@article{tokenlevelpolicyoptimizationlinkinggroup, title = {Token-Level Policy Optimization: Linking Group-Level Rewards to Token-Level Aggregation via Sequence-Level Likelihood}, author = {Xingyu Lin and Yilin Wen and Du Su and Jinchang Hou and En Wang and Wenbin Liu and Chenfu Bao and Zhonghou Lv}, year = {2026}, eprint = {2604.12736}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2604.12736}, }