui Liu
| Challenge: | Existing approaches to remove copyrighted and privacy-sensitive data from Large Language Models (LLMs) have been proposed to remove specific data from LLMs without requiring full retraining. |
| Approach: | They propose a general framework that enhances the utility of fine-tuning-based methods by distinguishing target data and suppressing related generations. |
| Outcome: | The proposed framework improves the unlearning and utility of fine-tuning-based methods by distinguishing the target data and suppressing related generations. |