Research map / Vision-language and multimodal / Robot action and navigation
RoboInter: A Holistic Intermediate Representation Suite Towards Robotic Manipulation
Hao Li, Ziqin Wang, Zi-han Ding, Shuai Yang, Yilun Chen, Yang Tian, Xiaolin Hu, Tai Wang, Dahua Lin, Feng Zhao, Si Liu, Jiangmiao Pang
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 pretrained VLM planner (implicit intermediate, IC-E2E) | Vanilla executor · Open-loop In-the-Wild | mOLS: 0.3086 -> 0.3218 | Improved |
| make intermediate representations explicit (EC-E2E) | IC-E2E · Open-loop In-the-Wild | mOLS: 0.3218 -> 0.3340 | Improved |
| decouple planning and execution (Te-Modular) | EC-E2E · Open-loop In-the-Wild | mOLS: 0.3340 -> 0.3543 | Improved |
| use visual prompts instead of textual F-CoT (Im-Modular) (ablation) | Te-Modular · Open-loop In-the-Wild | mOLS: 0.3543 -> 0.3439 | Hurt |
| add Subtask (ablation) | Vanilla · Open-loop Oracle+Executor | mOLS: 0.3086 -> 0.3146 | Improved |
| add Primitive Skill on top of Subtask (ablation) | Vanilla · Open-loop Oracle+Executor | mOLS: 0.3146 -> 0.3159 | Improved |
| add Object Box (ablation) | Subtask+Skill · Open-loop Oracle+Executor | mOLS: 0.3159 -> 0.3289 | Improved |
| add Trace (ablation) | all other representations · Open-loop Oracle+Executor | mOLS: 0.3435 -> 0.3861 | Improved |
| add Affordance alone (ablation) | Vanilla · Open-loop | mOLS: 0.3086 -> 0.3322 | Improved |
| add Primitive skill alone (ablation) | Vanilla · Open-loop | OLS@0.01: 0.0189 -> 0.0156 | Hurt |
| initialize from Planner (IC-E2E) | Vanilla · Real-world OOD | success rate: 38.3% -> 58.3% | Improved |
| jointly generate explicit CoT (EC-E2E) | IC-E2E · Real-world ID | success rate: 77.3% -> 68.3% | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Robobrain-2.0-7B — RoboBrain: A Unified Brain Model for Robotic Manipulation from Abstract to Concrete (CVPR 2025) · better on 6 of 6 shared results
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
- FineVLA: Fine-Grained Instruction Alignment for Steerable Vision-Language-Action Policies (NeurIPS 2026 · similarity 0.91)
- RoboMamba: Efficient Vision-Language-Action Model for Robotic Reasoning and Manipulation (NeurIPS 2024 · similarity 0.91)
- Robots Pre-train Robots: Manipulation-Centric Robotic Representation from Large-Scale Robot Datasets (ICLR 2025 · similarity 0.91)
- Robotic Visual Instruction (CVPR 2025 · similarity 0.90)
- RoboProcessBench: Benchmarking Process-Aware Understanding in Visual-Language Robotic Manipulation (NeurIPS 2026 · similarity 0.90)
- NS-VLA: Towards Neuro-Symbolic Vision-Language-Action Models (NeurIPS 2026 · similarity 0.90)
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.
