arXivXiaojun Wu, Cehao Yang, Xuhui Jiang, Chengjin Xu, Jia Li, Jian GuoThu, Nov 5, 2026, 8:18 AM PST
score 16.9
New benchmark tests AI agents on full math modeling contests
Original: Terminal-style Mathematical Modeling Bench: The First Agentic and Verifiable Benchmark for Mathematical Modeling
Source: arxiv.org ↗
Writing ELI5 summary…