Reproduce “The Metanym Game: An LLM Benchmark Without Ground Truth That Rises With the Models It Measures” 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
David Nordfors
The Metanym Game creates a benchmark where models write slot-based descriptions meant to remain true across domains, then judge one another’s work. Making claims checkable sentence by sentence lets the system score work without an external answer key. Its score correlates with G-P-Q-A accuracy across 12 models (Pearson r = 0.98), but this does not show the tests measure the same mechanism.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.