Tag
The paper proposes FrameFT, a parameter-efficient fine-tuning method using sparse coefficients in a Fusion Frame basis, reducing memory footprint while achieving performance on par with or exceeding state-of-the-art PEFT techniques.