Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Expert Pyramid Tuning: Efficient Parameter Fine-Tuning for Expertise-Driven Task Allocation

Jia-Chen Zhang, Zhen-Wei Yan, Yu-Jie Xiong, Chun-Ming Xia

Abstract

Parameter-Efficient Fine-Tuning (PEFT) has become a dominant paradigm for deploying LLMs in multi-task scenarios due to its extreme parameter efficiency. While Mixture-of-Experts (MoE) based LoRA variants have achieved promising results by dynamically routing tokens to different low-rank experts, they largely overlook the hierarchical nature of task complexity. Existing methods typically employ experts with uniform architectures, limiting their ability to capture diverse feature granularities required by distinct tasks--where some tasks demand high-level semantic abstraction while others require fine-grained syntactic manipulation. To bridge this gap, we propose Expert Pyramid Tuning (EPT), a novel architecture that integrates the multi-scale feature pyramid concept from computer vision into the realm of PEFT. Unlike standard LoRA, EPT decomposes task adaptation into two stages: (1) A shared meta-knowledge Subspace that encodes universal linguistic patterns in low dimensions; (2) A Pyramid Projection Mechanism that utilizes learnable up-projection operators to reconstruct high-dimensional features at varying scales. A task-aware router then dynamically selects the optimal combination of these multi-scale features. Extensive experiments across multiple multi-task benchmarks demonstrate that EPT significantly outperforms SOTA MoE-LoRA variants. Crucially, thanks to the re-parameterization capability of our design, EPT achieves this performance improvement while simultaneously reducing the number of training parameters.

Expert Pyramid Tuning: Efficient Parameter Fine-Tuning for Expertise-Driven Task Allocation

Abstract

Parameter-Efficient Fine-Tuning (PEFT) has become a dominant paradigm for deploying LLMs in multi-task scenarios due to its extreme parameter efficiency. While Mixture-of-Experts (MoE) based LoRA variants have achieved promising results by dynamically routing tokens to different low-rank experts, they largely overlook the hierarchical nature of task complexity. Existing methods typically employ experts with uniform architectures, limiting their ability to capture diverse feature granularities required by distinct tasks--where some tasks demand high-level semantic abstraction while others require fine-grained syntactic manipulation. To bridge this gap, we propose Expert Pyramid Tuning (EPT), a novel architecture that integrates the multi-scale feature pyramid concept from computer vision into the realm of PEFT. Unlike standard LoRA, EPT decomposes task adaptation into two stages: (1) A shared meta-knowledge Subspace that encodes universal linguistic patterns in low dimensions; (2) A Pyramid Projection Mechanism that utilizes learnable up-projection operators to reconstruct high-dimensional features at varying scales. A task-aware router then dynamically selects the optimal combination of these multi-scale features. Extensive experiments across multiple multi-task benchmarks demonstrate that EPT significantly outperforms SOTA MoE-LoRA variants. Crucially, thanks to the re-parameterization capability of our design, EPT achieves this performance improvement while simultaneously reducing the number of training parameters.
Paper Structure (23 sections, 11 equations, 2 figures, 8 tables)

This paper contains 23 sections, 11 equations, 2 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Parameter-Efficient Fine-Tuning
  4. Mixture of LoRA
  5. Method
  6. Expert Pyramid Tuning
  7. Adaptive LoRA Pruner
  8. Task Embedding
  9. Experiment
  10. Experimental Settings
  11. Datasets
  12. Overall Performance
  13. Expert Allocation Analysis
  14. Task Embeddings Analysis
  15. EPT Analysis
  16. ...and 8 more sections

Figures (2)

  • Figure 1: The overall framework of EPT. The overall architecture of EPT resembles a parameter pyramid, consisting of a shared meta-knowledge subspace and multiple deconvolutional projection layers of varying dimensions.
  • Figure 2: We present a visualization of the dataset analysis. Figure (a) shows the activation of each expert during the GLUE benchmark testing. Figure (b) displays the task embeddings from the final layer of the self-attention module.