The paper derives convergence rates for sigmoid gating mixture-of-experts with quadratic scores and uses them to argue sigmoid self-attention is more sample-efficient than softmax, but the link to attention is an unproven analogy.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Sigmoid Self-Attention has Lower Sample Complexity than Softmax Self-Attention: A Mixture-of-Experts Perspective
The paper derives convergence rates for sigmoid gating mixture-of-experts with quadratic scores and uses them to argue sigmoid self-attention is more sample-efficient than softmax, but the link to attention is an unproven analogy.