Research map / Label-efficient and robust learning / Recognition and long-tail
World4Drive: End-to-End Autonomous Driving via Intention-aware Physical Latent World Model
Yupeng Zheng, Pengxuan Yang, Zebin Xing, Qichao Zhang, Yuhang Zheng, Yinfeng Gao, Pengfei Li, Teng Zhang, Zhongpu Xia, Peng Jia, XianPeng Lang, Dongbin Zhao
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 driving intentions to latent world model (ablation) | LAW-like WM · nuScenes open-loop | avg L2 / collision: 0.61 -> 0.55 L2; 0.30 -> 0.25 collision | Improved |
| add depth spatial prior (ablation) | WM + intentions · nuScenes open-loop | avg L2 / collision: 0.55 -> 0.51 L2; 0.25 -> 0.29 collision | Mixed |
| add semantic prior (ablation) | WM + intentions + depth · nuScenes open-loop | avg L2 / collision: 0.29 -> 0.16 collision | Improved |
| add depth and semantic priors without intentions (ablation) | WM only · nuScenes open-loop | avg L2 / collision: 0.61 -> 0.49 L2; 0.30 -> 0.26 collision | Improved |
| add intentions on top of priors and WM (ablation) | depth+semantic WM · nuScenes open-loop | avg L2 / collision: 0.49 -> 0.50 L2; 0.26 -> 0.16 collision | Mixed |
| add world model to intentions with priors (ablation) | depth+semantic+intentions · nuScenes open-loop | avg L2 / collision: 0.61 -> 0.50 L2; 0.36 -> 0.16 collision | Improved |
| scale backbone ResNet-34 to ResNet-50 (ablation) | World4Drive · nuScenes open-loop | avg L2 / collision: 0.52 -> 0.50 L2; 0.25 -> 0.16 collision | Improved |
| scale backbone ResNet-50 to ResNet-101 (ablation) | World4Drive · nuScenes open-loop | avg L2 / collision: 0.50 -> 0.47 L2; 0.16 -> 0.14 collision | Improved |
| increase hidden dimension 128 to 256 (ablation) | World4Drive · nuScenes open-loop | avg L2 / collision: 0.55 -> 0.50 L2; 0.27 -> 0.16 collision | Improved |
| increase hidden dimension 256 to 384 (ablation) | World4Drive · nuScenes open-loop | avg L2 / collision: 0.50 -> 0.49 L2; 0.16 -> 0.10 collision | Improved |
| World4Drive vs LAW at night | World4Drive · nuScenes Night | avg L2 / collision: 0.67 -> 0.76 L2; 0.22 -> 0.08 collision | Mixed |
| World4Drive vs LAW in rain | World4Drive · nuScenes Rainy | avg L2 / collision: 0.54 -> 0.49 L2; 0.16 -> 0.05 collision | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PPAD — PPAD: Iterative Interactions of Prediction and Planning for End-to-end Autonomous Driving (ECCV 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Enhancing End-to-End Autonomous Driving with Latent World Model (ICLR 2025 · similarity 0.92)
- Driving in the Occupancy World: Vision-Centric 4D Occupancy Forecasting and Planning via World Models for Autonomous Driving (AAAI 2025 · similarity 0.92)
- Future-Aware End-to-End Driving: Bidirectional Modeling of Trajectory Planning and Scene Evolution (NeurIPS 2025 · similarity 0.91)
- S4-Driver: Scalable Self-Supervised Driving Multimodal Large Language Model with Spatio-Temporal Visual Representation (CVPR 2025 · similarity 0.90)
- Prioritizing Perception-Guided Self-Supervision: A New Paradigm for Causal Modeling in End-to-End Autonomous Driving (NeurIPS 2025 · similarity 0.90)
- $AutoDrive\text{-}P^3$: Unified Chain of Perception–Prediction–Planning Thought via Reinforcement Fine-Tuning (ICLR 2026 · similarity 0.90)
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.
