Reproduce “The Plausibility Trap: Anti-Informative Review Beyond an Unlocatable Frontier, and a Pre-Registered Null” 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
Lionel Arce
The authors model when accepting a plausible AI-generated claim can become evidence against its truth: under equal-variance signals and a fixed rule favoring higher plausibility, acceptance has a likelihood ratio below one beyond a reviewer–generator frontier. In a preregistered reanalysis of public open-answer mathematics data, however, all 135 verifier/tier/difficulty cells remained above one; the frontier’s location in systems remains unknown.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.