通过随机线性插值提升模型合并效果，自动寻找最优贡献比例。
Mixup Model Merge: Enhancing Model Merging Performance through Randomized Linear Interpolation
- 在参数空间中用Beta分布随机插值两个大模型，探索不同贡献比。
- 性能显著优于等权重合并，在多任务、抗干扰上均更优。
- 适合想提升模型合并效果的研究者与工程师，可与剪枝方法结合使用。
模型合并旨在将多个任务专用模型整合为一个统一模型，无需额外训练即可继承各模型能力。现有方法通常忽略不同模型对最终合并结果的贡献差异。本文提出 Mixup Model Merge（M3），受 Mixup 数据增强中随机线性插值启发，对两个任务专用的大语言模型（LLM）在参数空间进行随机线性插值，插值系数从 Beta 分布采样，以探索多样化的贡献比例。这种可控随机性使 M3 能发现更优的组合，超越标准等权重合并。大量实验表明，M3 显著（1）提升合并后 LLM 在多任务上的表现，（2）增强分布外和对抗鲁棒性，（3）优于剪枝方法 DARE 的正向作用，且可进一步与 DARE 结合实现更优效果，（4）通过调节 Beta 分布形状参数，平衡探索效率与贡献比例多样性。代码见附录。
原文摘要 · Abstract (English)
Model merging aims to integrate multiple task-specific models into a unified model that inherits the capabilities of the task-specific models, without additional training. Existing model merging methods often lack consideration of the varying contribution ratios of different task-specific models to the final merged model. In this paper, we propose Mixup Model Merge (M3), a simple yet effective method inspired by the randomized linear interpolation strategy from the Mixup data augmentation technique. M3 performs randomized linear interpolation in parameter space between two task-specific LLMs, where interpolation coefficients are sampled from a Beta distribution to explore diverse contribution ratios. This controllable randomness allows M3 to outperform standard equal-ratio merging by discovering better contribution ratio combinations. Extensive experiments show that M3 significantly (1) improves merged LLM performance across tasks, (2) enhances out-of-distribution and adversarial robustness, (3) outperforms the positive effects of the sparsification method DARE on model merging and can be further combined with DARE to achieve superior results, and (4) balances exploration efficiency and diversity in contribution ratios by tuning the Beta distribution's shape parameters. The code is provided in the supplementary materials.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。