Reproduce “Adversarial Multiplicity: Cherry-Picked Models and the Limits of Disclosure” 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
Cassandra C. Chou, Benjamin Q. Huynh, Allison Koenecke, Elizabeth T. Chin
The authors show how motivated parties can manipulate high-stakes predictions through soft model selection or hard parameter editing. In selected tabular benchmarks, hard edits reversed nearly every validation individual’s decision, and on COMPAS soft selection reversed decisions for 38.4% of test defendants using Brier-score eligibility, falling to 22.9% with additional checks. These findings demonstrate capacity, not real-world prevalence.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.