{"slug": "same-weights-different-robot-a-deployment-safety-view-of-vla-policies", "title": "Same Weights, Different Robot: A Deployment Safety View of VLA Policies", "summary": "Researchers have identified a deployment-safety gap in vision-language-action (VLA) robot policies, showing that identical model checkpoints can produce different physical actions due to variations in action unnormalization and controller conventions. In experiments on LIBERO-Goal and LIBERO-Spatial benchmarks, substituting a single metadata key caused mean action-space drift of 0.199 and reduced task success from 28/28 to 2/28 and 0/26, respectively. The findings demonstrate that action-space metadata must be treated as part of the executable policy and verified before deployment to prevent safety failures.", "body_md": "# Computer Science > Cryptography and Security\n\n[Submitted on 2 Jun 2026]\n\n# Title:Same Weights, Different Robot: A Deployment Safety View of VLA Policies\n\n[View PDF](/pdf/2606.03724)\n\n[HTML (experimental)](https://arxiv.org/html/2606.03724v1)\n\nAbstract:Vision-language-action (VLA) policies are often treated as checkpoint-defined objects: if the weights, prompt, and benchmark suite match, the deployment is assumed to be the same policy. Robot execution breaks this assumption because the same normalized model output can become a different physical action after action unnormalization and controller conventions are applied. This creates a deployment-safety gap: safety review can certify the checkpoint while missing the executable robot policy that reaches the controller. We formalize this gap as an executable policy specification problem: a VLA policy includes the learned model, action representation, metadata-selected unnormalizer, and controller-facing conventions. Under this view, identical checkpoints can be executable-inequivalent. For quantile-style action normalization, we derive a closed-form metadata mismatch transform and an ExecSpec certificate that measures action-space semantic drift without model inference or rollout. On LIBERO-Goal replay, substituting a plausible sibling metadata key yields mean drift 0.199 over six non-gripper action dimensions and reduces success from 28/28 to 2/28 under full substitution. On LIBERO-Spatial replay, the same substituted key reduces success from 26/26 to 0/26. The same full-substitution protocol gives 0/28 success for all four Object substitutions and 0/23 or 1/23 success on Long. Identity-key, replay-validity, no-op filtering, raw-vs-correct replay, mask/gripper, synthetic upper-bound, and OpenVLA-style unnormalizer interface checks rule out several simpler explanations. These results do not certify closed-loop or hardware safety. They support a narrower deployment-safety view: action-space metadata is part of the executable policy and should be checked before rollout.\n\n### References & Citations\n\nLoading...\n\n# Bibliographic and Citation Tools\n\nBibliographic Explorer\n\n*(*[What is the Explorer?](https://info.arxiv.org/labs/showcase.html#arxiv-bibliographic-explorer))\nConnected Papers\n\n*(*[What is Connected Papers?](https://www.connectedpapers.com/about))\nLitmaps\n\n*(*[What is Litmaps?](https://www.litmaps.co/))\nscite Smart Citations\n\n*(*[What are Smart Citations?](https://www.scite.ai/))# Code, Data and Media Associated with this Article\n\nalphaXiv\n\n*(*[What is alphaXiv?](https://alphaxiv.org/))\nCatalyzeX Code Finder for Papers\n\n*(*[What is CatalyzeX?](https://www.catalyzex.com))\nDagsHub\n\n*(*[What is DagsHub?](https://dagshub.com/))\nGotit.pub\n\n*(*[What is GotitPub?](http://gotit.pub/faq))\nHugging Face\n\n*(*[What is Huggingface?](https://huggingface.co/huggingface))\nScienceCast\n\n*(*[What is ScienceCast?](https://sciencecast.org/welcome))# Demos\n\n# Recommenders and Search Tools\n\nInfluence Flower\n\n*(*[What are Influence Flowers?](https://influencemap.cmlab.dev/))\nCORE Recommender\n\n*(*[What is CORE?](https://core.ac.uk/services/recommender))# arXivLabs: experimental projects with community collaborators\n\narXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.\n\nBoth individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.\n\nHave an idea for a project that will add value for arXiv's community? [ Learn more about arXivLabs](https://info.arxiv.org/labs/index.html).", "url": "https://wpnews.pro/news/same-weights-different-robot-a-deployment-safety-view-of-vla-policies", "canonical_source": "https://arxiv.org/abs/2606.03724", "published_at": "2026-06-04 05:31:42+00:00", "updated_at": "2026-06-04 05:47:09.248755+00:00", "lang": "en", "topics": ["ai-safety", "robotics", "machine-learning", "artificial-intelligence"], "entities": ["LIBERO-Goal", "LIBERO-Spatial", "VLA"], "alternates": {"html": "https://wpnews.pro/news/same-weights-different-robot-a-deployment-safety-view-of-vla-policies", "markdown": "https://wpnews.pro/news/same-weights-different-robot-a-deployment-safety-view-of-vla-policies.md", "text": "https://wpnews.pro/news/same-weights-different-robot-a-deployment-safety-view-of-vla-policies.txt", "jsonld": "https://wpnews.pro/news/same-weights-different-robot-a-deployment-safety-view-of-vla-policies.jsonld"}}