ECHO: Continuous Hierarchical Memory for Vision-Language-Action Models
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866916001706147840 |
|---|---|
| author | Hu, Yanbin Cui, Jin Lu, Jiayi Yang, Ruixuan Ye, Jun Zhao, Boran Chen, Xingyu Lan, Xuguang Ren, Pengju |
| author_facet | Hu, Yanbin Cui, Jin Lu, Jiayi Yang, Ruixuan Ye, Jun Zhao, Boran Chen, Xingyu Lan, Xuguang Ren, Pengju |
| contents | Memory capacity is a critical factor determining the performance of Vision-Language-Action (VLA) models in long-horizon manipulation tasks. Existing memory-augmented architectures primarily rely on linear or flat storage, lacking structural priors for manipulation categories and hierarchical organization. This deficiency hinders efficient experience retrieval and limits generalization to unseen long-horizon task compositions. Inspired by the hierarchical organization of human experience, we propose ECHO (Experience Consolidation and Hierarchical Organization), a novel memory framework operating within a Continuous Hierarchical Space. By employing a hyperbolic autoencoder, ECHO maps VLA hidden states into this space. Leveraging hyperbolic metrics and entailment constraint mechanisms, experience vectors are organized into a semantic memory tree that supports efficient top-down retrieval. In parallel, a background consolidation mechanism continuously refines the memory tree through geometric interpolation and structural splitting, supporting virtual memory synthesis in the continuous space. We integrate ECHO into the $π_0$ foundation model. Evaluations on LIBERO and preliminary real-world experiments demonstrate the effectiveness of our approach, notably achieving a 12.8% absolute improvement in execution success rate over the $π_0$ baseline on LIBERO-Long, while improving compositional generalization on cross-suite unseen long-horizon tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_10993 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | ECHO: Continuous Hierarchical Memory for Vision-Language-Action Models Hu, Yanbin Cui, Jin Lu, Jiayi Yang, Ruixuan Ye, Jun Zhao, Boran Chen, Xingyu Lan, Xuguang Ren, Pengju Robotics Memory capacity is a critical factor determining the performance of Vision-Language-Action (VLA) models in long-horizon manipulation tasks. Existing memory-augmented architectures primarily rely on linear or flat storage, lacking structural priors for manipulation categories and hierarchical organization. This deficiency hinders efficient experience retrieval and limits generalization to unseen long-horizon task compositions. Inspired by the hierarchical organization of human experience, we propose ECHO (Experience Consolidation and Hierarchical Organization), a novel memory framework operating within a Continuous Hierarchical Space. By employing a hyperbolic autoencoder, ECHO maps VLA hidden states into this space. Leveraging hyperbolic metrics and entailment constraint mechanisms, experience vectors are organized into a semantic memory tree that supports efficient top-down retrieval. In parallel, a background consolidation mechanism continuously refines the memory tree through geometric interpolation and structural splitting, supporting virtual memory synthesis in the continuous space. We integrate ECHO into the $π_0$ foundation model. Evaluations on LIBERO and preliminary real-world experiments demonstrate the effectiveness of our approach, notably achieving a 12.8% absolute improvement in execution success rate over the $π_0$ baseline on LIBERO-Long, while improving compositional generalization on cross-suite unseen long-horizon tasks. |
| title | ECHO: Continuous Hierarchical Memory for Vision-Language-Action Models |
| topic | Robotics |
| url | https://arxiv.org/abs/2605.10993 |