Event
Seoul National University releases omni-modal VLA model Dynin-Robotics
1 reports · 1 sources
The team of Professor Jaeheung Park at the AIDAS Lab of Seoul National University has developed a robot foundation model, Dynin-Robotics, which unifies language, images, video and robot actions in a single model, generating and selecting actions by predicting future scene images after actions; in evaluations across four real robot manipulation scenarios it achieved an average success rate of 78.4%, surpassing the NVIDIA and Physical Intelligence, a U.S. robot AI company, models compared in the experiments, and the paper has been accepted by NeurIPS 2026 and will be published in December. This result has so far been validated only in simulation and preset real robot experimental environments; the team says that before deployment in industrial applications, performance still needs to be evaluated on more objects, tasks and manipulation environments.
Report timeline
- Seoul National University's Dynin-Robotics robot foundation model guides actions by predicting future scenes, achieving 78.4% average success rate in real-world operation
Professor He Jiong's team at Seoul National University's AIDAS Lab proposed Dynin-Robotics, a robot foundation model that unifies language, images, video, and robot actions in a single model, generating and selecting actions by predicting future scene images after an action; the paper has been accepted by NeurIPS 2026.
Tech Xplore Robotics ·