scale-selective-test-time

【开发与工程】Decomposes math problems into sequential sub-problems, assesses difficulty, and allocates simple ones to fast System 1 reasoning while directing complex ones to deliberate System 2. Save 33-53% tokens while improving accuracy by up to 13.75 points on AIME.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准