Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

DualOptim+: Bridging Shared and Decoupled Optimizer States for Better Machine Unlearning in Large Language Models

About

We propose DualOptim+, a novel optimization framework for improving machine unlearning in large language models. It introduces a base state to capture common representations shared by forgetting and retaining objectives and delta states to preserve objective-specific residuals. This architecture allows the optimizer to adaptively bridge shared and decoupled states based on the directional conflict between forgetting and retaining gradients. We further introduce DualOptim+ 8bit, a quantized variant that reduces memory overhead without compromising performance. Extensive experiments across fictitious and real-world unlearning, safety alignment, and multi-task learning tasks demonstrate that DualOptim+ consistently achieves a superior trade-off between different objectives. Codes are available at https://github.com/CityU-MLO/DualOptimPlus.

Xuyang Zhong, Qizhang Li, Yiwen Guo, Chen Liu• 2026

Related benchmarks

TaskDatasetResultRank
Machine UnlearningTOFU Forget05 Phi-1.5B model (5%)
Model Utility (MU)51.52
32
Machine UnlearningTOFU (forget 1% data)
UFE97.88
24
Machine UnlearningTOFU (forget 5% data)
UFE96.91
24
Machine UnlearningTOFU (forget 10% data)
UFE Score96.85
24
UnlearningTOFU Phi 1.5 1.0 (forget 10%)
UFE66.88
12
UnlearningTOFU Llama 2 1.0 (forget 1%)
UFE89.47
12
UnlearningTOFU Llama 2 1.0 (forget 5%)
UFE Score84.49
12
UnlearningTOFU Llama 2 1.0 (forget 10%)
UFE82.18
12
UnlearningTOFU Phi 1.5 1.0 (forget 1%)
UFE75.51
12
Model UtilityAlpaca-Llama 3 Utility Benchmarks
ARC-c Score47.27
7
Showing 10 of 11 rows

Other info

Follow for update