GazeVLA: Learning Human Intention for Robotic Manipulation
GazeVLA: Learning Human Intention for Robotic Manipulation
This paper introduces GazeVLA, a framework that leverages human gaze cues to model intentions for robotic manipulation. By capturing the synergy between high-level task semantics and low-level actions encoded in gaze, the method derives mid-level representations that facilitate cross-embodiment knowledge transfer. The approach employs a Chain-of-Thought reasoning paradigm during inference to handle long-horizon and fine-grained tasks across diverse robot embodiments. Experiments in both simulation and real-world environments demonstrate consistent outperformance over baseline methods and generalization to unseen settings, all without requiring robotic intention annotations.
The authors propose GazeVLA, a method that models human gaze to derive mid-level intention representations for robotic manipulation, enabling cross-embodiment knowledge transfer and long-horizon task execution without explicit robot intention annotations; results are reported from extensive experiments in both simulation and real-world settings.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Dataset details reported by the source
- Tasks
- robotic manipulation Open source S1Open source S4
- Embodiments
- diverse embodiments Open source S4
- paper: not_reported
What remains unknown
Dataset licensing and commercial-use rights are not established here.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
learning. In this work, we model human intention to enable efficient learning from human demonstrations. Among human behavioral cues, gaze simultaneously encodes high-level task semantics and exhibits a synergistic coupling with low-level action. By modeling intention through gaze, we derive a mid-level representation that facilitates cross-embodiment knowledge transfer. We demonstrate that robots can internalize hu
Open source S3
GazeVLA: Learning Human Intention for Robotic Manipulation
Open source S1
man intentions to enhance manipulation, even without robotic intention annotations. To capture the synergy between intention and action, we adopt a Chain-of-Thought reasoning paradigm during inference. Our approach enables the execution of long-horizon and fine-grained tasks across diverse embodiments. Extensive experiments in both simulation and real-world demonstrate that our method consistently outperforms baselin
Open source S4
e approaches and generalizes to unseen settings. Project page: https://gazevla.github.io .
Open source S5
Implications for data suppliers
RoboSignal interpretation and collection questions, not statements of buyer demand.
- Confirm the required data type and collection setting with the buyer; this source does not establish a complete collection specification.
- Check the exact dataset version, license, modalities and task coverage against the intended use.
- Validate demand and acceptance criteria with a buyer before scaling. Publication, popularity and a research result do not establish a purchase commitment.
Source：arXiv Robotics — research abstracts · arxiv.org