arXiv Robotics — research abstracts· Yilin Wu, Rutav Shah, Shuaiyi Huang, Sumedh A. Sontakke, Homanga Bharadhwaj, Andrea Bajcsy, Max Simchowitz, Guanya Shi·· 7 hours agoEditorial score63
Memory Without Curation: Learning Multi-Stream Latent Memory for Robot Policies
Memory Without Curation: Learning Multi-Stream Latent Memory for Robot Policies
Summary
This research proposes IMPRINT, a policy-agnostic latent memory framework that augments vision-language-action (VLA) models with multi-stream memory without explicitly constraining what the policy should remember. IMPRINT improves average success over the base VLA by at least 26% in both simulation and real-world tasks, while maintaining the lowest inference latency among memory-augmented VLAs.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
Source：arXiv Robotics — research abstracts · arxiv.org