CoMoE trains LoRA-based mixture-of-experts with a contrastive objective that pulls activated experts together and pushes inactivated experts apart, improving multi-task fine-tuning accuracy.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
method 1
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
REJECT 1roles
method 1polarities
unclear 1representative citing papers
citing papers explorer
-
CoMoE: Contrastive Representation for Mixture-of-Experts in Parameter-Efficient Fine-tuning
CoMoE trains LoRA-based mixture-of-experts with a contrastive objective that pulls activated experts together and pushes inactivated experts apart, improving multi-task fine-tuning accuracy.