训练时长影响大模型合并效果，过拟合反而能提升合并性能。
Are we Merging the Right Models? Impact of Expert Training Duration on Model Merging for LLMs
- 不同训练时长的专家模型合并，效果差异显著
- 稀疏化方法在过拟合后表现更优，简单平均则相反
- 合并策略需与训练时长协同设计，非独立选择
多任务模型合并将各自训练的专家模型整合为一个可处理所有任务的统一模型，无需联合训练。当前通常在专家模型达到最优验证损失时进行合并。本文系统研究专家模型训练时长对合并质量的影响。我们在五个领域（数学、代码、指令遵循、多语言、安全）上，针对Qwen 3.5 0.8B、2B和4B三个规模的模型，从25%到500%的最优训练步数保存检查点，并在每个训练时长下评估五种合并方法。结果发现：简单平均法随过拟合急剧下降，而基于稀疏化的合并方法在超过验证最优后仍持续提升。通过偏差-方差分解分析，我们揭示其与随机森林的类比关系——高方差个体模型更适合平均。结论表明，训练时长与合并方法应联合选择，而非独立决定。
原文摘要 · Abstract (English)
Multi-task model merging combines separately trained expert models into a single model that handles all tasks without co-training. Standard practice merges experts at their optimal validation loss. We challenge this convention by systematically studying how training duration of domain experts affects the quality of the merged model. We fine-tune experts on five domains (Math, Code, Instruction Following, Multilingual, and Safety) across three model sizes (Qwen 3.5 0.8B, 2B, and 4B), saving checkpoints from 25% to 500% of the optimal training steps and evaluating five merging methods at each duration. Our findings reveal a striking method-dependent pattern: simple averaging degrades sharply with overfitting, while sparsification-based methods achieve their best performance well past the validation optimum. We formalize this through bias-variance decomposition analysis, drawing a parallel to random forests where averaging benefits from high-variance individual learners. These results suggest that training duration and merging method should be chosen jointly rather than independently.
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。