Metraipolis › archive › dc_db88a01a3e4f
OnEvoMemory: Evolving Memory through Online Robot Rollouts for Pretrained Robot Policies — arXiv:2608.08749
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.08749
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-19T09:12:37.477Z
- SHA-256 of the archived text
e5113a4e67f7d269e801258564d76483f3105a942fbd6c2961e25588c9fffafb- Stable URL of this page
https://metraipolis.com/documento/dc_db88a01a3e4f
How to cite
OnEvoMemory: Evolving Memory through Online Robot Rollouts for Pretrained Robot Policies — arXiv:2608.08749 — arXiv (en), https://arxiv.org/abs/2608.08749 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_db88a01a3e4f (Metraipolis archive dc_db88a01a3e4f, sha256 e5113a4e67f7d269e801258564d76483f3105a942fbd6c2961e25588c9fffafb)
BibTeX
@misc{dcdb88a01a3e4f,
title = {OnEvoMemory: Evolving Memory through Online Robot Rollouts for Pretrained Robot Policies — arXiv:2608.08749},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.08749},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_db88a01a3e4f; sha256 e5113a4e67f7d269e801258564d76483f3105a942fbd6c2961e25588c9fffafb},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.08749. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Long-horizon robot manipulation requires policies to track completed subtasks and critical interaction events. However, existing memory mechanisms heavily rely on external models or predefined update rules. To address this, we propose OnEvoMemory, a value-guided memory module for pretrained robot policies. It maintains recent context, high-value experiences, and salient transitions, while learning which experiences should be retained from trajectory outcomes. Offline demonstrations initialize the memory prior, whereas successful and unsuccessful online rollouts refine memory selection, helping the policy recognize task-stage transitions and avoid repeating completed subtasks. Experiments on long-horizon manipulation benchmarks show that OnEvoMemory improves the performance of the base VLA policy through both offline initialization and online memory evolution. Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-08-11 in cs.RO). Categorie arXiv: Robotics (cs.RO). Pagina del lavoro: https://arxiv.org/abs/2608.08749 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-19T09:11:58.753Z dalla pagina https://arxiv.org/catchup/cs.RO/2026-08-11?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).