Research map / Vision-language and multimodal / Robot action and navigation
XR-1: Towards Versatile Vision-Language-Action Models via Learning Unified Vision-Motion Representations
Shichao Fan, Kun Wu, Zhengping Che, Xinhua Wang, Di Wu, Fei Liao, Ning Liu, Yixue Zhang, Zhen Zhao, Zhiyuan Xu, Meng Li, Qingjie Liu and 3 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 UVMC-guided pretraining (Stages 1-2) to lightweight model (ablation) | XR-1-Light · Dual-Arm UR-5e, 6 tasks | average success rate (%): 42.5 -> 57.5 | Improved |
| add cross-modal KL alignment loss between vision and motion codes (ablation) | XR-1 · Dual-Arm UR-5e, 6 tasks | average success rate (%): 48.3 -> 66.7 | Improved |
| add UVMC learning on downstream data (no large-scale pretraining) (ablation) | XR-1 · Dual-Arm UR-5e, 6 tasks | average success rate (%): 28.3 -> 48.3 (w/o KL) -> 66.7 | Improved |
| scale Stage-1 pretraining data from 1% to 100% (ablation) | XR-1 · Dual-Arm UR-5e, 6 tasks, no fine-tuning | average success rate (%): 29.2 -> 38.3 -> 53.3 -> 65.0 | Improved |
| pretrain Stage-2 on XR-D cross-embodiment data and fine-tune, versus downstream-only Stage-1/2 (ablation) | XR-1 · Dual-Arm UR-5e, 6 tasks | average success rate (%): 66.7 -> 81.6 | Improved |
| use XR-1 instead of pi0 on unseen scenarios | XR-1 · Dual-Arm Franka SweepTrash, dynamic distractors | success rate (%): 5 -> 55 | Improved |
| use XR-1 instead of baselines on unseen embodiment Tien Kung 2.0 | XR-1 · Tien Kung 2.0, 20 tasks | average success rate (%): 41.0 (pi0.5) -> 72.0 | Improved |
Closest papers by meaning
- Beyond Data Scaling: Representation-Centric Pre-training for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.93)
- ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models (NeurIPS 2026 · similarity 0.92)
- NS-VLA: Towards Neuro-Symbolic Vision-Language-Action Models (NeurIPS 2026 · similarity 0.92)
- RotVLA: Rotational Latent Action for Vision-Language-Action Model (NeurIPS 2026 · similarity 0.91)
- FineVLA: Fine-Grained Instruction Alignment for Steerable Vision-Language-Action Policies (NeurIPS 2026 · similarity 0.91)
- CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Models (CVPR 2025 · similarity 0.91)
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.
