# Claim 3 — theoretical guarantees

## Official claim

> Machine-learned dual predictions enable improved theoretical guarantees over standard algorithms for online minimization problems.

## Recorded source anchors

- \label{thm:laminarsetcover}
- \label{thm:MTSthm}
- \label{lem:lemdualboundlaminar}
- \label{th:MTSBoundForBellmanError}
- \label{thm:supertheorem}

The pinned source crosswalk records these anchors for the stated
prediction-error-dependent guarantees.

## Limit

Status is **unverified**. Any interpretation of “improved” depends on the
authors' assumptions, comparators, and prediction-error regime. This package
does not validate theorems or reproduce experiments.