Research map / 3D vision / Gaussian splatting
ManiGaussian: Dynamic Gaussian Splatting for Multi-task Robotic Manipulation
Guanxing Lu, Shiyi Zhang, Ziwei Wang, Changliu Liu, Jiwen Lu, Yansong Tang
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 regressor with geometry reconstruction loss (ablation) | PerceiverIO manipulation policy · RLBench 10 tasks | average success rate (%): 23.6 -> 39.2 | Improved |
| add semantic feature distillation from Stable Diffusion (ablation) | ManiGaussian (geometry only) · RLBench 10 tasks | average success rate (%): 39.2 -> 41.6 | Improved |
| add deformation predictor with future scene consistency loss (ablation) | ManiGaussian (geometry only) · RLBench 10 tasks | average success rate (%): 39.2 -> 43.6 | Improved |
| combine geometry, semantics and dynamics (ablation) | PerceiverIO policy · RLBench 10 tasks | average success rate (%): 23.6 -> 44.8 | Improved |
| add dynamics loss on top of geometry and semantics (ablation) | ManiGaussian · RLBench Planning tasks | success rate (%): 46.0 -> 40.0 | Hurt |
| add dynamics loss on top of geometry and semantics (ablation) | ManiGaussian · RLBench Long tasks | success rate (%): 8.0 -> 14.0 | Improved |
| increase semantic loss weight from 0.0001 to 0.001 (ablation) | ManiGaussian (no dynamics) · RLBench 10 tasks | average success rate (%): 41.6 -> 37.6 | Hurt |
| decrease semantic loss weight from 0.0001 to 0.00001 (ablation) | ManiGaussian (no dynamics) · RLBench 10 tasks | average success rate (%): 41.6 -> 35.2 | Hurt |
| use ManiGaussian instead of GNFactor | GNFactor · RLBench 10 tasks | average success rate (%): 31.7 -> 44.8 | Improved |
| use explicit Gaussian scene reconstruction instead of NeRF | GNFactor · RLBench learning curve | training speed: 2.29x faster training, 1.18x performance | Improved |
Later papers that compared with this method
3 later accepted papers report ManiGaussian as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- PhysMani: Physics-principled 3D World Model for Dynamic Object Manipulation (ECCV 2026 · better on 3 of 3 shared results)
- GeoMoLa: Geometry-Aware Motion Latents for Learning Robust Manipulation Policies (ICML 2026 · better on 1 of 1 shared result)
- Pre-training Auto-regressive Robotic Models with 4D Representations (ICML 2025 · 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
- Self-Correcting Robot Manipulation via Gaussian-Splatted Foresight (AAAI 2025 · similarity 0.91)
- Gaussian Splatting Lucas-Kanade (ICLR 2025 · similarity 0.88)
- Robots Pre-train Robots: Manipulation-Centric Robotic Representation from Large-Scale Robot Datasets (ICLR 2025 · similarity 0.87)
- Unleashing Large-Scale Video Generative Pre-training for Visual Robot Manipulation (ICLR 2024 · similarity 0.87)
- Exploring the Limits of Vision-Language-Action Manipulation in Cross-task Generalization (NeurIPS 2025 · similarity 0.87)
- GravMAD: Grounded Spatial Value Maps Guided Action Diffusion for Generalized 3D Manipulation (ICLR 2025 · 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.
