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

• URL: http://arxiv.org/abs/2605.21539v1
• Date: Wed, 20 May 2026 07:45:08 GMT
• Title: DualOptim+: Bridging Shared and Decoupled Optimizer States for Better Machine Unlearning in Large Language Models
• Abstract summary: We propose DualOptim+, a novel optimization framework for improving machine unlearning in large language models.<n>It introduces a base state to capture common representations shared by forgetting and retaining objectives and delta states to preserve objective-specific residuals.<n>We further introduce DualOptim+ 8bit, a quantized variant that reduces memory overhead without compromising performance.
• Score: 28.795194964735412
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: 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.

Related papers

• MOSCOPT: Mixture-of-Skills Collective Optimization for LLM Agents [0.9883261192383612]
  MOSCOPT is a text-native, parameter-free algorithm that jointly optimize a pool of $N$ skills and a gating skill $G$ that dynamically selects $K$ skills per step.<n>Experiments and detailed ablations across 5 benchmarks and 3 target LLMs demonstrate that MOSCOPT consistently outperforms all baselines.
  arXiv  Detail & Related papers  (2026-09-13T09:57:59Z)
• From One-to-One to Many-to-Many: Dynamic Cross-Layer Injection for Deep Vision-Language Fusion [91.35078719566472]
  Vision-Language Models (VLMs) create a severe visual feature bottleneck by using a crude, asymmetric connection.<n>We introduce Cross-Layer Injection (CLI), a novel and lightweight framework that forges a dynamic many-to-many bridge between the two modalities.
  arXiv  Detail & Related papers  (2026-01-15T18:59:10Z)
• Prompt and Parameter Co-Optimization for Large Language Models [35.72638351230096]
  We introduce MetaTuner, a novel framework that jointly integrates prompt optimization and fine-tuning for Large Language Models (LLMs) training.<n>Our framework is optimized to discover the optimal combinations between the prompts and parameters.<n>Experiments across diverse benchmarks show that our method consistently outperforms the baselines.
  arXiv  Detail & Related papers  (2025-09-29T03:38:25Z)
• Objective Soups: Multilingual Multi-Task Modeling for Speech Processing [69.52720282028385]
  Training a single model for multilingual, multi-task speech processing (MSP) is severely hampered by conflicting objectives between tasks.<n>This paper investigates three multi-objective MSP formulations, which we refer to as textbfobjective soup recipes.<n>Our work demonstrates that hierarchical MOO is a more effective and scalable approach for building state-of-the-art MSP models.
  arXiv  Detail & Related papers  (2025-08-12T07:01:09Z)
• DualOptim: Enhancing Efficacy and Stability in Machine Unlearning with Dual Optimizers [7.365841971996289]
  We propose DualOptim, which incorporates adaptive learning rate and decoupled momentum factors.<n>We show that DualOptim can significantly boost efficacy and stability across diverse tasks, including image classification, image generation, and large language models.
  arXiv  Detail & Related papers  (2025-04-22T12:18:26Z)
• Multi-Objective Deep Reinforcement Learning for Optimisation in Autonomous Systems [3.2826250607043796]
  Multi-Objective Reinforcement Learning (MORL) techniques exist but they have mostly been applied in RL benchmarks rather than real-world AS systems. In this work, we use a MORL technique called Deep W-Learning (DWN) to find the optimal configuration for runtime performance optimization. We compare DWN to two single-objective optimization implementations: epsilon-greedy algorithm and Deep Q-Networks.
  arXiv  Detail & Related papers  (2024-08-02T11:16:09Z)
• Decoding-Time Language Model Alignment with Multiple Objectives [116.42095026960598]
  Existing methods primarily focus on optimizing LMs for a single reward function, limiting their adaptability to varied objectives. Here, we propose $textbfmulti-objective decoding (MOD)$, a decoding-time algorithm that outputs the next token from a linear combination of predictions. We show why existing approaches can be sub-optimal even in natural settings and obtain optimality guarantees for our method.
  arXiv  Detail & Related papers  (2024-06-27T02:46:30Z)
• Towards Efficient Pareto Set Approximation via Mixture of Experts Based Model Fusion [53.33473557562837]
  Solving multi-objective optimization problems for large deep neural networks is a challenging task due to the complexity of the loss landscape and the expensive computational cost. We propose a practical and scalable approach to solve this problem via mixture of experts (MoE) based model fusion. By ensembling the weights of specialized single-task models, the MoE module can effectively capture the trade-offs between multiple objectives.
  arXiv  Detail & Related papers  (2024-06-14T07:16:18Z)
• Meta Mirror Descent: Optimiser Learning for Fast Convergence [85.98034682899855]
  We take a different perspective starting from mirror descent rather than gradient descent, and meta-learning the corresponding Bregman divergence. Within this paradigm, we formalise a novel meta-learning objective of minimising the regret bound of learning. Unlike many meta-learned optimisers, it also supports convergence and generalisation guarantees and uniquely does so without requiring validation data.
  arXiv  Detail & Related papers  (2022-03-05T11:41:13Z)
• Optimization-Inspired Learning with Architecture Augmentations and Control Mechanisms for Low-Level Vision [74.9260745577362]
  This paper proposes a unified optimization-inspired learning framework to aggregate Generative, Discriminative, and Corrective (GDC) principles. We construct three propagative modules to effectively solve the optimization models with flexible combinations. Experiments across varied low-level vision tasks validate the efficacy and adaptability of GDC.
  arXiv  Detail & Related papers  (2020-12-10T03:24:53Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.