We are a group focused on vision-language-action models (VLAs). We wish to bring insights to the community with our research.

OpenHelix-Team introduces a novel family of fully open-source Vision-Language-Action Models (VLAs) that achieves state-of-the-art performance with substantially lower cost.

• ReconVLA (AAAI 2026 Best Paper Award): Reconstructive Vision-Language-Action Model as Effective Robot Perceiver
• Spatial Forcing (ICLR 2026): Implicit Spatial Representation Alignment for Vision-Language-Action Model

• OpenTrajBooster (ICRA 2026): Official implementation of TrajBooster

• Unified Diffusion VLA (ICLR 2026): The first open-sourced diffusion Vision-Language-Action model
• HiF-VLA: An efficient, bidirectional spatiotemporal expansion Vision-Language-Action Model
• frappe: Infusing World Modeling into Generalist Policies via Multiple Future Representation Alignment
• VLA-RFT: Vision-Language-Action Models with Reinforcement Fine-Tuning

• VLA-Adapter (AAAI 2026 (Oral)): An Effective Paradigm for Tiny-Scale Vision-Language-Action Model
• LLaVA-VLA (ICRA 2026): A Simple Yet Powerful Vision-Language-Action Model

• CEED-VLA: Consistency Vision-Language-Action Model with Early-Exit Decoding
• OpenHelix: An Open-Source Dual-System Vision-Language-Action Model for Robotic Manipulation

• VLA-2: Empowering Vision-Language-Action Models with an Agentic Framework for Unseen Concept Manipulation
• LongVLA (CoRL 2025): Unleashing Long-Horizon Capability of Vision-Language-Action Models for Robot Manipulation

• Awesome-Force-Tactile-VLA: A paper list of multimodal VLAs
• Awesome-VLA-RL: A taxonomy and summary of recent advances in VLA + RL

This initiative is jointly established and co-developed with the following research institutions:

• Westlake University
• The Hong Kong University of Science and Technology (Guangzhou)
• Zhejiang University
• Tsinghua University
• Beijing Academy of Artificial Intelligence (BAAI)
• Xi’an Jiaotong University
• Beijing University of Posts and Telecommunications

If you are interested in discussion or joining us, please send emails to songwenxuan0115@gmail.com.