Research map / Label-efficient and robust learning / Knowledge distillation
Transferring Causal Driving Patterns for Generalizable Traffic Simulation with Diffusion-Based Distillation
Yuhang Chen, Jie Sun, Jialin Fan, Jian Sun
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 feature self-distillation (FSD) (ablation) | Teacher (VBD) · WOMD closed-loop | collision rate (%): 10.378 -> 8.311 | Improved |
| add feature self-distillation (FSD) (ablation) | Teacher (VBD) · INTERACTION closed-loop | collision rate (%): 12.441 -> 10.207 | Improved |
| add response self-distillation (RSD) (ablation) | Teacher (VBD) · WOMD closed-loop | collision rate (%): 10.378 -> 12.118 | Hurt |
| add response self-distillation (RSD) (ablation) | Teacher (VBD) · INTERACTION closed-loop | wrong-way rate (%): 7.312 -> 6.112 | Improved |
| add response self-distillation (RSD) (ablation) | Teacher (VBD) · WOMD closed-loop | kinematic infeasibility (%): 0.230 -> 0.016 | Improved |
| add contrastive self-distillation (CSD) (ablation) | Teacher (VBD) · WOMD closed-loop | collision rate (%): 10.378 -> 9.475 | Improved |
| add contrastive self-distillation (CSD) (ablation) | Teacher (VBD) · INTERACTION closed-loop | kinematic infeasibility (%): 0.341 -> 0.141 | Improved |
| add feature self-distillation (FSD) (ablation) | Teacher (VBD) · WOMD closed-loop | kinematic infeasibility (%): 0.230 -> 0.329 | Hurt |
| add all distillation components and continual distillation (full CDPT) (ablation) | Teacher (VBD) · INTERACTION closed-loop | collision rate (%): 12.441 -> 6.853 | Improved |
| add all distillation components and continual distillation (full CDPT) (ablation) | Teacher (VBD) · INTERACTION closed-loop | off-road rate (%): 7.125 -> 3.117 | Improved |
| add continual distillation on top of FSD (ablation) | FSD variant · INTERACTION closed-loop | collision rate (%): 10.207 -> 6.853 | Improved |
| CDPT distillation instead of plain diffusion teacher | VBD · WOMD val open-loop | minADE: 1.199 -> 0.831 | Improved |
Closest papers by meaning
- Causal Composition Diffusion Model for Closed-loop Traffic Generation (CVPR 2025 · similarity 0.91)
- BehaviorGPT: Smart Agent Simulation for Autonomous Driving with Next-Patch Prediction (NeurIPS 2024 · similarity 0.88)
- ChainFlow-VLA: Causal Flow Planning with Vision-Language Models (NeurIPS 2026 · similarity 0.88)
- Prioritizing Perception-Guided Self-Supervision: A New Paradigm for Causal Modeling in End-to-End Autonomous Driving (NeurIPS 2025 · similarity 0.88)
- Scenario Dreamer: Vectorized Latent Diffusion for Generating Driving Simulation Environments (CVPR 2025 · similarity 0.87)
- DriveGPT: Scaling Autoregressive Behavior Models for Driving (ICML 2025 · similarity 0.87)
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.
