Research map / Vision-language and multimodal / Robot action and navigation
AtomicVLA: Unlocking the Potential of Atomic Skill Learning in Robots
Likui Zhang, Tao Tang, Zhihao Zhan, Xiuwei Chen, Zisheng Chen, Jianhua Han, Jiangtong Zhu, Pei Xu, Hang Xu, 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 |
|---|---|---|---|
| add skill-guided MoE (SG-MoE) routing on atomic skills (ablation) | pi0 · LIBERO-Long | success rate (%): 85.2 -> 95.2 | Improved |
| add generic token-level MoE (ablation) | pi0 · LIBERO-Long | success rate (%): 85.2 -> 88.6 | Improved |
| add timestep-conditioned MoDE routing (ablation) | pi0 · LIBERO-Long | success rate (%): 85.2 -> 89.5 | Improved |
| route experts by atomic skill instead of token-level MoE (ablation) | AtomicVLA · LIBERO-Long | success rate (%): 88.6 -> 95.2 | Improved |
| route experts by atomic skill instead of timestep (MoDE) (ablation) | AtomicVLA · LIBERO-Long | success rate (%): 89.5 -> 95.2 | Improved |
| add atomic-skill MoE and planning on pi0 | pi0 · LIBERO average | success rate (%): 94.2 -> 96.6 | Improved |
| add atomic-skill MoE and planning on pi0.5 | pi0.5 · LIBERO average | success rate (%): 96.9 -> 97.8 | Improved |
| add atomic-skill MoE on pi0 | pi0 · CALVIN ABC->D | average length: 3.87 -> 4.09 | Improved |
| add atomic-skill MoE on pi0.5 | pi0.5 · CALVIN ABC->D | average length: 4.02 -> 4.27 | Improved |
| add atomic-skill MoE for multi-task long-horizon training | pi0.5 · Real-world long-horizon tasks | success rate (%) average: 45 -> 63.3 | Improved |
| learn new skill by adding an expert instead of fine-tuning the single decoder | pi0.5 · Real-world continual learning | success rate (%) average on base tasks: -15.0 -> -1.3 | Improved |
| add atomic-skill MoE | pi0.5 · Real-world complex scenes | success rate (%) average: 33.3 -> 43.3 | Improved |
Later papers that compared with this method
2 later accepted papers report AtomicVLA as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- VLA-Pro: Cross-Task Procedural Memory Transfer for Vision-Language-Action Models (NeurIPS 2026 · better on 2 of 3 shared results)
- SkillNet: Hierarchical Skill Modeling for Compositional Generalization in Vision-Language Action Models (ICML 2026 · better on 1 of 1 shared result)
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.
- CoT-VLA — CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Models (CVPR 2025) · better on 2 of 2 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
- ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models (NeurIPS 2026 · similarity 0.89)
- NS-VLA: Towards Neuro-Symbolic Vision-Language-Action Models (NeurIPS 2026 · similarity 0.89)
- UniDomain: Pretraining a Unified PDDL Domain from Real-World Demonstrations for Generalizable Robot Task Planning (NeurIPS 2025 · similarity 0.89)
- FineVLA: Fine-Grained Instruction Alignment for Steerable Vision-Language-Action Policies (NeurIPS 2026 · similarity 0.89)
- PriorVLA: Prior-Preserving Adaptation for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.89)
- STAR: Learning Diverse Robot Skill Abstractions through Rotation-Augmented Vector Quantization (ICML 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.
