HiMe: Hierarchical Embodied Memory for Long-Horizon Vision-Language-Action Control
Lists
Only you see this
Abstract
Current Vision-Language-Action (VLA) models excel at robotic manipulation but often struggle with non-Markovian tasks requiring long-term memory and reasoning due to their reliance on immediate observations. Existing solutions face a ''frequency-competence paradox,'' where stronger reasoning models are too slow for real-time control, while faster models lack sufficient reasoning capabilities. To resolve this architectural misalignment, we propose HiMe, a Hierarchical Embodied Memory framework that decouples embodied intelligence into a high-frequency Executor for execution, a Sentry for working memory, and a Planner for long-term strategy. We also introduce a dynamic knowledge system based on cross-modal semantic schemas and active management mechanisms, allowing robots to maintain memory plasticity through ''Add, Update, and Delete'' operations. This hierarchical design effectively balances the conflict between real-time execution and slow thinking planning, significantly improving success rates in long-horizon tasks. Experiments demonstrate that this approach not only outperforms flat memory baselines but also exhibits the novel ability to self-correct its internal knowledge based on human preferences.
BibTeX
@inproceedings{ji2026hime,
author = {Li Ji and Siyin Wang and Pengfang Qian and Xiaopeng Yu and Yihai Tian and Zhaoye Fei and Jingjing Gong and Xipeng Qiu},
title = {{HiMe: Hierarchical Embodied Memory for Long-Horizon Vision-Language-Action Control}},
booktitle = {International Conference on Machine Learning},
year = {2026},
url = {https://openreview.net/forum?id=vVVbGj9cMC},
}Loading related content...