@misc{indiciae29fa1b61b3d1, title = {See Every Block Once: Variance-Reduced Post-Training for Masked Diffusion Language Models}, author = {Yunzhen He and Hiroaki Yamagiwa and Hidetoshi Shimodaira}, year = {2026}, url = {https://searcharxiv.com/abs/2611.17754}, note = {Source identifier: 2611.17754} }