Research map / Vision-language and multimodal / Robot action and navigation
EnsembleVLA: Ensemble Learning for Vision-Language Action Models
Mingchen Song, Xiang Deng, Jie Wei, Dongmei Jiang, Liqiang Nie, Weili Guan
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 |
|---|---|---|---|
| ensemble DP and DP3 via energy composition | DP3 · RoboTwin 2.0 (8 tasks avg) | success rate (%): 68.4 -> 82.8 | Improved |
| ensemble pi0 and pi0.5 via energy composition | pi0.5 · RoboTwin 2.0 (8 tasks avg) | success rate (%): 65.1 -> 74.6 | Improved |
| ensemble diffusion DP with flow-based pi0.5 | pi0.5 · RoboTwin 2.0 (8 tasks avg) | success rate (%): 65.1 -> 70.1 | Improved |
| add confidence gate (ablation) | EnsembleVLA · Handover Block (DP+DP3) | success rate (%): 64.0 -> 70.0 | Improved |
| add bounded residual Delta correction (ablation) | EnsembleVLA · Handover Block (DP+DP3) | success rate (%): 56.0 -> 64.0 | Improved |
| add learnable composition weights (ablation) | EnsembleVLA · Handover Block (DP+DP3) | success rate (%): 42.0 -> 56.0 | Improved |
| use energy-based composition instead of action-level averaging (ablation) | EnsembleVLA · Handover Block (DP+DP3) | success rate (%): 12.0 -> 42.0 | Improved |
| add OpenVLA-OFT as third policy to DP+DP3 ensemble | EnsembleVLA · RoboTwin 2.0 click alarmclock, move playingcard, open laptop | success rate (%): 96/89/93 -> 99/95/98 | Improved |
| ensemble DP and pi0.5 on real robot | pi0.5 · Real-world Cobot Mobile ALOHA | success rate (%): 57.5 -> 65.8 | Improved |
Closest papers by meaning
- VLA-REPLICA: A Low-Cost, Reproducible Benchmark for Real-World Evaluation of Vision-Language-Action Models (NeurIPS 2026 · similarity 0.91)
- ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models (NeurIPS 2026 · similarity 0.91)
- PriorVLA: Prior-Preserving Adaptation for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.90)
- ActionUNet: Improving Robustness of VLA Models with Efficient Multi-scale Fine-tuning (NeurIPS 2026 · similarity 0.90)
- Beyond World-Frame Action Heads: Motion-Centric Action Frames for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.90)
- VLA-Cache: Efficient Vision-Language-Action Manipulation via Adaptive Token Caching (NeurIPS 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.
