{"ID":2834416,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2512.01461","arxiv_id":"2512.01461","title":"Stay Unique, Stay Efficient: Preserving Model Personality in Multi-Task Merging","abstract":"Model merging has emerged as a promising paradigm for enabling multi-task capabilities without additional training. However, existing methods often experience substantial performance degradation compared with individually fine-tuned models, even on similar tasks, underscoring the need to preserve task-specific information. This paper proposes Decomposition, Thresholding, and Scaling (DTS), an approximation-based personalized merging framework that preserves task-specific information with minimal storage overhead. DTS first applies singular value decomposition to the task-specific information and retains only a small subset of singular values and vectors. It then introduces a novel thresholding strategy that partitions singular vector elements into groups and assigns a scaling factor to each group. To enable generalization to unseen tasks, we further extend DTS with a variant that fuses task-specific information in a data-free manner based on the semantic similarity of task characteristics. Extensive experiments demonstrate that DTS consistently outperforms state-of-the-art baselines while requiring only 1\\% additional storage per task. Furthermore, experiments on unseen tasks show that the DTS variant achieves significantly better generalization performance. Our code is available at https://github.com/krumpguo/DTS.","short_abstract":"Model merging has emerged as a promising paradigm for enabling multi-task capabilities without additional training. However, existing methods often experience substantial performance degradation compared with individually fine-tuned models, even on similar tasks, underscoring the need to preserve task-specific informat...","url_abs":"https://arxiv.org/abs/2512.01461","url_pdf":"https://arxiv.org/pdf/2512.01461v1","authors":"[\"Kuangpu Guo\",\"Yuhe Ding\",\"Jian Liang\",\"Zilei Wang\",\"Ran He\"]","published":"2025-12-01T09:47:17Z","proceeding":"cs.LG","tasks":"[\"cs.LG\",\"cs.CV\"]","methods":"[]","has_code":false,"code_links":[{"ID":606413,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_id":2834416,"paper_url":"https://arxiv.org/abs/2512.01461","paper_title":"Stay Unique, Stay Efficient: Preserving Model Personality in Multi-Task Merging","repo_url":"https://github.com/krumpguo/DTS","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}