Research map / Probabilistic and scientific modeling / Treatment effects
CausalVAD: De-confounding End-to-End Autonomous Driving via Causal Intervention
Jiacheng Tang, Zhiyuan Zhou, Zhuolin He, Jia Zhang, Kai Zhang, Jian Pu
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 Perception de-confounding module (PDM) (ablation) | VAD-tiny · nuScenes val | avg collision rate (%): 0.44 -> 0.26 | Improved |
| add Interaction de-confounding module (IDM) (ablation) | VAD-tiny · nuScenes val | avg L2 (m): 0.74 -> 0.57 | Improved |
| add both PDM and IDM (ablation) | VAD-tiny · nuScenes val | avg L2 (m): 0.74 -> 0.54 | Improved |
| use dictionary sizes (10,3,6) instead of smaller (5,2,3) (ablation) | CausalVAD · nuScenes val | avg L2 (m): 0.58 -> 0.54 | Improved |
| use larger dictionary (20,5,10) instead of (10,3,6) (ablation) | CausalVAD · nuScenes val | avg L2 (m): 0.54 -> 0.56 | Hurt |
| swap K-means++ for K-means or K-medoids clustering (ablation) | CausalVAD · nuScenes val | avg L2 (m): 0.54 vs 0.53 / 0.54 | No clear effect |
| apply causal de-confounding | VAD-tiny · nuScenes turning scenarios | avg L2 (m): 1.07 -> 0.69 | Improved |
| apply causal de-confounding | VAD-tiny · nuScenes ego-velocity zeroed | avg L2 (m): 6.94 -> 4.80 | Improved |
| apply causal de-confounding | VAD-tiny · nuScenes ego-velocity 100m/s | avg collision rate (%): 11.62 -> 5.19 | Improved |
| apply causal de-confounding | VAD-tiny · nuScenes ego-velocity x0.5 | avg L2 (m): 3.80 -> 2.84 | Improved |
| apply causal de-confounding | VAD-tiny · nuScenes val | avg collision rate (%): 0.44 -> 0.11 | Improved |
| apply causal de-confounding | VAD-tiny · NAVSIM | PDMS: 80.5 -> 87.2 | Improved |
Closest papers by meaning
- Prioritizing Perception-Guided Self-Supervision: A New Paradigm for Causal Modeling in End-to-End Autonomous Driving (NeurIPS 2025 · similarity 0.90)
- ChainFlow-VLA: Causal Flow Planning with Vision-Language Models (NeurIPS 2026 · similarity 0.88)
- Continuously Learning, Adapting, and Improving: A Dual-Process Approach to Autonomous Driving (NeurIPS 2024 · similarity 0.87)
- CaDeT: a Causal Disentanglement Approach for Robust Trajectory Prediction in Autonomous Driving (CVPR 2024 · similarity 0.87)
- nuReasoning: A Reasoning-Centric Dataset and Benchmark for Long-Tail Autonomous Driving (NeurIPS 2026 · similarity 0.87)
- DriveGPT: Scaling Autoregressive Behavior Models for Driving (ICML 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.
