CoMoL: Efficient Mixture of LoRA Experts via Dynamic Core Space Merging
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915825498193920 |
|---|---|
| author | Cao, Jie Fan, Zhenxuan Wang, Zhuonan Lin, Tianwei Zhao, Ziyuan Yan, Rolan Zhang, Wenqiao Shao, Feifei Wang, Hongwei Xiao, Jun Tang, Siliang |
| author_facet | Cao, Jie Fan, Zhenxuan Wang, Zhuonan Lin, Tianwei Zhao, Ziyuan Yan, Rolan Zhang, Wenqiao Shao, Feifei Wang, Hongwei Xiao, Jun Tang, Siliang |
| contents | Large language models (LLMs) achieve remarkable performance on diverse downstream and domain-specific tasks via parameter-efficient fine-tuning (PEFT). However, existing PEFT methods, particularly MoE-LoRA architectures, suffer from limited parameter efficiency and coarse-grained adaptation due to the proliferation of LoRA experts and instance-level routing. To address these issues, we propose Core Space Mixture of LoRA (\textbf{CoMoL}), a novel MoE-LoRA framework that incorporates expert diversity, parameter efficiency, and fine-grained adaptation. Specifically, CoMoL introduces two key components: core space experts and core space routing. Core space experts store each expert in a compact core matrix, preserving diversity while controlling parameter growth. Core space routing dynamically selects and activates the appropriate core experts for each token, enabling fine-grained, input-adaptive routing. Activated core experts are then merged via a soft-merging strategy into a single core expert, which is combined with a shared LoRA to form a specialized LoRA module. Besides, the routing network is projected into the same low-rank space as the LoRA matrices, further reducing parameter overhead without compromising expressiveness. Extensive experiments demonstrate that CoMoL retains the adaptability of MoE-LoRA architectures while achieving parameter efficiency comparable to standard LoRA, consistently outperforming existing methods across multiple tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_00573 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | CoMoL: Efficient Mixture of LoRA Experts via Dynamic Core Space Merging Cao, Jie Fan, Zhenxuan Wang, Zhuonan Lin, Tianwei Zhao, Ziyuan Yan, Rolan Zhang, Wenqiao Shao, Feifei Wang, Hongwei Xiao, Jun Tang, Siliang Computation and Language Large language models (LLMs) achieve remarkable performance on diverse downstream and domain-specific tasks via parameter-efficient fine-tuning (PEFT). However, existing PEFT methods, particularly MoE-LoRA architectures, suffer from limited parameter efficiency and coarse-grained adaptation due to the proliferation of LoRA experts and instance-level routing. To address these issues, we propose Core Space Mixture of LoRA (\textbf{CoMoL}), a novel MoE-LoRA framework that incorporates expert diversity, parameter efficiency, and fine-grained adaptation. Specifically, CoMoL introduces two key components: core space experts and core space routing. Core space experts store each expert in a compact core matrix, preserving diversity while controlling parameter growth. Core space routing dynamically selects and activates the appropriate core experts for each token, enabling fine-grained, input-adaptive routing. Activated core experts are then merged via a soft-merging strategy into a single core expert, which is combined with a shared LoRA to form a specialized LoRA module. Besides, the routing network is projected into the same low-rank space as the LoRA matrices, further reducing parameter overhead without compromising expressiveness. Extensive experiments demonstrate that CoMoL retains the adaptability of MoE-LoRA architectures while achieving parameter efficiency comparable to standard LoRA, consistently outperforming existing methods across multiple tasks. |
| title | CoMoL: Efficient Mixture of LoRA Experts via Dynamic Core Space Merging |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2603.00573 |