Research map / 3D vision / 3D generation
DLWM: Dual Latent World Models enable Holistic Gaussian-centric Pre-training in Autonomous Driving
Yiyao Zhu, Ying Xue, Haiming Zhang, Guangfeng Jiang, Wending Zhou, Xu Yan, Jiantao Gao, Yingjie Cai, Bingbing Liu, Zhen Li, Shaojie Shen
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 two-stage pre-training | DLWM · 3D occupancy perception | mIoU: 20.83 -> 21.85 | Improved |
| add two-stage pre-training | DLWM · 4D occupancy forecasting | mIoU: 15.09 -> 17.77 | Improved |
| add two-stage pre-training | DLWM · motion planning | L2: 0.55 -> 0.46 | Improved |
| add dense pseudo-depth supervision (ablation) | DLWM · perception ablation | mIoU: 18.40 -> 18.74 | Improved |
| add semantic supervision (ablation) | DLWM · perception ablation | mIoU: 18.74 -> 18.99 | Improved |
| add Gaussian-flow-guided latent world model (stage 2) (ablation) | DLWM · perception ablation | mIoU: 18.99 -> 19.30 | Improved |
| use latent BEV instead of raw Gaussian features for planning (ablation) | DLWM · motion planning | L2: 0.55 -> 0.52 | Improved |
| add future BEV forecasting to planning (ablation) | DLWM · motion planning | L2: 0.52 -> 0.50 | Improved |
| add stage-1 reconstruction pre-training to planning (ablation) | DLWM · motion planning | L2: 0.50 -> 0.46 | Improved |
| use dual decoupled world models instead of a unified one (ablation) | DLWM · planning | L2: 0.58 -> 0.46 | Improved |
| raise Gaussian count from 25600 to 51200 (ablation) | DLWM · 3D occupancy perception | IoU: 30.56 -> 29.79 | Mixed |
| predict 1 future frame instead of 0 (ablation) | DLWM · 3D occupancy perception | mIoU: 18.83 -> 19.30 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 6 of 6 on most shared results.
- GaussianFormer-2 — GaussianFormer-2: Probabilistic Gaussian Superposition for Efficient 3D Occupancy Prediction (CVPR 2025) · better on 2 of 2 shared results
- GaussianWorld — GaussianWorld: Gaussian World Model for Streaming 3D Occupancy Prediction (CVPR 2025) · better on 2 of 2 shared results
- GaussianFormer — GaussianFormer: Scene as Gaussians for Vision-Based 3D Semantic Occupancy Prediction (ECCV 2024) · better on 1 of 1 shared result
- OccFormer — OccFormer: Dual-path Transformer for Vision-based 3D Semantic Occupancy Prediction (ICCV 2023) · better on 1 of 1 shared result
- SurroundOcc — SurroundOcc: Multi-camera 3D Occupancy Prediction for Autonomous Driving (ICCV 2023) · better on 1 of 1 shared result
- MonoScene — MonoScene: Monocular 3D Semantic Scene Completion (CVPR 2022) · better on 1 of 1 shared result
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
- GaussianWorld: Gaussian World Model for Streaming 3D Occupancy Prediction (CVPR 2025 · similarity 0.89)
- Semi-Supervised Vision-Centric 3D Occupancy World Model for Autonomous Driving (ICLR 2025 · similarity 0.88)
- Enhancing End-to-End Autonomous Driving with Latent World Model (ICLR 2025 · similarity 0.88)
- Visual Point Cloud Forecasting enables Scalable Autonomous Driving (CVPR 2024 · similarity 0.88)
- ODG: Occupancy Prediction Using Dual Gaussians (NeurIPS 2025 · similarity 0.88)
- Achieving Speed-Accuracy Balance in Vision-based 3D Occupancy Prediction via Geometric-Semantic Disentanglement (AAAI 2025 · 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.
