A new study explores alternatives to LoRA, the most popular fine-tuning technique, assessing whether other methods can achieve better performance with less computational cost. The research finds that while some approaches show promise, none consistently outperform LoRA across diverse tasks and datasets.