Mobility Trajectory Generation on Oklahoma CA -> OK Zero-shot transfer
74.3AccuracyGRPO_Qwen
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GRPO_QwenTraining Strategy=GRPO, Backbone Model=Qwen, Multi-agent architecture=true2026.04 | 74.3 | 22.3 | 11.6 | |
| SFT_QwenTraining Strategy=SFT, Backbone Model=Qwen, Multi-agent architecture=true2026.04 | 71.4 | 24.1 | 12.7 | |
| SFT_CoPBTraining Strategy=SFT, Backbone Model=CoPB, Multi-agent architecture=true2026.04 | 71.1 | 32.5 | 11.4 | |
| GRPO_LLaMATraining Strategy=GRPO, Backbone Model=LLaMA, Multi-agent architecture=true2026.04 | 69 | 26.8 | 14.2 | |
| SFT_RAGHomeTraining Strategy=SFT, Backbone Model=RAGHome, Multi-agent architecture=true2026.04 | 67.2 | 24.6 | 12 | |
| SFT_LLaMATraining Strategy=SFT, Backbone Model=LLaMA, Multi-agent architecture=true2026.04 | 65 | 22.7 | 16.5 |