ClementiGroup/range: [STABLE] rangemp v1.0.0

[STABLE] rangemp v1.0.0 Release Date: 02 Feb 2026 License: MIT Paper: arXiv:2502.13797 Overview Initial public release of RANGE (Relaying Attention Nodes for Global Encoding), a model-agnostic framework for extending Graph Neural Networks with long-range message-passing capabilities. Features Virtual node architecture with attention-based aggregation-broadcast mechanism that significantly reduces oversquashing effects in message-passing neural networks. Linear time complexity O(N) with respect to system size. Model-agnostic design. Installation Requires the mlcg package (v0.1.0+) installed in a clean environment. git clone git@github.com:ClementiGroup/range.git cd range pip install . Dependencies Python 3.12 mlcg >= 0.1.0 PyTorch, PyTorch Geometric (via mlcg) Citation @article{caruso2025extending, title={Extending the RANGE of Graph Neural Networks: Relaying Attention Nodes for Global Encoding}, author={Caruso, Alessandro and Venturin, Jacopo and Giambagli, Lorenzo and Rolando, Edoardo and No{\'e}, Frank and Clementi, Cecilia}, journal={arXiv preprint arXiv:2502.13797}, year={2025} } Authors Alessandro Caruso, Jacopo Venturin, Lorenzo Giambagli, Edoardo Rolando, Zakariya El-Machachi, Frank Noé, Cecilia Clementi Freie Universität Berlin • Rice University • AI4Science, Microsoft Research * These authors contributed equally.