VLANeXt codebase release reveals recipes for building strong VLA models
ccloy · x · 2026-08-31
Released VLANeXt, a research-oriented codebase for robotics research, aiming to address the fragmentation in the Vision-Language-Action (VLA) field and uncover what makes these models strong.
Key Highlights:
- Experimental Basis: Insights from over 500 controlled experiments, forming the basis of an ICML paper.
- Design Exploration: Systematically explored the VLA design space starting from a simple RT-2/OpenVLA baseline to distill practical recipes.
- Extended Scope: The codebase covers emerging directions in robotics foundation models, including various VLA backbones, latent action learning, latent-space predictive modeling, and world-action modeling.
More from Embodied
- Peking University's SimForcing distills simulation motion priors into robot world models — PekingUniversity · 2026-10-16
- NVIDIA and Mercedes-Benz Expand Partnership on Next-Gen Automated Driving — nvidia · 2026-10-16
- ~25,000 Humanoid Robots Shipped in H1 2026, Over 95% from Chinese Vendors — ingliguori · 2026-10-16
- openpilot brings stoplight-stopping autonomy to a 2019 Corolla — k7agar · 2026-10-16
- Hironobu Fujiyoshi publishes deck: from AI agents to Agentic Robotics — HirokatuKataoka · 2026-10-16
- CoDA co-evolves simulation and robot policies via shared failure memory, lifting OOD success from 5.3% to 27.6% — PLAN-Lab · 2026-10-16