Uncertainty-aware Latent Safety Filters for Avoiding Out-of-Distribution Failures
Junwon Seo, Kensuke Nakamura, Andrea Bajcsy · 2025
Read Uncertainty-aware Latent Safety Filters for Avoiding Out-of-Distribution Failures by Junwon Seo, Kensuke Nakamura, Andrea Bajcsy (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-00779.txt.
Description
Junwon Seo, Kensuke Nakamura, and Andrea Bajcsy build safety filters that heed doubt in two thousand twenty-five. World models let classic safe control work from raw sensor views. But training can never cover all risky scenes. A filter on a flawed model may miss new hazards and call them safe.
Their filter uses the model's own doubt as a sign of the unseen. A threshold for doubt is set with a formal method that gives coverage bounds. Reach tests then run in a space that holds both state and doubt. Tests in simulation and on a real robot arm show early warnings. The filter steers the robot to safe, known acts.
Catalog record
| CREATOR | Junwon Seo, Kensuke Nakamura, Andrea Bajcsy |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 112 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-00779