Metraipolis › archive › dc_cda79175a31a
Behavioral Steering of Vision-Language-Action Models in the Low-Parameter Regime — arXiv:2611.00522
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.00522
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-03T11:24:51.138Z
- SHA-256 of the archived text
5a86f6446421b42129b2c8c27e98f1d888bfcbbe340dd5513b5b0e5cfbdc955e- Stable URL of this page
https://metraipolis.com/documento/dc_cda79175a31a
How to cite
Behavioral Steering of Vision-Language-Action Models in the Low-Parameter Regime — arXiv:2611.00522 — arXiv (en), https://arxiv.org/abs/2611.00522 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_cda79175a31a (Metraipolis archive dc_cda79175a31a, sha256 5a86f6446421b42129b2c8c27e98f1d888bfcbbe340dd5513b5b0e5cfbdc955e)
BibTeX
@misc{dccda79175a31a,
title = {Behavioral Steering of Vision-Language-Action Models in the Low-Parameter Regime — arXiv:2611.00522},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.00522},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_cda79175a31a; sha256 5a86f6446421b42129b2c8c27e98f1d888bfcbbe340dd5513b5b0e5cfbdc955e},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.00522. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Reinforcement learning (RL) is increasingly used to adapt pretrained vision-language-action models (VLAs). We investigate the problem of behavioral steering: adapting a VLA's execution style to preferences such as carrying objects higher. To investigate how little trainable capacity such steering requires and how much pretrained capability can be retained, we freeze an OpenVLA-OFT 7B policy and optimize TinyLoRA adapters with 8 to 384 trainable parameters using either policy-gradient RL (GRPO) or evolutionary search (CMA-ES). With just 256 trainable parameters, we raise carry height by 120 mm on trained tasks at 92% success and by 74 mm on a held-out task at 80% success. By rescaling the learned update, we can tune the trade-off between style gains and capability retention without retraining. Increasing adapter capacity generally improves this trade-off, retaining more pretrained capability at comparable style gains. The choice of optimizer also depends on capacity, with CMA-ES being competitive at low parameter counts, while GRPO offers better trade-offs as capacity grows. These findings show that extremely low-dimensional updates can steer pretrained VLA behavior, with adapter capacity, optimizer choice, and update scaling offering practical control over adaptation and retention. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno). Categorie arXiv: Robotics (cs.RO) ; Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2611.00522 — 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 2026-11-03T11:02:37.772Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).