SimpleMemVLA
Collection
A Simple but Effective Native-Video Memory for Vision-Language-Action Models • 11 items • Updated • 1
SimpleMemVLA
SimpleMemVLA is a Vision-Language-Action (VLA) model for long-horizon robot manipulation with a native-video memory mechanism. It keeps the sampled history intact and processes it as timestamped video through the backbone, using the hidden states of a generated sub-task to condition a flow-matching action head.
Paper: SimpleMemVLA: A Simple but Effective Native-Video Memory for Vision-Language-Action Models
Project page: https://huggingface.co/collections/yinchenghust/simplememvla
- Downloads last month
- 23