Reproduce “Empirical Bayes Conformal Prediction for Vision and Language Models” 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
Jiapeng Zeng, Yogesh Prabhu, Zhanpeng Zeng, Michael A. Newton, Vikas Singh
The authors introduce r-value conformal prediction, which ranks candidate labels using scores and score variability before calibration. It aims to retain marginal coverage under exchangeability while producing smaller sets when variability is informative. On ImageNet with SigLIP2-B/32 at nominal 95% coverage, it reports mean set size 13.0 versus 19.6 for standard C-P, with 94.9% coverage; gains are setting-dependent.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.