Reproduce “Test-Time Compute Games” 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
Ander Artola Velasco, Dimitrios Rontogiannis, Stratis Tsirtsis, Manuel Gomez-Rodriguez
The authors model competition among language-model providers choosing test-time compute under pay-for-compute pricing and analyze a reverse second-price auction. The simulated GSM8K market reaches equilibrium with price of anarchy 1.19, indicating welfare below the optimum. The auction raises modeled social welfare from $5.8×10⁻³ to $7.3×10⁻³. These results rely on estimated costs and prices and do not establish deployment performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.