Next-state prediction on Nine benchmark datasets (average)
0.9492R2T3R
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| T3RBackbone=GAT, Adaptation steps=12026.06 | 0.9492 | 0.3095 | 0.9721 | 0.2196 | -0.9875 | |
| ERMBackbone=GAT, Adaptation steps=02026.06 | 0.9051 | 0.5023 | 0.9497 | 0.3916 | -2.36 | |
| TentBackbone=GAT, Adaptation steps=12026.06 | 0.8985 | 0.5135 | 0.9459 | 0.3915 | -2.6587 | |
| TTTBackbone=GAT, Adaptation steps=12026.06 | 0.8832 | 0.4806 | 0.9376 | 0.364 | -1.014 | |
| Joint TrainingBackbone=GAT, Adaptation steps=02026.06 | 0.8831 | 0.4782 | 0.9376 | 0.3621 | -1.0399 | |
| TTT + RotogradBackbone=GAT, Adaptation steps=12026.06 | 0.8541 | 0.5195 | 0.9136 | 0.4122 | -0.1343 |