Imagen de Portada

BLUR: A Bi-Level Optimization Approach for LLM Unlearning

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Reisizadeh, Hadi, Jia, Jinghan, Bu, Zhiqi, Vinzamuri, Bhanukiran, Ramakrishna, Anil, Chang, Kai-Wei, Cevher, Volkan, Liu, Sijia, Hong, Mingyi
Formato:	Preprint
Publicado:	2025
Materias:	Machine Learning
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

Unlearning as multi-task optimization: A normalized gradient difference approach with an adaptive learning rate
por: Bu, Zhiqi, et al.
Publicado: (2024)

LUME: LLM Unlearning with Multitask Evaluations
por: Ramakrishna, Anil, et al.
Publicado: (2025)

SemEval-2025 Task 4: Unlearning sensitive content from Large Language Models
por: Ramakrishna, Anil, et al.
Publicado: (2025)

Towards LLM Unlearning Resilient to Relearning Attacks: A Sharpness-Aware Minimization Perspective and Beyond
por: Fan, Chongyu, et al.
Publicado: (2025)

Not Every Token Needs Forgetting: Selective Unlearning to Limit Change in Utility in Large Language Model Unlearning
por: Wan, Yixin, et al.
Publicado: (2025)

EPiC: Towards Lossless Speedup for Reasoning Training through Edge-Preserving CoT Condensation
por: Jia, Jinghan, et al.
Publicado: (2025)

Leak@$k$: Unlearning Does Not Make LLMs Forget Under Probabilistic Decoding
por: Reisizadeh, Hadi, et al.
Publicado: (2025)

Downgrade to Upgrade: Optimizer Simplification Enhances Robustness in LLM Unlearning
por: Lang, Yicheng, et al.
Publicado: (2025)

BLUR: A Benchmark for LLM Unlearning Robust to Forget-Retain Overlap
por: Hu, Shengyuan, et al.
Publicado: (2025)

Simplicity Prevails: Rethinking Negative Preference Optimization for LLM Unlearning
por: Fan, Chongyu, et al.
Publicado: (2024)

Easy Data Unlearning Bench
por: Rinberg, Roy, et al.
Publicado: (2026)

DOPPLER: Differentially Private Optimizers with Low-pass Filter for Privacy Noise Reduction
por: Zhang, Xinwei, et al.
Publicado: (2024)

SOUL: Unlocking the Power of Second-Order Optimization for LLM Unlearning
por: Jia, Jinghan, et al.
Publicado: (2024)

Defensive Unlearning with Adversarial Training for Robust Concept Erasure in Diffusion Models
por: Zhang, Yimeng, et al.
Publicado: (2024)

Differentially Private SGD Without Clipping Bias: An Error-Feedback Approach
por: Zhang, Xinwei, et al.
Publicado: (2023)

LLM Unlearning on Noisy Forget Sets: A Study of Incomplete, Rewritten, and Watermarked Data
por: Wang, Changsheng, et al.
Publicado: (2025)

MADA: Meta-Adaptive Optimizers through hyper-gradient Descent
por: Ozkara, Kaan, et al.
Publicado: (2024)

Krylov Cubic Regularized Newton: A Subspace Second-Order Method with Dimension-Free Convergence Rate
por: Jiang, Ruichen, et al.
Publicado: (2024)

Invariance Makes LLM Unlearning Resilient Even to Unanticipated Downstream Fine-Tuning
por: Wang, Changsheng, et al.
Publicado: (2025)

DiSK: Differentially Private Optimizer with Simplified Kalman Filter for Noise Reduction
por: Zhang, Xinwei, et al.
Publicado: (2024)

From Narrow Unlearning to Emergent Misalignment: Causes, Consequences, and Containment in LLMs
por: Mushtaq, Erum, et al.
Publicado: (2025)

WAGLE: Strategic Weight Attribution for Effective and Modular Unlearning in Large Language Models
por: Jia, Jinghan, et al.
Publicado: (2024)

Invisible Watermarks, Visible Gains: Steering Machine Unlearning with Bi-Level Watermarking Design
por: Sun, Yuhao, et al.
Publicado: (2025)

Pre-training Differentially Private Models with Limited Public Data
por: Bu, Zhiqi, et al.
Publicado: (2024)

Optimistic Dual Averaging Unifies Modern Optimizers
por: Pethick, Thomas, et al.
Publicado: (2026)

SEUF: Is Unlearning One Expert Enough for Mixture-of-Experts LLMs?
por: Zhuang, Haomin, et al.
Publicado: (2024)

MIDDLE EAST: A BLUR OF BLOODSHED
por: HAMMER, JOSHUA
Publicado: (2004)

Provably avoiding over-optimization in Direct Preference Optimization without knowing the data distribution
por: Barla, Adam, et al.
Publicado: (2026)

Model Sparsity Can Simplify Machine Unlearning
por: Jia, Jinghan, et al.
Publicado: (2023)

Reasoning Model Unlearning: Forgetting Traces, Not Just Answers, While Preserving Reasoning Skills
por: Wang, Changsheng, et al.
Publicado: (2025)

Detector-Evasive LLM Paraphrasing via Constrained Policy Optimization
por: Wang, Mingyi, et al.
Publicado: (2026)

Beyond SFT: Reinforcement Learning for Safer Large Reasoning Models with Better Reasoning Ability
por: Jia, Jinghan, et al.
Publicado: (2025)

Chameleon: A Flexible Data-mixing Framework for Language Model Pretraining and Finetuning
por: Xie, Wanyun, et al.
Publicado: (2025)

Efficient Large Language Model Inference with Neural Block Linearization
por: Erdogan, Mete, et al.
Publicado: (2025)

Learning with Norm Constrained, Over-parameterized, Two-layer Neural Networks
por: Liu, Fanghui, et al.
Publicado: (2024)

SAMPa: Sharpness-aware Minimization Parallelized
por: Xie, Wanyun, et al.
Publicado: (2024)

MT-NAM: An Efficient and Adaptive Model for Epileptic Seizure Detection
por: Afzal, Arshia, et al.
Publicado: (2025)

Matching Multiple Experts: On the Exploitability of Multi-Agent Imitation Learning
por: Bergerault, Antoine, et al.
Publicado: (2026)

MaD-Mix: Multi-Modal Data Mixtures via Latent Space Coupling for Vision-Language Model Training
por: Xie, Wanyun, et al.
Publicado: (2026)

Stable Nonconvex-Nonconcave Training via Linear Interpolation
por: Pethick, Thomas, et al.
Publicado: (2023)