Parameter Efficient Mamba Tuning via Projector-targeted Diagonal-centric Linear Transformation
This addresses the problem of efficient adaptation of Mamba models for new tasks, offering a domain-specific solution that is incremental but novel for this architecture.
The paper tackles parameter-efficient fine-tuning for Mamba architecture by discovering that Projectors, not state-space models, are the primary contributors to transfer learning, and proposes ProDiaL, which optimizes only pretrained Projectors using diagonal-centric linear transformations, achieving strong performance with less than 1% of total parameters across vision and language models.
Despite the growing interest in Mamba architecture as a potential replacement for Transformer architecture, parameter-efficient fine-tuning (PEFT) approaches for Mamba remain largely unexplored. In our study, we introduce two key insights-driven strategies for PEFT in Mamba architecture: (1) While state-space models (SSMs) have been regarded as the cornerstone of Mamba architecture, then expected to play a primary role in transfer learning, our findings reveal that Projectors -- not SSMs -- are the predominant contributors to transfer learning. (2) Based on our observation, we propose a novel PEFT method specialized to Mamba architecture: Projector-targeted Diagonal-centric Linear Transformation (ProDiaL). ProDiaL focuses on optimizing only the pretrained Projectors for new tasks through diagonal-centric linear transformation matrices, without directly fine-tuning the Projector weights. This targeted approach allows efficient task adaptation, utilizing less than 1% of the total parameters, and exhibits strong performance across both vision and language Mamba models, highlighting its versatility and effectiveness.