Bayesian Conformal Prediction as a Decision Risk Problem
Fanyi Wu, Veronika Lohmanova, Samuel Kaski, Michele Caprio · 2026
Read Bayesian Conformal Prediction as a Decision Risk Problem by Fanyi Wu, Veronika Lohmanova, Samuel Kaski, Michele Caprio (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-2602-03331.txt.
Description
Prediction sets that can split into separate pieces are new by Fanyi Wu, with 3 co-authors, in a system called Bayesian Conformal Prediction. Standard conformal methods use one fixed threshold, which typically yields connected prediction sets. When the predictive spread has several separated modes, a connected set wastes space on low-density regions between modes, even though it still satisfies its coverage guarantee on average across repeated uses.
Their system formulates conformal prediction as a decision-risk tuning problem. Optimized highest posterior density sets can be disjoint, concentrating mass on separated high-density regions. A PAC-style risk constraint enforces validity even when the Bayesian model is misspecified. In a multimodal test, mean set size fell by more than half while meeting the target rate. Further tests cover regression, classification, and spread shift.
Catalog record
| CREATOR | Fanyi Wu, Veronika Lohmanova, Samuel Kaski, Michele Caprio |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-03331