Reproduce “this-that-model-1.0: A typed decision model that decides in 30 ms, for a millionth of a cent” 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
Zehua Cheng, Wei Dai, Jiahao Sun
The model scores caller-declared options from hidden states, producing probabilities without generating answer tokens or requiring parsing. It runs locally, with latency of 30.9 ms per question. On 68 recorded local-geometry questions, it reached 0.941 accuracy and 0.042 Brier score, versus Jev’s 0.765 and 0.133. Parity accuracy was 0.513, near chance, limiting the evidence to bounded decisions rather than search.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.