Traveling Salesman Problem on Uniform-TSP1000
0Optimality GapConcorde
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ConcordeType=Exact2024.12 | 0 | 1.2 | |
| LKH3Type=Heuristic2024.12 | 0.2 | 0.4 | |
| HyperNS2025.10 | 0.3 | 53.4 | |
| DRHG2025.10 | 0.31 | 138.6 | |
| PCIsampling size=x8, steps=5, renoises=52026.06 | 0.31 | 16.5 | |
| Fast-T2Tsampling size=x4, steps=5, renoises=52026.06 | 0.36 | 25.5 | |
| IDEQType=SL+S+2-opt, sampling size=42024.12 | 0.49 | 6.1 | |
| Efficient Decoder Scaling StrategyDepth=42, Width=512, Decoding Strategy=beam search, Beam Size=22026.02 | 0.493 | 57 | |
| PCIsampling size=x2, steps=5, renoises=52026.06 | 0.53 | 4.3 | |
| Fast-T2Tsampling size=x1, steps=5, renoises=52026.06 | 0.57 | 7 | |
| Efficient Decoder Scaling StrategyDepth=42, Width=512, Decoding Strategy=greedy2026.02 | 0.576 | 28 | |
| IDEQType=SL+G+2-opt2024.12 | 0.63 | 2.3 | |
| PCIsampling size=x8, steps=1, renoises=12026.06 | 0.67 | 4 | |
| PCIsampling size=x1, steps=5, renoises=52026.06 | 0.7 | 2.4 | |
| Efficient Decoder Scaling StrategyDepth=42, Width=128, Decoding Strategy=beam search, Beam Size=22026.02 | 0.735 | 22 | |
| Fast-T2Tsampling size=x4, steps=1, renoises=12026.06 | 0.75 | 7.8 | |
| DualOpt2025.10 | 0.82 | 18.6 | |
| Efficient Decoder Scaling StrategyDepth=42, Width=128, Decoding Strategy=greedy2026.02 | 0.869 | 11 | |
| PCIsampling size=x2, steps=1, renoises=12026.06 | 0.94 | 1.1 | |
| Fast-T2Tsampling size=x1, steps=1, renoises=12026.06 | 1 | 2.1 | |
| Efficient Decoder Scaling StrategyDepth=6, Width=512, Decoding Strategy=beam search, Beam Size=22026.02 | 1.055 | 8.7 | |
| PCIsampling size=x1, steps=1, renoises=12026.06 | 1.13 | 0.69 | |
| UTSPType=UL + MCTS2024.12 | 1.2 | 6 | |
| T2TCOType=SL+S+2-opt, sampling size=42024.12 | 1.2 | 4.9 | |
| Efficient Decoder Scaling StrategyDepth=6, Width=512, Decoding Strategy=greedy2026.02 | 1.291 | 4.2 | |
| BQDecoding Strategy=beam search, Beam Size=162026.02 | 1.35 | 18 | |
| BQ-NCOType=RL +BS, beam search size=162024.12 | 1.4 | 38 | |
| DIFUSCOType=SL+S+2-opt, sampling size=42024.12 | 1.5 | 3.3 | |
| T2TCOType=SL+G+2-opt2024.12 | 1.5 | 3 | |
| SIGDDecoding Strategy=beam search, Beam Size=162026.02 | 1.569 | 39 | |
| ICAM aug.Type=RL, sampling size=82024.12 | 1.6 | 3.8 | |
| Efficient Decoder Scaling StrategyDepth=6, Width=128, Decoding Strategy=beam search, Beam Size=22026.02 | 1.819 | 3.8 | |
| LEHDDecoding Strategy=beam search, Beam Size=162026.02 | 1.874 | 22 | |
| DIFUSCOType=SL+G+2-opt2024.12 | 2 | 0.8 | |
| Efficient Decoder Scaling StrategyDepth=6, Width=128, Decoding Strategy=greedy2026.02 | 2.172 | 1.6 | |
| BQDecoding Strategy=greedy2026.02 | 2.274 | 1.18 | |
| BQ-NCOType=RL2024.12 | 2.3 | 2 | |
| SIGDDecoding Strategy=greedy2026.02 | 2.791 | 2.71 | |
| ICAMType=RL2024.12 | 2.9 | 2 | |
| GLOPType=RL2024.12 | 3.1 | 3.3 | |
| LEHDDecoding Strategy=greedy2026.02 | 3.168 | 1.17 | |
| INViT2025.10 | 5.99 | 15.6 | |
| 2-optType=Heuristic2024.12 | 13 | 13 | |
| POMOType=RL + S, sampling size=82024.12 | 41 | 8.5 |