Traveling Salesperson Problem on TSP 80 nodes
3.9Gap to Optimal (%)GNARLBC
Evaluation Results
| Method | Links | |
|---|---|---|
| GNARLBCtraining=Behavioral Cloning2025.09 | 3.9 | |
| GNARLPPOtraining=PPO2025.09 | 7.1 | |
| GNARLPPO+heurtraining=PPO, insertion_heuristic=true2025.09 | 9.8 | |
| Christofidestype=handcrafted heuristic2025.09 | 11.3 | |
| Georgiev et al.beam_search_width=12802025.09 | 19 |