fastcurl-curriculum-rl-efficient-reasoning

【数据与研究】Train efficient reasoning models using stage-wise context scaling and complexity-aware data selection. Achieves 49.6% accuracy on AIME 2024 while reducing training steps by 50% through alternating compress-extend cycles that progressively refine reasoning quality.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准