Mobility trajectory generation on Georgia CA -> GA Zero-shot transfer
73.6AccuracyGRPO_Qwen
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GRPO_QwenTraining Strategy=GRPO, Backbone Model=Qwen, Multi-agent architecture=true2026.04 | 73.6 | 22.9 | 11.9 | |
| SFT_QwenTraining Strategy=SFT, Backbone Model=Qwen, Multi-agent architecture=true2026.04 | 70.8 | 24.8 | 13 | |
| GRPO_LLaMATraining Strategy=GRPO, Backbone Model=LLaMA, Multi-agent architecture=true2026.04 | 68.1 | 27.6 | 14.7 | |
| SFT_RAGHomeTraining Strategy=SFT, Backbone Model=RAGHome, Multi-agent architecture=true2026.04 | 64.3 | 32.4 | 13.4 | |
| SFT_CoPBTraining Strategy=SFT, Backbone Model=CoPB, Multi-agent architecture=true2026.04 | 64.2 | 30.3 | 15.3 | |
| SFT_LLaMATraining Strategy=SFT, Backbone Model=LLaMA, Multi-agent architecture=true2026.04 | 63.2 | 25.3 | 19.5 |