Metraipolis › archive › dc_e4ec270bc439
CaPtain: A Pipeline for Post-Training Code-as-Policy Robot Agents — arXiv:2610.40799
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.40799
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-03T05:47:11.584Z
- SHA-256 of the archived text
fa5a7c7e579f2b7376b27a9a633918b7bebeadd8c42c7e0bc838fc89874ad33f- Stable URL of this page
https://metraipolis.com/documento/dc_e4ec270bc439
How to cite
CaPtain: A Pipeline for Post-Training Code-as-Policy Robot Agents — arXiv:2610.40799 — arXiv (en), https://arxiv.org/abs/2610.40799 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_e4ec270bc439 (Metraipolis archive dc_e4ec270bc439, sha256 fa5a7c7e579f2b7376b27a9a633918b7bebeadd8c42c7e0bc838fc89874ad33f)
BibTeX
@misc{dce4ec270bc439,
title = {CaPtain: A Pipeline for Post-Training Code-as-Policy Robot Agents — arXiv:2610.40799},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.40799},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_e4ec270bc439; sha256 fa5a7c7e579f2b7376b27a9a633918b7bebeadd8c42c7e0bc838fc89874ad33f},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.40799. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Across coding, web, and tool-use tasks, language-model agents are now routinely trained with the same recipe: supervised fine-tuning (SFT) to learn the interface, then multi-turn reinforcement learning (RL) to master it. Code-as-policy (CaP) casts robot manipulation in the same form, as the policy writes code against robot APIs and reads back the results, so the same recipe should apply. Yet existing CaP agents sidestep training: they either prompt closed frontier models or delegate manipulation to pre-trained vision-language-action models (VLAs) and high-level skills, while small open vision-language models (VLMs) fail outright. We present CaPtain, an open post-training pipeline that turns a small open VLM into a capable CaP agent, with no VLA, frontier model, or privileged state at test time. CaPtain trains and deploys the policy inside CaPtain-Harness, which exposes low-level perception, geometry, and guarded-motion primitives across simulation and real hardware. We first align the base model via SFT on synthetic trajectories generated with GPT-6 Astra, then apply multi-turn RL. Multi-turn RL must credit tens of agent requests, whose physical effects often surface turns later, from a single end-of-episode outcome. We therefore introduce a physical critic that reads privileged simulator state, never seen by the policy, and credits each request by how much it changes the predicted chance of success. Starting from Qwen3.5-4B (0% zero-shot), CaPtain-4B reaches 55.4% held-out LIBERO success after SFT and 73.8% after RL. On LIBERO-PRO, it reaches 78.5%, surpassing several frontier-model agents, and its RL gains persist across three perturbation suites (+17.6 to +21.2 points). Without target-domain training, it attains 83.3% on SimplerEnv WidowX, exceeds frontier models on CaP-Bench, and runs on a real robot after simulation-only adaptation. We release all code, data, and model weights. Tema degli esploratori: statistica e scienza dei dati (elenco arXiv cs.RO, nuovi lavori del giorno). Categorie arXiv: Robotics (cs.RO). Pagina del lavoro: https://arxiv.org/abs/2610.40799 — 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 2026-11-03T05:33:48.000Z dalla pagina https://arxiv.org/list/cs.RO/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).