---
license: mit
pipeline_tag: graph-ml
---
# GNN-based Multigraph Solver - GMS

Model weights and training/test files for the ICLR2026 paper [Beyond Simple Graphs: Neural Multi-Objective Routing on Multigraphs](https://openreview.net/forum?id=55laGcPNZZ).

In short, we extend end-to-end neural vehicle routing to problems defined on multigraphs. 

Code and usage instructions can be found in the project [Github repository](https://github.com/filiprydin/GMS). 

## Reference

If you find our work useful, please consider citing our paper:
```
@inproceedings{rydin2026beyond,
  title={Beyond Simple Graphs: Neural Multi-Objective Routing on Multigraphs},
  author={Filip Rydin and Attila Lischka and Jiaming Wu and Morteza Haghir Chehreghani and Balázs Kulcsár},
  booktitle={International Conference on Learning Representations},
  year={2026},
  url={https://openreview.net/forum?id=55laGcPNZZ}
}
```