Reproduce “TIRE: Top-Down Formula Instantiation with Reliable Execution for Physics Problem Synthesis” 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

Xiaoyu Wu, Zhaoxin Yu, Xinglin Xiao, Nan Xu

TIRE constructs physics training data from reusable formulas and their applicability conditions. It synthesizes and tests programs, executes admissible parameter assignments to establish reference answers, then generates questions and keeps teacher traces matching those answers. On Qwen3-8B, TIRE-19k reaches 66.56% average accuracy, 2.86 points above the strongest baseline in equal-sized comparisons; SciBench is in-domain.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.