Learning to Translate from Soft to Hard LLM Prompts
For practitioners using soft prompt tuning, this method improves interpretability and enables portability of prompts to larger models, but the approach is incremental.
This work trains a model to translate soft prompts into natural language, achieving fluent and accurate verbalizations that outperform existing training-free methods like InSPEcT. The translated prompts, when deployed on larger closed-API models, exceed the performance of the original soft prompt and sometimes even few-shot learning.
Soft prompt tuning is a parameter-efficient method for adapting LLMs to specific tasks, but suffers from a lack of interpretability. Building on recent work on interpreting soft prompts (Ramati et al., 2024), we explore how training a dedicated soft prompt to natural language translation model can yield higher translation quality. In particular, in both quantitative and qualitative comparisons on multiple Datasets of Datasets (DoDs), we demonstrate that our translator produces fluent, accurate verbalizations that outperforms existing training-free methods like InSPEcT. In addition to advancing interpretability, our work suggests a promising downstream application: soft prompts optimized on small, open-source models can be translated into portable text prompts that, when deployed on larger closed-API models, exceed the performance of the original soft prompt and, in some cases, even few-shot learning.