BoRA: Bi-dimensional Weight-Decomposed Low-Rank Adaptation
This is an incremental improvement for parameter-efficient fine-tuning of large-scale pre-trained models, addressing a specific limitation in existing methods.
The paper tackled the asymmetry in weight-decomposed low-rank adaptation by introducing BoRA, which symmetrically adjusts column-wise and row-wise magnitudes, achieving superior results over state-of-the-art methods like LoRA and DoRA across various benchmarks.
In recent years, Parameter-Efficient Fine-Tuning (PEFT) methods like Low-Rank Adaptation (LoRA) have significantly enhanced the adaptability of large-scale pre-trained models. Weight-Decomposed Low-Rank Adaptation (DoRA) improves upon LoRA by separating the magnitude and direction components of the weight matrix, leading to superior performance. However, DoRA's improvements are limited to the vertical dimension, resulting in an asymmetrical pattern between horizontal and vertical dimensions. This paper introduces BoRA, an innovative extension of LoRA and DoRA, characterized by symmetrical properties across horizontal and vertical dimensions. Our approach optimizes the weight matrix symmetrically by adjusting both column-wise and row-wise magnitudes. Extensive experiments demonstrate that BoRA surpasses state-of-the-art PEFT methods, including LoRA and DoRA, achieving superior results across various benchmarks.