Reproduce “Macro Mondrian Conformal Risk Control” 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
Julien Combes, Jean-Francois Coeurjolly, Alexandre Dervill{é}
The authors introduce Macro-CRC to control a user-weighted average of segmentation risks across chosen groups. It reweights calibration losses by group importance when selecting a shared threshold. On polyp segmentation at target risk 0.10, weighted loss was 0.089 ± 0.026, versus 0.197 ± 0.027 for standard CRC; the guarantee does not cover each group’s risk individually.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.