HELLoRA: Hot Experts Layer-Level Low-Rank Adaptation for Mixture-of-Experts Models
HELLoRA is a parameter-efficient fine-tuning method for Mixture-of-Experts (MoE) models that attaches LoRA modules only to the most frequently activated experts per layer. It reduces trainable parameters and adapter FLOPs while improving downstream performance. Tested on OlMoE, Mixtral, and DeepSeekMoE across math, code, and safety tasks, HELLoRA significantly outperforms vanilla LoRA, e.g., using 15.7% of the parameters on OlMoE with 9.2% higher accuracy.
Article intelligence
Key points
- HELLoRA attaches LoRA only to the most active experts per layer in MoE models.
- It achieves superior performance with far fewer trainable parameters and FLOPs.
- On OlMoE, it uses 15.7% of vanilla LoRA's parameters and improves accuracy by 9.2%.
Why it matters
This matters because hELLoRA attaches LoRA only to the most active experts per layer in MoE models.
Technical impact
May affect model selection, inference cost, product capability, and evaluation benchmarks.
[2605.18795] HELLoRA: Hot Experts Layer-Level Low-Rank Adaptation for Mixture-of-Experts Models
[Submitted on 11 May 2026]
Title:HELLoRA: Hot Experts Layer-Level Low-Rank Adaptation for Mixture-of-Experts Models
View a PDF of the paper titled HELLoRA: Hot Experts Layer-Level Low-Rank Adaptation for Mixture-of-Experts Models, by Jia Wei and 7 other authors
View PDF HTML (experimental)
Abstract:Low-Rank Adaptation (LoRA) dominates parameter-efficient fine-tuning of large language models, yet most variants target dense architectures. Mixture-of-Experts (MoE) models scale parameters at near-constant per-token compute, and their sparse activation patterns create untapped opportunities for more efficient adaptation. We propose Hot-Experts Layer-level Low-Rank Adaptation (HELLoRA), which attaches LoRA modules only to the most frequently activated experts at each layer. This simple mechanism reduces trainable parameters and adapter-induced FLOPs while improving downstream performance, an effect we attribute to a form of structured regularization that preserves pretrained expert specialization. To stress-test HELLoRA under extreme parameter budgets, we further compose it with LoRI to form HELLoRI, which freezes the up-projection and sparsifies the down-projection. Across three MoE backbones, namely OlMoE-1B-7B, Mixtral-8x7B, and DeepSeekMoE, and three task families covering mathematical reasoning, code generation, and safety alignment, HELLoRA consistently outperforms strong PEFT baselines. Relative to vanilla LoRA on OlMoE, HELLoRA uses 15.7% of the trainable parameters, reduces adapter FLOPs by 38.7%, achieves 1.9x the training throughput, and improves accuracy by 9.2%. On DeepSeekMoE, HELLoRA outperforms LoRA while using only 23.2% of its trainable parameters. These results demonstrate that activation-aware adapter placement is an effective and practical route to scaling PEFT for MoE language models.
Subjects:
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
Cite as: arXiv:2605.18795 [cs.LG]
(or arXiv:2605.18795v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2605.18795
arXiv-issued DOI via DataCite
Submission history
From: Jia Wei [view email] [v1] Mon, 11 May 2026 06:43:14 UTC (821 KB)
Full-text links:
Access Paper:
View a PDF of the paper titled HELLoRA: Hot Experts Layer-Level Low-Rank Adaptation for Mixture-of-Experts Models, by Jia Wei and 7 other authors
View PDF
HTML (experimental)
TeX Source
view license
Current browse context:
cs.LG
new | recent | 2026-05
Change to browse by:
cs cs.AI
References & Citations
NASA ADS
Google Scholar
Semantic Scholar
Loading...
Data provided by:
Bibliographic Tools
Bibliographic and Citation Tools
Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media
Code, Data and Media Associated with this Article
alphaXiv Toggle
alphaXiv (What is alphaXiv?)
Links to Code Toggle
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub Toggle
DagsHub (What is DagsHub?)
GotitPub Toggle
Gotit.pub (What is GotitPub?)
Huggingface Toggle
Hugging Face (What is Huggingface?)
ScienceCast Toggle
ScienceCast (What is ScienceCast?)
Demos
Demos
Replicate Toggle
Replicate (What is Replicate?)
Spaces Toggle
Hugging Face Spaces (What is Spaces?)
Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
Related Papers
Recommenders and Search Tools
Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
IArxiv recommender toggle
IArxiv Recommender (What is IArxiv?)
Author
Venue
Institution
Topic
About arXivLabs
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)