| tags: | |
| - robotics | |
| - vision-language-action | |
| - embodied-chain-of-thought | |
| # ERVLA Pretrained | |
| [Revisiting Embodied Chain-of-Thought for Generalizable Robot Manipulation](https://arxiv.org/abs/2606.03784) · [Project page](https://taoshuaiz.github.io/ERVLA/) | |
| ERVLA is a vision-language-action model for robot manipulation. It learns from embodied chain-of-thought that links task understanding to end-effector movements and image-space trajectories. With reasoning dropout, the model can generate actions directly at inference without decoding a chain of thought. | |
| This is the pretrained ERVLA model. Our embodied reasoning corpus contains 978,743 trajectories, 226.3 million samples, and 2,592.5 hours of robot data. | |