Learning-Augmented Online Allocation under Unreliable Advice: Robustness, Exposure Fairness, and Distribution Shift

The paper introduces a learning‑augmented algorithm for online allocation that handles unreliable predictions. It addresses finite candidate sets, irreversible decisions, and exposure constraints by combining advice with a conservative fallback and a fairness correction. The authors prove consistency and robustness under bounded‑error assumptions and demonstrate experimentally that the method remains stable against adversarial advice while substantially reducing exposure disparity.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

• safety