Research map / 3D vision / 3D generation
RoboTron-Mani: All-in-One Multimodal Large Model for Robotic Manipulation
Feng Yan, Fanfan Liu, Yiyang Huang, Zechao Guan, Liming Zheng, Yufeng Zhong, Chengjian Feng, Lin Ma
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 frame-by-frame action (FFA) output (ablation) | RoboTron-Mani · CALVIN | average length: 1.77 -> 2.37 | Improved |
| add image output on top of FFA (ablation) | RoboTron-Mani · CALVIN | average length: 2.37 -> 3.13 | Improved |
| add UVFormer on top of FFA (ablation) | RoboTron-Mani · CALVIN | average length: 2.37 -> 2.88 | Improved |
| add UVFormer to FFA plus image output (ablation) | RoboTron-Mani · CALVIN | average length: 3.13 -> 3.21 | Improved |
| add occupancy (OCC) supervision on FFA plus UVFormer (ablation) | RoboTron-Mani · CALVIN | average length: 2.88 -> 3.18 | Improved |
| add OCC to FFA, image and UVFormer (ablation) | RoboTron-Mani · CALVIN | average length: 3.21 -> 3.31 | Improved |
| replace MLP action head with DiT action head (ablation) | RoboTron-Mani · CALVIN | average length: 3.31 -> 3.51 | Improved |
| train on aligned RoboData instead of non-aligned (ablation) | RoboTron-Mani · LIBERO | success rate: 64.2% -> 90.7% | Improved |
| train on aligned RoboData instead of non-aligned (ablation) | RoboTron-Mani · CALVIN | success rate: 74.7% -> 91.0% | Improved |
| train on aligned data instead of non-aligned (ablation) | RoboTron-Mani · Meta-World | success rate: 79.3% -> 78.6% | No clear effect |
| align data for a 2D-input model (ablation) | RoboFlamingo* · LIBERO | success rate: 69.1% -> 72.1% | No clear effect |
| use 3D perception model over 2D RoboFlamingo on real robot | RoboTron-Mani · Real-world Hard | success rate: 6.6% -> 46.6% | Improved |
Closest papers by meaning
- Robots Pre-train Robots: Manipulation-Centric Robotic Representation from Large-Scale Robot Datasets (ICLR 2025 · similarity 0.87)
- RDT-1B: a Diffusion Foundation Model for Bimanual Manipulation (ICLR 2025 · similarity 0.86)
- RoboBrain: A Unified Brain Model for Robotic Manipulation from Abstract to Concrete (CVPR 2025 · similarity 0.86)
- ManipLLM: Embodied Multimodal Large Language Model for Object-Centric Robotic Manipulation (CVPR 2024 · similarity 0.86)
- BridgeVLA: Input-Output Alignment for Efficient 3D Manipulation Learning with Vision-Language Models (NeurIPS 2025 · similarity 0.86)
- SAM2Act: Integrating Visual Foundation Model with A Memory Architecture for Robotic Manipulation (ICML 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.
