Trajectory Generation on Porto Dataset (test)
52.2JacTrajGPT-R
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| TrajGPT-Rstage=Fine-tuned, tuning_scheme=RMFT2026.02 | 52.2 | 47 | 43.2 | 0.013 | 0.032 | 10.13 | 10.75 | |
| TrajGPT-DPOstage=Fine-tuned, tuning_scheme=DPO2026.02 | 33.7 | 35.3 | 33.2 | 0.024 | 0.04 | 8.06 | 9.06 | |
| TrajGPTstage=Pre-trained2026.02 | 31.9 | 35 | 32.2 | 0.023 | 0.042 | 8.13 | 9.17 | |
| IQL2026.02 | 21.4 | 25.9 | 21.5 | 0.023 | 0.042 | 7.77 | 8.52 | |
| TrajGAIL2026.02 | 10.2 | 24.7 | 5 | 0.021 | 0.035 | 8.46 | 10.47 | |
| D3PM2026.02 | 8.3 | 33.3 | 25.9 | 0.111 | 0.225 | 6.81 | 6.82 | |
| Markov2026.02 | 8 | 33.2 | 9.9 | 0.077 | 0.209 | 6.71 | 6.71 | |
| TrajVAE2026.02 | 8 | 32 | 15.8 | 0.093 | 0.226 | 6.63 | 6.71 |