Reproduce “A Unifying Post-Processing Framework for Multi-Objective Learn-to-Defer Problems” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Mohammad-Amin Charusaie, Samira Samadi

The authors present d-GNP, a post-processing framework that jointly selects a model’s prediction rule and when to defer to a human, optimizing performance under constraints such as fairness or expert workload. It combines objective and constraint scores to choose actions, allowing randomization. In experiments on COMPAS and ACSIncome, it controls constraint violations, but does not guarantee higher accuracy or establish deployment effectiveness.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.