GiVA: Gradient-Informed Bases for Vector-Based Adaptation

TL;DR

GiVA is a gradient-informed initialization method for vector-based adaptation that reduces rank requirements and training time, matching or surpassing LoRA's performance across multiple benchmarks.

Contribution

Introduces GiVA, a novel gradient-based initialization strategy that enhances vector-based adaptation efficiency and performance, reducing rank needs by a factor of eight.

Findings

GiVA achieves training times comparable to LoRA.

GiVA reduces rank requirements by a factor of eight.

GiVA outperforms or matches existing vector-based methods and LoRA.

Abstract

As model sizes continue to grow, parameter-efficient fine-tuning has emerged as a powerful alternative to full fine-tuning. While LoRA is widely adopted among these methods, recent research has explored vector-based adaptation methods due to their extreme parameter efficiency. However, these methods typically require substantially higher ranks than LoRA to match its performance, leading to increased training costs. This work introduces GiVA, a gradient-based initialization strategy for vector-based adaptation. It achieves training times comparable to LoRA and maintains the extreme parameter efficiency of vector-based adaptation. We evaluate GiVA across diverse benchmarks, including natural language understanding, natural language generation, and image classification. Experiments show that our approach consistently outperforms or achieves performance competitive with existing vector-based adaptation methods and LoRA while reducing rank requirements by a factor of eight ($8\times$).