EPnG reallocates LoRA capacity in MoE models by pruning experts with low router gate probabilities and expanding high-importance ones via rank growth, outperforming standard LoRA and nearing full fine-tuning performance with 0.55-0.72% parameters updated.
Title resolution pending
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.LG 2years
2026 2verdicts
UNVERDICTED 2representative citing papers
ProbMoE frames MoE routing as probabilistic inference over cardinality-constrained subsets, enabling Exact-k sampling with marginal-probability gradients and a dynamic-k variant that matches training and inference cardinality.
citing papers explorer
-
EPnG: Adaptive Expert Prune-and-Grow for Parameter-Efficient MoE Fine-tuning
EPnG reallocates LoRA capacity in MoE models by pruning experts with low router gate probabilities and expanding high-importance ones via rank growth, outperforming standard LoRA and nearing full fine-tuning performance with 0.55-0.72% parameters updated.
-
ProbMoE: Differentiable Probabilistic Routing for Mixture-of-Experts
ProbMoE frames MoE routing as probabilistic inference over cardinality-constrained subsets, enabling Exact-k sampling with marginal-probability gradients and a dynamic-k variant that matches training and inference cardinality.