Trajectory Generation on T-Drive
63.5JaccardTrajGPT-R
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| TrajGPT-Rtraining_mode=RMFT (Reward Modeling Fine-Tuning)2026.02 | 63.5 | 57 | 34.5 | 0.005 | 0.013 | 8.57 | 10.22 | |
| TrajGPTtraining_mode=Base2026.02 | 61.7 | 56.7 | 31.7 | 0.005 | 0.014 | 8.06 | 9.87 | |
| TrajGPT-DPOtraining_mode=DPO fine-tuning2026.02 | 61.2 | 57.3 | 33.3 | 0.005 | 0.011 | 8.49 | 10.12 | |
| IQL2026.02 | 31.4 | 37.4 | 19.5 | 0.005 | 0.01 | 8.4 | 9.81 | |
| D3PM2026.02 | 23.1 | 24.6 | 12.7 | 0.691 | 0.691 | 7.89 | 8.65 | |
| Markov2026.02 | 22.5 | 42.8 | 0 | 0.007 | 0.137 | 3.66 | 3.63 | |
| TrajVAE2026.02 | 20 | 40.5 | 5.4 | 0.054 | 0.144 | 8.27 | 9 | |
| TrajGAIL2026.02 | 17.2 | 27.4 | 0 | 0.166 | 0.328 | 8.06 | 9.06 |