Metraipolis › archive › dc_d23ef0eabb85
SUN: Agentic Robot Policy Learning with Persistent Task Programs — arXiv:2608.31167
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.31167
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-17T07:55:12.375Z
- SHA-256 of the archived text
18c21a4684b5618700948980554a7471e0739098cbaa213174f39a009bcbe801- Stable URL of this page
https://metraipolis.com/documento/dc_d23ef0eabb85
How to cite
SUN: Agentic Robot Policy Learning with Persistent Task Programs — arXiv:2608.31167 — arXiv (en), https://arxiv.org/abs/2608.31167 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d23ef0eabb85 (Metraipolis archive dc_d23ef0eabb85, sha256 18c21a4684b5618700948980554a7471e0739098cbaa213174f39a009bcbe801)
BibTeX
@misc{dcd23ef0eabb85,
title = {SUN: Agentic Robot Policy Learning with Persistent Task Programs — arXiv:2608.31167},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.31167},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d23ef0eabb85; sha256 18c21a4684b5618700948980554a7471e0739098cbaa213174f39a009bcbe801},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.31167. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Model-based control can directly execute specified objectives, while learning can amortize such behaviors into reactive policies, making their combination a natural solution to multi-stage manipulation. We introduce Semantically UNified (SUN) Programs, typed executables that compile grounded relations into aligned optimal control objectives, satisfaction predicates, and learning rewards. Our harness, Kuafu, equips a foundation model as a task-level agent to orchestrate scene preparation, verification, residual RL, and data production. The agent uses program feedback to repair candidate programs and training diagnostics to calibrate relative reward weights, retaining accepted task semantics across tool calls. Across nine multi-stage manipulation tasks, Kuafu achieves 82.03% average success rate, significantly outperforming all learned baselines. Its learned controllers generate demonstrations at 10.57x the human-teleoperation rate, yielding data that improve visualpolicy success by 23.6 percentage points over the strongest baseline. The policies transfer zero-shot to physical Franka and Kinova robots, demonstrating sim-to-real generalization. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-01 in cs.AI). Categorie arXiv: Robotics (cs.RO) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2608.31167 — 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 2026-10-17T07:28:26.945Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).