Research map / Vision-language and multimodal / Robot action and navigation
RoBridge: A Hierarchical Architecture Bridging Cognition and Execution for General Robotic Manipulation
Kaidong Zhang, Rongtao Xu, Pengzhen Ren, Junfan Lin, Hefeng Wu, Liang Lin, Xiaodan Liang
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 |
|---|---|---|---|
| use first-view masked depth instead of original depth (ablation) | RoBridge · Metaworld ablation | mean success rate (%): 79.74 -> 82.12 | Improved |
| use first-view masked depth instead of original depth (ablation) | RoBridge · Metaworld unseen camera pose | success rate (%): 60.8 -> 74.8 | Improved |
| use third-view mask instead of DINOv2 features (ablation) | RoBridge · Metaworld ablation | mean success rate (%): 50.36 -> 82.12 | Improved |
| use IOR instead of keypoints (ablation) | RoBridge · Metaworld ablation | mean success rate (%): 59.88 -> 82.12 | Improved |
| use IOR instead of language-only input (ablation) | RoBridge · Metaworld ablation | mean success rate (%): 39.28 -> 82.12 | Improved |
| use a larger VLM planner instead of a smaller VLM (ablation) | RoBridge · Metaworld ablation | mean success rate (%): 63.84 -> 82.12 | Improved |
| add DAgger training (ablation) | RoBridge · Metaworld ablation | mean success rate (%): 65.12 -> 82.12 | Improved |
| add online learning (ablation) | RoBridge · Metaworld ablation | mean success rate (%): 73.68 -> 82.12 | Improved |
| add domain randomization (ablation) | RoBridge · Metaworld ablation | mean success rate (%): 56.08 -> 82.12 | Improved |
| use original depth input on MT50 (in-distribution) (ablation) | RoBridge · Metaworld MT50 | success rate (%): 85.4 -> 87.2 | Hurt |
| replace ReKep with RoBridge on zero-shot unseen tasks | ReKep · Metaworld zero-shot unseen tasks | mean success rate (%): 28 -> 75 | Improved |
| replace best baseline with RoBridge on real-world tasks | ReKep · Real-world manipulation | mean success rate (%): 49.2 -> 83.3 | Improved |
Closest papers by meaning
- RoboCerebra: A Large-scale Benchmark for Long-horizon Robotic Manipulation Evaluation (NeurIPS 2025 · similarity 0.91)
- RoboMP$^2$: A Robotic Multimodal Perception-Planning Framework with Multimodal Large Language Models (ICML 2024 · similarity 0.90)
- RoboProcessBench: Benchmarking Process-Aware Understanding in Visual-Language Robotic Manipulation (NeurIPS 2026 · similarity 0.88)
- PIVOT-R: Primitive-Driven Waypoint-Aware World Model for Robotic Manipulation (NeurIPS 2024 · similarity 0.88)
- Embodied-R1.5: Evolving Physical Intelligence via Embodied Foundation Models (NeurIPS 2026 · similarity 0.88)
- RoboWits: Unexpected Challenges for Robotic Creative Problem Solving (NeurIPS 2026 · 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.
