PolyMath-en: leaderboard

English split of PolyMath (Qwen Team, Alibaba, 2025): 500 math problems in 4 difficulty tiers of 125 each; numeric or symbolic final answers scored by difficulty-weighted accuracy.

Source: arxiv.org.

Interactive version: theaggregate.ai/benchmark?slug=polymath-en · How It Works · Data refreshed daily, snapshot 2026-09-05.