Theoretical analysis demonstrates constant competitive ratios for dynamic mixture-of-experts serving, highlighting optimal bounds for online model inference.
We prove a factor-16 competitive guarantee for a deterministic fractional algorithm for Dynamic MoE Serving and, against an oblivious adversary, a randomized factor-48 guarantee for the original integral problem. Thus the optimal randomized competitive ratio is Theta(1) under the stated algebraic real-RAM model. This record contains the rendered preprint and its complete LaTeX source.
No takes yet. Share an insight, caveat, or question.
Thomas Vergeres (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: