Research map / Point clouds / Point cloud classification
DriveX: Omni Scene Modeling for Learning Generalizable World Knowledge in Autonomous Driving
Chen Shi, Shaoshuai Shi, Kehua Sheng, Bo Zhang, Li Jiang
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 semantic supervision to depth supervision (ablation) | DriveX-S world encoder · FlowOcc3D occupancy and flow | mIoU: 3.96 -> 42.53 | Improved |
| add color reconstruction supervision (ablation) | DriveX-S world encoder · FlowOcc3D occupancy and flow | mIoU: 42.53 -> 43.47 | Improved |
| add future prediction (2 frames) in FSA (ablation) | DriveX-S with ViewFormer · FlowOcc3D occupancy and flow | mAVE: 0.415 -> 0.385 | Improved |
| extend prediction horizon from 2 to 4 frames (ablation) | DriveX-S with ViewFormer · FlowOcc3D occupancy and flow | mIoU: 43.47 -> 43.59 | No clear effect |
| use decoupled instead of joint training (ablation) | DriveX · nuScenes point cloud forecasting | avg Chamfer Distance (m^2): 1.41 -> 1.06 | Improved |
| add dynamic-aware ray sampling (ablation) | DriveX decoupled · nuScenes point cloud forecasting | avg Chamfer Distance (m^2): 1.06 -> 1.02 | Improved |
| use direct instead of autoregressive future prediction (ablation) | DriveX · nuScenes point cloud forecasting | avg Chamfer Distance (m^2): 1.32 -> 1.02 | Improved |
| scale training data from 50% to 100% (ablation) | DriveX · nuScenes point cloud forecasting | avg Chamfer Distance (m^2): 1.53 -> 1.02 | Improved |
| scale training data from 10% to 50% (ablation) | DriveX · nuScenes point cloud forecasting | avg Chamfer Distance (m^2): 2.30 -> 1.53 | Improved |
| add DriveX-S features to ViewFormer | ViewFormer · FlowOcc3D occupancy (no flow head) | mIoU: 43.61 -> 44.38 | Improved |
| add ViDAR pretraining to ViewFormer | ViewFormer · FlowOcc3D occupancy (no flow head) | mIoU: 43.61 -> 43.72 | No clear effect |
| add DriveX-S to DiffusionDrive-C | DiffusionDrive-C · NAVSIM test | PDMS: 83.6 -> 84.5 | Improved |
Closest papers by meaning
- DriveVLA-W0: World Models Amplify Data Scaling Law in Autonomous Driving (ICLR 2026 · similarity 0.90)
- World4Drive: End-to-End Autonomous Driving via Intention-aware Physical Latent World Model (ICCV 2025 · similarity 0.89)
- Enhancing End-to-End Autonomous Driving with Latent World Model (ICLR 2025 · similarity 0.88)
- X-Scene: Large-Scale Driving Scene Generation with High Fidelity and Flexible Controllability (NeurIPS 2025 · similarity 0.88)
- DriveSpatial: A Benchmark for Spatiotemporal Intelligence in VLMs for Autonomous Driving (NeurIPS 2026 · similarity 0.88)
- SGDrive: Scene-to-Goal Hierarchical World Cognition for Autonomous Driving (CVPR 2026 · similarity 0.88)
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.
