Hot eventLive
BiToK-SD:双层Top-K令牌选择的自蒸馏方法
1 reports1 sources3 hr ago updated
Get the story
AI overview
2026年10月7日,研究者在 arXiv(Machine Learning Theory)发布 BiToK-SD(Bilevel Top-K Token Selection for Self-Distillation)。该方法基于双层优化,用于在 on-policy 自蒸馏中学习应在哪些位置施加蒸馏,通过双层 Top-K 令牌选择改进大语言模型的自蒸馏过程。目前报道仅给出方法提出与基本思路,未披露实验数据或进一步验证结果。
Generated from reports · updated 3 hr ago
Timeline
Follow the coverage from different angles.
Oct 7, 2026
- arXiv · Machine Learning TheoryBiToK-SD:用双层 Top-K token 选择改进大语言模型自蒸馏
研究者提出 BiToK-SD(Bilevel Top-K Token Selection for Self-Distillation),一种基于双层优化的 token 选择方法,用于在 on-policy 自蒸馏中学习应在哪些位置施加蒸馏。
Heat trend
Current heat 9·Comparable peak 10(Oct 7)·Comparable change over 24 hours –
The trend compares only the same participants observed continuously; its range may be smaller than the current heat count. Move or click on the chart to inspect hourly heat; use the left and right arrow keys to switch.