TY  - RPRT
TI  - Optimal Sparsity of Mixture-of-Experts Language Models for Reasoning Tasks
AU  - Taishi Nakamura
AU  - Satoki Ishikawa
AU  - Masaki Kawamura
AU  - Takumi Okamoto
AU  - Daisuke Nohara
AU  - Jun Suzuki
AU  - Rio Yokota
PY  - 2026
UR  - https://arxiv.org/abs/2508.18672
ID  - 2508.18672
ER  -