AlgoWorlds: Benchmarking Tool Use for Global Optimization in Algorithmic Worlds
AlgoWorlds is introduced, a benchmark that transforms formally specified combinatorial optimization problems into partially observed decision environments with verifiable global optima, and seven leading LLMs are evaluated, including Claude Opus 4.8 and GPT-5.6 Sol.
Zi-Xiang Xu, Jiaan Wang, Fanfei Meng
· 1 citation