Research map / 3D vision / 3D generation
GEM: Generating LiDAR World Model via Deformable Mamba
Yang Wu, Zhaojiang Liu, Qiang Meng, Youquan Liu, Renliang Weng, Jianjun Qian, Jian Yang, Jin Xie
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 dynamic extractor and dynamic deformable Mamba (ablation) | GEM · nuScenes 3s | CD: 1.01 -> 0.67 | Improved |
| add static extractor and static deformable Mamba (ablation) | GEM · nuScenes 3s | CD: 0.78 -> 0.67 | Improved |
| guide deformable Mamba paths with dynamic-static features (ablation) | GEM · nuScenes 3s | CD: 1.17 -> 0.67 | Improved |
| add adaptive gated attention fusion (ablation) | GEM · nuScenes 3s | CD: 0.88 -> 0.67 | Improved |
| use tri-path deformable Mamba instead of triple standard Mamba (ablation) | GEM · nuScenes 3s | CD: 0.72 -> 0.67 | Improved |
| use tri-path deformable Mamba instead of UNet (ablation) | GEM · nuScenes 3s | CD: 1.07 -> 0.67 | Improved |
| use tri-path deformable Mamba instead of DiT (ablation) | GEM · nuScenes 3s | AbsRel: 10.67 -> 11.13 | Mixed |
| use Mamba tokenizer instead of transformer with discriminator (ablation) | LiDAR scene tokenizer · KITTI-360 reconstruction | FSVD: 7.9 -> 6.9 | Improved |
| add a discriminator to the Mamba tokenizer (ablation) | LiDAR scene tokenizer · KITTI-360 reconstruction | FSVD: 6.9 -> 6.7 | Improved |
| GEM versus Copilot4D at 1s | GEM · nuScenes 1s | CD: 2.01 -> 0.38 | Improved |
| GEM versus Copilot4D at 3s | GEM · nuScenes 3s | AbsRel: 10.38 -> 11.13 | Hurt |
| GEM versus 4D-Occ for generation speed | GEM · nuScenes | FPS: 7.41 -> 9.23 | Improved |
Closest papers by meaning
- Towards foundational LiDAR world models with efficient latent flow matching (NeurIPS 2025 · similarity 0.88)
- WeatherGen: A Unified Diverse Weather Generator for LiDAR Point Clouds via Spider Mamba Diffusion (CVPR 2025 · similarity 0.87)
- OLiDM: Object-aware LiDAR Diffusion Models for Autonomous Driving (AAAI 2025 · similarity 0.86)
- DIO: Decomposable Implicit 4D Occupancy-Flow World Model (CVPR 2025 · similarity 0.86)
- LiDAR-RT: Gaussian-based Ray Tracing for Dynamic LiDAR Re-simulation (CVPR 2025 · similarity 0.86)
- Outdoor Scene Extrapolation with Hierarchical Generative Cellular Automata (CVPR 2024 · 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.
