Can discrete information extraction prompts generalize across language models?

TL;DR

This paper investigates whether discrete prompts induced for one language model can generalize to others, proposing a training method that enhances cross-model transferability and analyzing properties of such prompts.

Contribution

It introduces a training approach mixing models to induce prompts that generalize across models and analyzes their linguistic and structural properties.

Findings

AutoPrompt prompts outperform manual prompts on slot-filling

Performance drops when prompts are transferred between models

Mixing models during training improves prompt generalization

Abstract

We study whether automatically-induced prompts that effectively extract information from a language model can also be used, out-of-the-box, to probe other language models for the same information. After confirming that discrete prompts induced with the AutoPrompt algorithm outperform manual and semi-manual prompts on the slot-filling task, we demonstrate a drop in performance for AutoPrompt prompts learned on a model and tested on another. We introduce a way to induce prompts by mixing language models at training time that results in prompts that generalize well across models. We conduct an extensive analysis of the induced prompts, finding that the more general prompts include a larger proportion of existing English words and have a less order-dependent and more uniform distribution of information across their component tokens. Our work provides preliminary evidence that it's possible to generate discrete prompts that can be induced once and used with a number of different models, and gives insights on the properties characterizing such prompts.