When Does Trimming Help Conformal Prediction? A Retained-Law Diagnostic under Calibration Contamination
Congye Wang · 2026
Read When Does Trimming Help Conformal Prediction? A Retained-Law Diagnostic under Calibration Contamination by Congye Wang (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-2605-06204.txt.
Introduction
Congye Wang asks when trimming suspicious calibration points actually helps conformal prediction. Conformal methods give prediction sets with coverage guarantees, but contaminated calibration data can distort them, and trimming looks like an obvious cleanup. Wang shows the effect runs through the law of the retained points, not through the taint level alone, which reframes trimming as conditioning rather than cleanup.
An exact finite sample identity reduces clean target coverage after trimming to a one dimensional transfer problem on score distributions. A piecewise bound splits the cost into a clean side covariance term and a retained taint term. Trimming helps when its score separates retention chances while staying neutral on clean data. Otherwise it cannot cut taint much, and the paper offers certificate templates for independent audits.
Catalog record
| CREATOR | Congye Wang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 154 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.06204