GFFMERGE让图神经网络力场模型高效合并，无需重训即可适配新分子体系。
GFFMERGE: Efficient Merging of Graph Neural Force Fields and Beyond
- 基于消息传递层的线性结构，提出闭式合并方法，可直接计算最优融合参数。
- 在多个数据集上实现5-27倍加速，性能接近联合训练的黄金标准。
- 适合需快速部署、模块化组合专用模型的研究者，尤其适合跨体系力场迁移。
图神经网络（GNN）已革新原子模拟中的神经力场，以较低成本实现接近量子精度的结果，但将这些模型适配到新化学体系仍需昂贵的再训练。受视觉与语言模型合并的启发，我们提出GFFMERGE，首个针对GNN的原理性闭式模型合并框架。利用消息传递层的线性结构，将合并问题建模为凸嵌入对齐问题，并获得解析解。通过首次系统性评估GNN模型合并，我们发现专为视觉与语言设计的方法在力场回归任务中灾难性失败，而GFFMERGE则恢复接近联合训练的性能。在分子（MD17、MD22）、固态（LiPS20）及大规模图基准上，GFFMERGE与通用版GNNMERGE实现5-27倍加速，支持专用模型的模块化组合。值得注意的是，仅闭式解本身即优于所有基线方法，且为快速、数据高效收敛提供优异初始化。
原文摘要 · Abstract (English)
Graph Neural Networks (GNNs) have revolutionized Neural Force Fields for atomistic simulations, achieving near-quantum accuracy at reduced cost, yet adapting these models to new chemical systems requires expensive retraining of foundation models. Inspired by model merging in vision and language processing, we introduce GFFMERGE, the first principled framework for closed-form model merging in GNNs. We exploit the linear structure of message-passing layers and formulate merging as a convex embedding-alignment problem with an analytical solution. Through the first systematic benchmarking of model merging for GNNs, we show that existing methods designed for vision and language catastrophically fail on force field regression, while GFFMERGE recovers performance approaching gold standard joint training. Across molecular (MD17, MD22), solid-state (LiPS20), and large-scale graph benchmarks, GFFMERGE and GNNMERGE (its generic GNN counterpart) achieve 5-27$\times$ speedups while enabling modular composition of specialized models. Remarkably, our closed-form solution alone outperforms all baseline methods before fine-tuning and provides superior initialization for faster, data-efficient convergence.
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。