publications

2026

1. CoRL

   

   GeomVLA: Unifying Scene, Motion, and Action in 3D

   Ziyin Xiong, Nikolaos Gkanatsios, Moritz Reuss, and Katerina Fragkiadaki

   In Conference on Robot Learning, 2026

   Abs arXiv Code Website

   We present GeomVLA, a vision-language-action model that unifies scene perception, latent future scene-motion prediction, and action generation in a shared metric 3D frame. GeomVLA achieves state-of-the-art results on CALVIN and strong performance on LIBERO, RoboTwin2.0, and real-world manipulation without robot-action pretraining.

2025

1. IROS

   

   Ag2x2: Robust Agent-Agnostic Visual Representations for Zero-Shot Bimanual Manipulation

   Ziyin Xiong^*, Yinghan Chen^*, Puhao Li, Yixin Zhu, Tengyu Liu, and Siyuan Huang

   In IEEE/RSJ International Conference on Intelligent Robots and Systems, 2025

   Abs arXiv Code Website

   We present Ag2x2, a framework that advances the autonomous acquisition of bimanual manipulation skills through agent-agnostic and coordination-aware visual representations that jointly encode object and hand motion patterns.