SuperPos-Prompt: Enhancing Soft Prompt Tuning of Language Models with Superposition of Multi Token Embeddings
This work addresses parameter-efficient tuning for language models, offering incremental improvements in soft prompt tuning methods.
The paper tackles the challenge of achieving optimal tuning with soft prompts for pretrained language models, particularly on smaller datasets, by introducing SuperPos-Prompt, a reparameterization technique using superposition of multiple token embeddings, which shows an average score increase of +6.4 in T5-Small and +5.0 in T5-Base on GLUE and SuperGLUE benchmarks, occasionally outperforming full fine-tuning.
Soft prompt tuning techniques have recently gained traction as an effective strategy for the parameter-efficient tuning of pretrained language models, particularly minimizing the required adjustment of model parameters. Despite their growing use, achieving optimal tuning with soft prompts, especially for smaller datasets, remains a substantial challenge. This study makes two contributions in this domain: (i) we introduce SuperPos-Prompt, a new reparameterization technique employing the superposition of multiple pretrained vocabulary embeddings to improve the learning of soft prompts. Our experiments across several GLUE and SuperGLUE benchmarks consistently highlight SuperPos-Prompt's superiority over Residual Prompt tuning, exhibiting an average score increase of $+6.4$ in T5-Small and $+5.0$ in T5-Base along with a faster convergence. Remarkably, SuperPos-Prompt occasionally outperforms even full fine-tuning methods. (ii) Additionally, we demonstrate enhanced performance and rapid convergence by omitting dropouts from the frozen network, yielding consistent improvements across various scenarios and tuning methods.