Learning More by Training Less: Uncertainty-Guided Progressive Contraction for Efficient Reasoning Post-Training ▲ 6 💬 1 2026-10-14 arXiv HF 原文 GitHub ★5 摘要 （无摘要）