Dual-Stream Memory for Vision-Language-Action Policies
cs.RO, cs.CV
Submitted: 2026-10-09
Updated: 2026-10-09
Comments: 8 pages, 6 figures, 3 tables. Evaluated on 16 RoboMME tasks and 4 real-world robotic manipulation tasks
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Sources
- RT-1: Robotics Transformer for Real-World Control at Scale
- Open X-Embodiment: Robotic Learning Datasets and RT-X Models
- Diffusion Policy: Visuomotor Policy Learning via Action Diffusion
- Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware
- MemoryVLA: Perceptual-Cognitive Memory in Vision-Language-Action Models for Robotic Manipulation
- MEM: Multi-Scale Embodied Memory for Vision Language Action Models
- NativeMEM: Native Memory Compression for Long-Horizon Robotic Manipulation
- CALVIN: A Benchmark for Language-Conditioned Policy Learning for Long-Horizon Robot Manipulation Tasks
- LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learning
- RoboMME: Benchmarking and Understanding Memory for Robotic Generalist Policies
- Perceiver: General Perception with Iterative Attention
- Flamingo: a Visual Language Model for Few-Shot Learning
- Attention Is All You Need
- Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context
- Compressive Transformers for Long-Range Sequence Modelling
- Gated Delta Networks: Improving Mamba2 with Delta Rule
- RoboTTT: Context Scaling for Robot Policies
- An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
- Emerging Properties in Self-Supervised Vision Transformers
- Feature Pyramid Networks for Object Detection
Related papers
- FMT x: An Efficient and Asymptotically Optimal Extension of the Fast Marching Tree for Dynamic Replanning
- MPCFormer: A physics-informed data-driven approach for explainable socially-aware autonomous driving
- RoboLab: A High-Fidelity Simulation Benchmark for Analysis of Task Generalist Policies
- ARC: A Reasoning Recipe for Robot Foundation Models
- HRDexDB: A 4D Dexterous Grasping Dataset Across Human and Multiple Robot Embodiments
- APT: Action Expert Pretraining Improves Instruction Generalization of Vision-Language-Action Policies