Distill-Belief: Closed-Loop Inverse Source Localization and Characterization in Physical Fields Conference: The 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining; August 9–13, 2026; International Convention Center Jeju (ICC Jeju), Jeju, Republic of Korea ISBN: 978-1-4503-XXXX-X/2018/06 CCS: Do Not Use This Code Generate the Correct Terms for Your Paper CCS: Do Not Use This Code Generate the Correct Terms for Your Paper CCS: Do Not Use This Code Generate the Correct Terms for Your Paper CCS: Do Not Use This Code Generate the Correct Terms for Your Paper
Yiwei Shi 1 , Zixing Song 1 , Mengyue Yang 1 , Cunjia Liu 2 , Weiru Liu 1 · 2026
Read Distill-Belief: Closed-Loop Inverse Source Localization and Characterization in Physical Fields Conference: The 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining; August 9–13, 2026; International Convention Center Jeju (ICC Jeju), Jeju, Republic of Korea ISBN: 978-1-4503-XXXX-X/2018/06 CCS: Do Not Use This Code Generate the Correct Terms for Your Paper CCS: Do Not Use This Code Generate the Correct Terms for Your Paper CCS: Do Not Use This Code Generate the Correct Terms for Your Paper CCS: Do Not Use This Code Generate the Correct Terms for Your Paper by Yiwei Shi 1 , Zixing Song 1 , Mengyue Yang 1 , Cunjia Liu 2 , Weiru Liu 1 (2026) 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-2604-26095.txt.
Description
Shi and four colleagues solve a belief dilemma in closed-loop localization. Exact Bayesian uncertainty estimates are expensive, while fast learned belief models invite reward hacking. In reward hacking, a policy exploits approximation errors instead of reducing uncertainty. Their Distill-Belief framework splits the roles. A Bayes-correct particle-filter teacher supplies a gain signal, while a student distills the posterior into control numbers and a stop check. At deployment only the student runs, at constant cost per step. Tests across seven modalities show lower sensing cost, better success.
Catalog record
| CREATOR | Yiwei Shi 1 , Zixing Song 1 , Mengyue Yang 1 , Cunjia Liu 2 , Weiru Liu 1 |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 148520 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.26095