Galaxea launches R1 Series for wheeled humanoid manipulation - 2025-02-25 - Article - Galaxea launched the R1 Series as a full-size wheeled dual-arm humanoid platform for data collection, research and real-world manipulation; the robot combines dual arms, a 4-DoF torso and a three-wheel steering chassis.
EvoScene-VLA: Evolving Scene Beliefs Inside the Action Decoder for Chunked Robot Control - 2026-05-21 - Research Publication - The authors argue for a persistent action-updated scene state across control calls, and introduce EvoScene-VLA. On the Galaxea R1-Lite real robot, EvoScene-VLA outperforms all baselines. - source: arxiv.org
CORAL: Scalable Multi-Task Robot Learning via LoRA Experts - 2026-03-10 - Research Publication - To address this dilemma, the authors present CORAL, a backbone- and embodiment-agnostic framework designed primarily to mitigate multi-task interference while remaining naturally extensible to a continuous stream of new… - source: arxiv.org
AtomVLA: Scalable Post-Training for Robotic Manipulation via Predictive Latent World Models - 2026-03-09 - Research Publication - To tackle this problem, the authors propose \method, the first subtask-aware VLA framework integrated with a scalable offline post-training pipeline. Finally, experiments conducted in the real world using the Galaxea R1… - source: arxiv.org
LingBot-VLA benchmarks generalist policies on Galaxea R1Pro tasks - 2026-01-26 - Research Publication - LingBot-VLA reports GM-100 benchmark results across Agibot G1, AgileX, and Galaxea R1Pro robot platforms. The released repository shows Galaxea R1Pro as one of the real-world benchmark platforms and publishes model… - source: github.com
Galaxea publishes Open-World Dataset and G0 dual-system VLA model for R1 - 2025-08-30 - Research Publication - Galaxea published an open-world robot-behavior dataset collected in authentic human living and working environments. The paper also introduces G0, a dual-system robotics framework combining VLM planning with VLA execution… - source: arxiv.org