1 paper · 1 filter
Lingkai Kong, Zijian Wu, Yuzhe Gu +11
Large language models (LLMs) have achieved remarkable performance on high-school and competition-level mathematics, yet their capabilities on advanced mathematics remain poorly und…