Inducer-tuning: Connecting Prefix-tuning and Adapter-tuning

Prefix-tuning, or more generally continuous prompt tuning, has become an essential paradigm of parameter-efficient transfer learning. Using a large pre-trained language model (PLM), prefix-tuning can obtain strong performance by training only a small portion of parameters. In this paper, we propose to understand and further develop prefix-tuning through the kernel lens. Specifically, we make an analogy between $\mathit{\text{prefixes}}$ and $\mathit{\text{inducing}}\mathit{\text{variables}}$ in kernel methods and hypothesize that $\mathit{\text{prefixes}}$ serving as $\mathit{\text{inducing}}\mathit{\text{variables}}$ would improve their overall mechanism. From the kernel estimator perspective, we suggest a new variant of prefix-tuning – $\mathit{\text{inducer-tuning}}$, which shares the exact mechanism as prefix-tuning while leveraging the residual form found in adapter-tuning. This mitigates the initialization issue in prefix-tuning. Through comprehensive empirical experiments on natural language understanding and generation tasks, we demonstrate that inducer-tuning can close the performance gap between prefix-tuning and fine-tuning.