Enabling Token-wise Importance Sampling in Off-Policy RLVR with Proximal KL Regularization

By Hao Bai, Jiarui Yao, Tong Zhang

· HF Daily Papers · Oct 13, 2026