Research map / Vision-language and multimodal / Robot action and navigation
MoLe-VLA: Dynamic Layer-skipping Vision Language Action Model via Mixture-of-Layers for Efficient Robot Manipulation
Rongyu Zhang, Menghang Dong, Yuan Zhang, Liang Heng, Xiaowei Chi, Gaole Dai, Li Du, Dan Wang, Yuan Du, Shanghang Zhang
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 STAR router instead of a linear router (ablation) | MoLe-OpenVLA (50% layer skip) · RLBench (OpenVLA) | mean success rate (%): 42.6 -> 45.6 | Improved |
| skip 50% of layers with a linear router (no distillation) (ablation) | OpenVLA · RLBench | mean success rate (%): 46.7 -> 42.6 | Hurt |
| add CogKD with MSE loss (ablation) | MoLe-OpenVLA with STAR · RLBench | mean success rate (%): 45.6 -> 51.4 | Improved |
| add CogKD with forward KL loss (ablation) | MoLe-OpenVLA with STAR · RLBench | mean success rate (%): 45.6 -> 49.6 | Improved |
| add CogKD with reverse KL loss (ablation) | MoLe-OpenVLA with STAR · RLBench | mean success rate (%): 45.6 -> 53.8 | Improved |
| combine MSE and reverse KL in CogKD (ablation) | MoLe-OpenVLA with STAR · RLBench | mean success rate (%): 53.8 -> 56.4 | Improved |
| apply MoLe layer-skipping to OpenVLA | OpenVLA · RLBench | mean success rate (%): 46.7 -> 56.4 | Improved |
| apply MoLe layer-skipping to pi0 | pi0 · RLBench | mean success rate (%): 50.0 -> 55.4 | Improved |
| apply MoLe layer-skipping to CogAct | CogAct · RLBench | mean success rate (%): 57.1 -> 60.2 | Improved |
| random layer skipping instead of learned routing | CogAct · RLBench | mean success rate (%): 57.1 -> 51.2 | Hurt |
| token-wise MoD routing instead of MoLe layer routing | CogAct · RLBench | mean success rate (%): 57.1 -> 56.5 | No clear effect |
| early-exit DeeR instead of MoLe | CogAct · RLBench | mean success rate (%): 58.8 -> 60.2 | No clear effect |
Closest papers by meaning
- ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models (NeurIPS 2026 · similarity 0.92)
- VLA-Cache: Efficient Vision-Language-Action Manipulation via Adaptive Token Caching (NeurIPS 2025 · similarity 0.92)
- NS-VLA: Towards Neuro-Symbolic Vision-Language-Action Models (NeurIPS 2026 · similarity 0.91)
- EfficientVLA: Training-Free Acceleration and Compression for Vision-Language-Action Models (NeurIPS 2025 · similarity 0.91)
- VLA-REPLICA: A Low-Cost, Reproducible Benchmark for Real-World Evaluation of Vision-Language-Action Models (NeurIPS 2026 · similarity 0.90)
- CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Models (CVPR 2025 · 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.
