Reproduce “Beyond Prediction Residuals: Conformal Regret Calibration for Robust Contextual Linear Optimization” 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

Lingjie Zhao, Hansheng Jiang, Wei Qi

The authors use conformal calibration of an SPO+ upper bound on decision regret to refine decisions in contextual linear optimization. With an independently trained predictor and exchangeable data, the finite-sample guarantee is marginal, not per-context. In an n=20 fractional-knapsack test, continuous refinement lowers Q95 from 38.664 to 35.137, while mean regret rises from 7.605 to 11.117. These are empirical results.

Summary by alphaXiv.

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