#KL正则化

2 stories · 1 in 7 days · 1 in 30 days

Stories, products and papers tagged "KL正则化", newest first.

Nov 11

搞 RLHF 或最大熵 RL 的朋友可以看这篇。以前理论界是 H^5，现在证明 H^3 就够，还首次给出带温度 β 的下界。做 RLHF 训练时要估采样预算的话，这篇能直接告诉你该按什么规模分配样本。

arXiv cs.LG@Andrew LiuSource

Aug 28

PapersSelected73°19:28

LAMA offers a new advertising model for generative AI platforms that could increase revenue while preserving user experience.

arXiv cs.LG@Hanbing Liu et al. (5)Source