Research map / Vision-language and multimodal / Robot action and navigation
PixelVLA: Advancing Pixel-level Understanding in Vision-Language-Action Model
Wenqi Liang, Gan Sun, Yao He, Jiahua Dong, Suyan Dai, Ivan Laptev, Salman Khan, Yang Cong
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 continuous action training stage with continuous action decoder (ablation) | OpenVLA after fine-tuning · SimplerEnv Google Robot (VA) | average success rate: 37.0 -> 43.8 | Improved |
| add pixel-level understanding enhancement on Pixel-160K (ablation) | OpenVLA after fine-tuning · SimplerEnv Google Robot (VA) | average success rate: 37.0 -> 48.0 | Improved |
| add continuous action training stage before pixel-level enhancement (full PixelVLA) (ablation) | Baseline+FT+PUE · SimplerEnv Google Robot (VA) | average success rate: 48.0 -> 50.1 | Improved |
| add continuous action training stage before pixel-level enhancement (ablation) | Baseline+FT+PUE · SimplerEnv Google Robot (VA) Open/Close Drawer | success rate: 21.3 -> 20.0 | Hurt |
| add pixel-level understanding enhancement on top of continuous action training (ablation) | Baseline+FT+CAT · SimplerEnv Google Robot (VA) | average success rate: 43.8 -> 50.1 | Improved |
| fine-tune OpenVLA on Fractal and Bridge only (ablation) | OpenVLA · SimplerEnv Google Robot (VA) | average success rate: 40.0 -> 37.0 | Hurt |
| apply PixelVLA to OpenVLA | OpenVLA · SimplerEnv Google Robot (VM) | average success rate: 32.7 -> 61.4 | Improved |
| apply PixelVLA to OpenVLA | OpenVLA · SimplerEnv Google Robot (VA) | average success rate: 40.0 -> 50.1 | Improved |
| apply PixelVLA to pi0 | pi0 · SimplerEnv Google Robot (VM) | average success rate: 58.8 -> 63.3 | Improved |
| apply PixelVLA to pi0 | pi0 · SimplerEnv Google Robot (VA) | average success rate: 54.8 -> 56.5 | Improved |
| apply PixelVLA to pi0 | pi0 · SimplerEnv WidowX | average task success rate: 27.1 -> 33.8 | Improved |
| apply PixelVLA to OpenVLA | OpenVLA · SimplerEnv WidowX | average task success rate: 1.0 -> 16.7 | Improved |
Later papers that compared with this method
1 later accepted paper reports PixelVLA as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Action-Sketcher: From Reasoning to Action via Visual Sketches for Robotic Manipulation (CVPR 2026 · better on 2 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 3 on most shared results.
- TraceVLA — TraceVLA: Visual Trace Prompting Enhances Spatial-Temporal Awareness for Generalist Robotic Policies (ICLR 2025) · better on 3 of 3 shared results
- Dita — Dita: Scaling Diffusion Transformer for Generalist Vision-Language-Action Policy (ICCV 2025) · better on 1 of 3 shared results
- CoT-VLA — CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Models (CVPR 2025) · better on 1 of 1 shared result
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
- CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Models (CVPR 2025 · similarity 0.92)
- ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models (NeurIPS 2026 · similarity 0.91)
- UP-VLA: A Unified Understanding and Prediction Model for Embodied Agent (ICML 2025 · similarity 0.91)
- FineVLA: Fine-Grained Instruction Alignment for Steerable Vision-Language-Action Policies (NeurIPS 2026 · similarity 0.91)
- PriorVLA: Prior-Preserving Adaptation 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.
