根据参数敏感度动态调整合并权重，提升大模型融合效果
Sens-Merging: Sensitivity-Guided Parameter Balancing for Merging Large Language Models
- 按任务内敏感度和跨任务迁移性动态分配参数权重
- 在Mistral 7B和LLaMA2上显著提升通用知识与代码生成能力
- 适合需要高效融合多个专业模型的开发者使用
近期大型语言模型涌现出大量任务专用微调版本，亟需高效的模型合并技术，在保留专长能力的同时避免昂贵的重新训练。现有基于任务向量的合并方法通常对所有参数采用统一系数，忽略了参数重要性的差异性，无论是在单个任务内还是跨任务之间。本文提出Sens-Merging，一种基于敏感度引导的系数调整方法，可在任务特定和跨任务两个层面优化现有模型合并技术。该方法分析各任务内的参数敏感度，并评估跨任务间的可迁移性，以确定最优合并系数。在Mistral 7B及LLaMA2-7B/13B模型上的大量实验表明，Sens-Merging显著提升了通用知识、数学推理和代码生成等任务的表现。尤其值得注意的是，当与现有合并技术结合时，合并模型在代码生成任务中甚至优于专门微调的模型。研究揭示了任务特异性与跨任务缩放之间的关键权衡，为未来模型合并策略提供重要启示。
原文摘要 · Abstract (English)
Recent advances in large language models have led to numerous task-specialized fine-tuned variants, creating a need for efficient model merging techniques that preserve specialized capabilities while avoiding costly retraining. While existing task vector-based merging methods show promise, they typically apply uniform coefficients across all parameters, overlooking varying parameter importance both within and across tasks. We present Sens-Merging, a sensitivity-guided coefficient adjustment method that enhances existing model merging techniques by operating at both task-specific and cross-task levels. Our method analyzes parameter sensitivity within individual tasks and evaluates cross-task transferability to determine optimal merging coefficients. Extensive experiments on Mistral 7B and LLaMA2-7B/13B models demonstrate that Sens-Merging significantly improves performance across general knowledge, mathematical reasoning, and code generation tasks. Notably, when combined with existing merging techniques, our method enables merged models to outperform specialized fine-tuned models, particularly in code generation tasks. Our findings reveal important trade-offs between task-specific and cross-task scalings, providing insights for future model merging strategies.
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。