Research map / Label-efficient and robust learning / Recognition and long-tail
Driving like yourself: A Benchmark for Closed-Loop Personalized End-to-End Autonomous Driving
Xiaoru Dong, Ruiqin Li, Xiao Han, Zhenxuan Wu, Jiamin Wang, Jian Chen, Qi Jiang, SM Yiu, Xinge Zhu, Yuexin Ma
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add style-reward-model guidance instead of direct fine-tuning (ablation) | PDSA-WB vs DFT · Person2Drive average of routes | MMDSS: 0.6760 -> 0.6785 | Improved |
| add BEV features to the style reward model (ablation) | PDSA · Person2Drive average of routes | MMDSS: 0.6785 -> 0.6800 | Improved |
| add BEV features to the style reward model (ablation) | PDSA · StyleDrive aggressive (A to GT) | MMDSS: 0.8601 -> 0.8605 | Improved |
| fine-tune with PDSA instead of direct fine-tuning (ablation) | DiffusionDrive · StyleDrive aggressive (A to GT) | MMDSS: 0.8551 -> 0.8605 | Improved |
| aggressive fine-tuning shifts toward aggressive style (ablation) | PDSA · StyleDrive normal (N to GT) | MMDSS: 0.9439 -> 0.9308 | Hurt |
| personal trajectories instead of coarse ANC category models (ablation) | fine-tuning · Person2Drive | MMDSS average: 0.6174 -> 0.6800 | Improved |
| use PDSA instead of inverse RL (ablation) | style learning · Person2Drive | MMDSS: 0.6591 -> 0.7055 | Improved |
| use PDSA instead of behavior cloning (ablation) | style learning · Person2Drive | MMDSS: 0.6811 -> 0.7055 | Improved |
| use PDSA instead of direct style-vector supervision (ablation) | style learning · Person2Drive | MMDSS: 0.6690 -> 0.7055 | Improved |
| apply PDSA personalization | DiffusionDrive · Human preference Route 1 | preferred (%): 24.0 -> 76.0 | Improved |
| apply PDSA personalization | DiffusionDrive · Person2Drive | Driving Score: 58.22 -> 70.13 | Improved |
| apply PDSA personalization | DiffusionDrive · Person2Drive | Success Rate: 98.9% -> 98.4% | No clear effect |
Closest papers by meaning
- Bench2Drive: Towards Multi-Ability Benchmarking of Closed-Loop End-To-End Autonomous Driving (NeurIPS 2024 · similarity 0.89)
- Prioritizing Perception-Guided Self-Supervision: A New Paradigm for Causal Modeling in End-to-End Autonomous Driving (NeurIPS 2025 · similarity 0.87)
- Beyond Self-Play and Scale: A Behavior Benchmark for Generalization in Autonomous Driving (NeurIPS 2026 · similarity 0.87)
- DriveGPT: Scaling Autoregressive Behavior Models for Driving (ICML 2025 · similarity 0.87)
- Robust Autonomy Emerges from Self-Play (ICML 2025 · similarity 0.86)
- DriveDPO: Policy Learning via Safety DPO For End-to-End Autonomous Driving (NeurIPS 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
