Continuous Prompt Generation from Linear Combination of Discrete Prompt Embeddings
This addresses interpretability issues in continuous prompts for sensitive tasks like resume screening, but it is incremental as it builds on existing prompt-based methods.
The paper tackles the problem of unpredictable behavior in continuous prompts for large language models by constructing them from linear combinations of discrete prompt embeddings, resulting in improved interpretability and inference accuracy on natural language understanding tasks.
The wayward quality of continuous prompts stresses the importance of their interpretability as unexpected and unpredictable behaviors appear following training, especially in the context of large language models automating people-sensitive tasks such as resume screening. In this paper we present a novel method of constructing continuous prompts via discrete prompt embeddings and evaluate improvements to continuous prompt interpretability and inference accuracy. For a set of manually designed discrete prompts $\mathcal{D}$, which we tokenize and embed each into tensor form, we train a model to predict the weights such that the linear combinations of those prompts correspond to higher performance on natural language understanding tasks.