arXivArya Tschand, Yaosheng Fu, Vikram Sharma Mailthody, Nicolai Oswald, Po-An Tsai, Ritchie Zhao, Oreste Villa, Vijay Janapa Reddi, Karu SankaralingamFri, Sep 11, 2026, 10:55 AM PDT
score 15.4
New serving scheme splits AI inference work across chips by attention type
Original: Rethinking Heterogeneous System Disaggregation for Subquadratic Attention
Source: arxiv.org ↗
Writing ELI5 summary…