@misc{indiciaeee6d956ddc7f,
  title = {QUADS: Stabilizing NVFP4 Reinforcement Learning for MoE via QUantization-error Alignment across Dual Sides},
  author = {Zhengyang Zhuge and Hao Yu and Xin Wang and Zheng Li and Yizhong Cao and Dayiheng Liu and Jianwei Zhang},
  year = {2026},
  url = {https://arxiv.org/abs/2607.15810},
  note = {Source identifier: 2607.15810}
}