Research map / Vision-language and multimodal / Robot action and navigation
GWM: Towards Scalable Gaussian World Models for Robotic Manipulation
Guanxing Lu, Baoxiong Jia, Puhao Li, Yixin Chen, Ziwei Wang, Yansong Tang, Siyuan Huang
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 Gaussian splatting representation to the image-based diffusion world model (ablation) | diffusion-transformer world model · RoboCasa PnP CabToCounter (H-50) | success rate (%): 4 -> 18 | Improved |
| add 3D Gaussian VAE (ablation) | GWM with Gaussian splatting · RoboCasa PnP CabToCounter (H-50) | success rate (%): 18 -> 24 | Improved |
| add 3D Gaussian VAE (ablation) | GWM with Gaussian splatting · RoboCasa PnP CabToCounter | FVD: 65.3 -> 62.4 | Improved |
| add Gaussian splatting (ablation) | image-based world model · RoboCasa PnP CabToCounter | PSNR: 27.2 -> 26.9 | Hurt |
| use GWM instead of iVideoGPT | world model · Meta-World future prediction | FVD: 75.0 -> 73.0 | Improved |
| use GWM instead of iVideoGPT | world model · Franka-PnP future prediction | FVD: 63.2 -> 61.5 | Improved |
| use GWM features as state encoding for BC-transformer | BC-transformer · RoboCasa 24 tasks (H-50) | success rate (%): average improvement of 10.5% | Improved |
| use GWM features as state encoding for BC-transformer | BC-transformer · RoboCasa 24 tasks (G-3000) | success rate (%): average gain of 7.6% | Improved |
| use GWM encoder on PnP CounterToCab | BC-transformer · RoboCasa PnP CounterToCab | success rate (%): 28 -> 22 | Hurt |
| use GWM instead of iVideoGPT as MBRL simulator | DrQ-v2 with MBPO · Meta-World six tasks | convergence speed: approximately 2x faster | Improved |
| add GWM representation to Diffusion Policy | Diffusion Policy · Franka-PnP real world | success rate: 35% -> 65% | Improved |
Closest papers by meaning
- Predictive Inverse Dynamics Models are Scalable Learners for Robotic Manipulation (ICLR 2025 · similarity 0.89)
- Zero-Shot Robotic Manipulation with Pre-Trained Image-Editing Diffusion Models (ICLR 2024 · similarity 0.89)
- Dream to Manipulate: Compositional World Models Empowering Robot Imitation Learning with Imagination (ICLR 2025 · similarity 0.89)
- GravMAD: Grounded Spatial Value Maps Guided Action Diffusion for Generalized 3D Manipulation (ICLR 2025 · similarity 0.88)
- EA-WM: Event-Aware Generative World Model with Structured Kinematic-to-Visual Action Fields (NeurIPS 2026 · similarity 0.88)
- EnerVerse: Envisioning Embodied Future Space for Robotics Manipulation (NeurIPS 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.
