Reproduce “AI Behavioral Evaluation Should Be Grounded in Psychophysics Across Marr’s Levels” 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
Jonas Leo Mueller, Bernhard Egger, Bjoern M. Eskofier, Pasquale Minervini, Antonio Rizzo, Marco Valentino, Dario Zanca
The authors propose a framework for evaluating AI systems that separates observed behavior, internal representations, and physical implementation. It adapts psychophysics’ controlled stimulus–response method and organizes claims using Marr’s three levels. It non-exhaustively classifies 21 representative paradigms. The framework is a position paper, not an experimental validation; its central warning is that outputs alone do not establish internal explanations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.