HELLoRA: Hot Experts Layer-Level Low-Rank Adaptation for Mixture-of-Experts Models

• URL: http://arxiv.org/abs/2605.18795v1
• Date: Mon, 11 May 2026 06:43:14 GMT
• Title: HELLoRA: Hot Experts Layer-Level Low-Rank Adaptation for Mixture-of-Experts Models
• Abstract summary: Low-Rank Adaptation (LoRA) dominates parameter-efficient fine-tuning of large language models.<n>We propose Hot-Experts Layer-level Low-Rank Adaptation (HELLoRA)<n>HELLoRA attaches LoRA modules only to the most frequently activated experts at each layer.
• Score: 11.336343540603616
• License: http://creativecommons.org/licenses/by/4.0/
• 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.

Related papers

• Taming Momentum: Rethinking Optimizer States Through Low-Rank Approximation [85.89510825889168]
  We introduce LoRA-Pre, a novel low-rank system for efficient pre-training.<n>LoRA-Pre decomposing the momentum matrix into a compact low-rank subspace within the online linear learner.<n>We empirically validate LoRA-Pre's efficacy by pre-training models from the Llama architecture family.
  arXiv  Detail & Related papers  (2026-02-27T18:57:06Z)
• ARD-LoRA: Dynamic Rank Allocation for Parameter-Efficient Fine-Tuning of Foundation Models with Heterogeneous Adaptation Needs [0.46040036610482665]
  This paper introduces Adaptive Rank Dynamic LoRA (ARD-LoRA), a novel framework that automates rank allocation through learnable scaling factors.<n>ARD-LoRA enables continuous, differentiable, per-head rank adaptation.<n> Experiments on LLAMA-3.1-70B and PaliGemma-2 demonstrate ARD-LoRA's efficacy, achieving up to 99.3% of full fine-tuning performance with only 0.32% trainable parameters.
  arXiv  Detail & Related papers  (2025-06-23T03:45:37Z)
• LoRA-Mixer: Coordinate Modular LoRA Experts Through Serial Attention Routing [17.171872354057694]
  We propose the LoRA-Mixer, a modular and lightweight MoE framework that integrates LoRA experts.<n>Our core innovation lies in replacing the projection matrices of the attention module's input/output linear layers with task-specific LoRA experts.<n>LoRA-Mixer achieves significant improvements on datasets such as GSM8K, HumanEval, and MedQA.
  arXiv  Detail & Related papers  (2025-06-17T14:58:54Z)
• DenseLoRA: Dense Low-Rank Adaptation of Large Language Models [14.133511131962786]
  Low-rank adaptation (LoRA) has been developed as an efficient approach for adapting large language models (LLMs)<n>We introduce Dense Low-Rank Adaptation (DenseLoRA), a novel approach that enhances parameter efficiency while achieving superior performance compared to LoRA.<n>We evaluate DenseLoRA on various benchmarks, showing that it achieves 83.8% accuracy with only 0.01% of trainable parameters, compared to LoRA's 80.8% accuracy with 0.70% of trainable parameters on LLaMA3-8B.
  arXiv  Detail & Related papers  (2025-05-27T08:19:07Z)
• BeamLoRA: Beam-Constraint Low-Rank Adaptation [51.52097743781401]
  Low-Rank Adaptation (LoRA) has been widely adopted as one of the most effective parameter-efficient fine-tuning methods.<n>We propose BeamLoRA, which conceptualizes each LoRA module as a beam where each rank naturally corresponds to a potential sub-solution.
  arXiv  Detail & Related papers  (2025-02-19T10:33:22Z)
• EDoRA: Efficient Weight-Decomposed Low-Rank Adaptation via Singular Value Decomposition [2.5269004336032186]
  Efficient Weight-Decomposed Low-Rank Adaptation (EDoRA) is a novel PEFT method that decomposes pre-trained weights into magnitude and directional components.<n>EDoRA achieves competitive or superior performance compared to state-of-the-art methods, such as LoRA and DoRA.
  arXiv  Detail & Related papers  (2025-01-21T11:42:09Z)
• Less is More: Extreme Gradient Boost Rank-1 Adaption for Efficient Finetuning of LLMs [75.11449420928139]
  Fine-tuning Large Language Models (LLMs) has become a crucial technique for adapting pre-trained models to downstream tasks. Low-Rank Adaptation (LoRA) has emerged as a promising solution, but there exists a gap between the practical performance of low-rank adaptations and its theoretical optimum. We propose eXtreme Gradient Boosting LoRA, a novel framework that bridges this gap by leveraging the power of ensemble learning.
  arXiv  Detail & Related papers  (2024-10-25T17:07:13Z)
• MLAE: Masked LoRA Experts for Visual Parameter-Efficient Fine-Tuning [45.93128932828256]
  Masked LoRA Experts (MLAE) is an innovative approach that applies the concept of masking to visual PEFT. Our method incorporates a cellular decomposition strategy that transforms a low-rank matrix into independent rank-1 submatrices. We show that MLAE achieves new state-of-the-art (SOTA) performance with an average accuracy score of 78.8% on the VTAB-1k benchmark and 90.9% on the FGVC benchmark.
  arXiv  Detail & Related papers  (2024-05-29T08:57:23Z)
• MELoRA: Mini-Ensemble Low-Rank Adapters for Parameter-Efficient Fine-Tuning [71.50432879573614]
  Low-rank adaptation (LoRA) is based on the idea that the adaptation process is intrinsically low-dimensional.<n>We present MELoRA, a mini-ensemble low-rank adapters that uses fewer trainable parameters while maintaining a higher rank.<n>Our experimental results show that, compared to LoRA, MELoRA achieves better performance with 8 times fewer trainable parameters on natural language understanding tasks and 36 times fewer trainable parameters on instruction following tasks.
  arXiv  Detail & Related papers  (2024-02-27T07:14:12Z)
• Sparse Low-rank Adaptation of Pre-trained Language Models [79.74094517030035]
  We introduce sparse low-rank adaptation (SoRA) that enables dynamic adjustments to the intrinsic rank during the adaptation process. Our approach strengthens the representation power of LoRA by initializing it with a higher rank, while efficiently taming a temporarily increased number of parameters. Our experimental results demonstrate that SoRA can outperform other baselines even with 70% retained parameters and 70% training time.
  arXiv  Detail & Related papers  (2023-11-20T11:56:25Z)
• LoRA: Low-Rank Adaptation of Large Language Models [71.75808607987281]
  Low-Rank Adaptation, or LoRA, freezes the pre-trained model weights and injects trainable rank decomposition into each layer of the Transformer architecture. For GPT-3, LoRA can reduce the number of trainable parameters by 10,000 times and the computation hardware requirement by 3 times compared to full fine-tuning.
  arXiv  Detail & Related papers  (2021-06-17T17:37:18Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.