Simultaneous LoRA gradient descent reaches ε-stationary points in O(ε^-4) full-gradient evaluations, and the new stochastic algorithms LoRA-NSGDM and LoRA-STORM achieve O(ε^-8) and O(ε^-6) oracle complexity.
Unless explicitly noted, the tasks, architectures, and LoRA configurations follow Mu and Klabjan (2026)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
On the Convergence of Stochastic Low-Rank Adaptation
Simultaneous LoRA gradient descent reaches ε-stationary points in O(ε^-4) full-gradient evaluations, and the new stochastic algorithms LoRA-NSGDM and LoRA-STORM achieve O(ε^-8) and O(ε^-6) oracle complexity.