Research paper
Verifiable Foundation Models for Robot Safety
FEARL separates a large perception/reasoning controller from a smaller action-producing safety module with limited sensor and context inputs. Three simulated domains and one physical transfer test examine the architecture. Verification addresses the defined small module, leaving interface assumptions and overall robot hazards to assess.
Reviewed 2026-11-05 · Source and bibliography checked
- Original title
- Verifiable Foundation Models for Robot Safety
- Publisher
- arXiv
- Authors
- Corsi, Davide, Kim, Kyungmin, Fox, Roy
- DOI
- 10.48550/arxiv.2606.23754
- First archive posting
- 2026-06-22
- Source language
- en
- Source country
- US
- Edition
- v1
- Access
- Open archive metadata and manuscript
- Publication status
- Preprint
Related resources and engineering notes
Source overview
FEARL separates a large perception/reasoning controller from a smaller action-producing safety module with limited sensor and context inputs. Three simulated domains and one physical transfer test examine the architecture. Verification addresses the defined small module, leaving interface assumptions and overall robot hazards to assess.
Use and review
Original summary of the primary archive abstract and bibliographic record. Specialist technical and native-language review remain pending. The reported method does not establish a deployed safety function or achieved PL.
Sources & editions
Verifiable Foundation Models for Robot Safety ↗
arXiv · v1
arXiv abstract, authors, submission history and DOI links; first posting 2026-06-22, abstract reviewed in v1. Publisher dates are separate.