Expert Pyramid Tuning: Efficient Parameter Fine-Tuning for Expertise-Driven Task Allocation
This work addresses the problem of efficient and effective multi-task adaptation for large language models, offering a novel architecture that enhances task-specific feature capture, though it appears incremental by building on existing MoE-LoRA frameworks.
The paper tackles the limitation of existing Mixture-of-Experts LoRA methods in handling hierarchical task complexity by proposing Expert Pyramid Tuning (EPT), which integrates multi-scale feature pyramids into parameter-efficient fine-tuning, resulting in significant performance improvements over state-of-the-art variants while reducing training parameters.
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.