Trajectory Generation on Toyota core area of Tokyo (test)
52.4Jaccard IndexTrajGPT-R
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| TrajGPT-Rfine-tuning=RMFT2026.02 | 52.4 | 57.5 | 38.3 | 0.016 | 0.042 | 14.85 | 14.82 | |
| TrajGPT-DPOfine-tuning=DPO2026.02 | 49.9 | 55.6 | 34.1 | 0.021 | 0.052 | 14.44 | 14.75 | |
| TrajGPT2026.02 | 39 | 45.5 | 22.5 | 0.028 | 0.07 | 14.07 | 14.41 | |
| IQL2026.02 | 23.6 | 27.1 | 7.2 | 0.026 | 0.075 | 14.17 | 14.28 | |
| Markov2026.02 | 19.8 | 29.1 | 0.8 | 0.34 | 0.692 | 13.4 | 13.75 | |
| TrajVAE2026.02 | 18.1 | 27.1 | 1.8 | 0.056 | 0.174 | 12.31 | 12.49 | |
| D3PM2026.02 | 17.3 | 22.5 | 2.1 | 0.692 | 0.691 | 12.02 | 12.24 | |
| TrajGAIL2026.02 | 12.4 | 20.6 | 0.1 | 0.072 | 0.234 | 12.22 | 12.61 |