@article{rethinkingimportancesamplinginllmpolicyo, title = {Rethinking Importance Sampling in LLM Policy Optimization: A Cumulative Token Perspective}, author = {Yuheng Zhang and Chenlu Ye and Shuowei Jin and Changlong Yu and Wei Xiong and Saurabh Sahu and Nan Jiang}, year = {2026}, eprint = {2605.07331}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2605.07331}, }