SimpleMemVLA

SimpleMemVLA is a vision-language-action (VLA) model for long-horizon robotic manipulation. It does not rely on a dedicated memory module; instead, it keeps the sampled history intact and passes it to the backbone in the timestamped video format the backbone was pretrained to process. The model is introduced in the paper SimpleMemVLA: A Simple but Effective Native-Video Memory for Vision-Language-Action Models.

• Paper (Hugging Face): https://huggingface.co/papers/2609.05533
• Paper (arXiv): https://arxiv.org/abs/2609.05533
• Project page: https://huggingface.co/collections/yinchenghust/simplememvla
• Code: https://github.com/wadeKeith/SimpleMemVLA
• License: MIT