arXiv:2504.02263 - MegaScale-Infer: Serving Mixture-of-Experts at Scale with Disaggregated Expert Parallelism

arXiv:2504.02263, v1 2025-04-03, latest v4 2025-07-26. arXiv comment is EMPTY. Venue established from the OpenReview DBLP-mirrored note 'SIGCOMM 2025' / dblp.org/conf/SIGCOMM/2025. CORROBORATION CAVEAT, and it is the interesting part: that SIGCOMM note's title reads 'MegaScale-Infer: Efficient Mixture-of-Experts Model Serving with Disaggregated Expert Parallelism', which is the pre-print TITLE of the workshop-era version of this work, not the arXiv title. The identification rests on the shared method name, the identical topic, and the same author group (Xin Jin appears on both), not on a title string match - recorded here so a reader does not mistake this for a clean title-index hit. Authors Ruidong Zhu, Ziheng Jiang, Chao Jin, Peng Wu, Cesar A. Stuardo, Dongyang Wang, Xinlei Zhang, Huaping Zhou, Haoran Wei, Yang Cheng, Jianzhe Xiao, Xinyi Zhang, Lingjun Liu, Haibin Lin, Li-Wen Chang, Jianxi Ye, Xiao Yu, Xuanzhe Liu, Xin Jin, Xin Liu.