Research map / Vision-language and multimodal / Robot action and navigation
A0: An Affordance-Aware Hierarchical Model for General Robotic Manipulation
Rongtao Xu, Jian Zhang, Minghao Guo, Youpeng Wen, Haoting Yang, Min Lin, Jianzheng Huang, Zhe Li, Kaidong Zhang, Liqiong Wang, Yuxuan Kuang, Meng Cao and 2 more
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 Position Offset Attention (ablation) | A0-1B · Maniskill-5k offline | MAE: 6.3 -> 5.5 | Improved |
| add Position Offset Attention (ablation) | A0-1B · HOI4D-22k offline | MAE: 47.9 -> 47.5 | Improved |
| add Position Offset Attention (ablation) | A0-1B · DROID-3k offline | MAE: 18.5 -> 17.5 | Improved |
| add Spatial Information Aggregation Layer (ablation) | A0-1B · HOI4D-22k offline | MAE: 61.1 -> 47.5 | Improved |
| add Spatial Information Aggregation Layer (ablation) | A0-1B · Maniskill-5k offline | MAE: 10.2 -> 5.5 | Improved |
| add Spatial Information Aggregation Layer (ablation) | A0-1B · DROID-3k offline | MAE: 19.6 -> 17.5 | Improved |
| add pre-training on 1M PixMo-One-Point contact points (ablation) | A0-1B · HOI4D, DROID, ManiSkill offline transfer | MAE | Improved |
| use A0 instead of Robopoint for first contact point | Robopoint · HOI4D first-point MAE | MAE: 121.09 -> 54.46 | Improved |
| use A0 instead of Robopoint for first contact point | Robopoint · DROID first-point MAE | MAE: 27.47 -> 14.13 | Improved |
| use A0-1B instead of Molmo points with the same action execution module | Molmo · Franka real-world (4 tasks, avg success) | success rate (%): 43.75 -> 62.50 | Improved |
| use A0-1B instead of MOKA on Kinova | MOKA · Kinova real-world (4 tasks, avg success) | success rate (%): 45.00 -> 53.75 | Improved |
| use A0-1B instead of MOKA/ReKep for Place Object | MOKA · Kinova Place Object | success rate (%): 70 -> 60 | 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.
- MOKA — MokA: Multimodal Low-Rank Adaptation for MLLMs (NeurIPS 2025) · better on 2 of 3 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
- AffordDP: Generalizable Diffusion Policy with Transferable Affordance (CVPR 2025 · similarity 0.93)
- Learning Precise Affordances from Egocentric Videos for Robotic Manipulation (ICCV 2025 · similarity 0.93)
- Afford-X: Generalizable and Slim Affordance Reasoning for Task-Oriented Manipulation (TPAMI 2026 · similarity 0.92)
- Robo-ABC: Affordance Generalization Beyond Categories via Semantic Correspondence for Robot Manipulation (ECCV 2024 · similarity 0.92)
- PALM: Progress-Aware Policy Learning via Affordance Reasoning for Long-Horizon Robotic Manipulation (CVPR 2026 · similarity 0.91)
- OmniManip: Towards General Robotic Manipulation via Object-Centric Interaction Primitives as Spatial Constraints (CVPR 2025 · similarity 0.89)
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.
