Effective Structured Prompting by Meta-Learning and Representative Verbalizer
This work addresses efficiency and performance issues in few-shot NLP tasks, offering a parameter-efficient solution that is incremental over prior meta-learning approaches.
The paper tackles the challenge of prompt tuning for masked language models in few-shot learning by introducing MetaPrompter, which uses a prompt pool and a novel soft verbalizer to improve performance and efficiency. Experimental results show MetaPrompter outperforms recent state-of-the-art methods and the verbalizer surpasses existing soft verbalizers.
Prompt tuning for pre-trained masked language models (MLM) has shown promising performance in natural language processing tasks with few labeled examples. It tunes a prompt for the downstream task, and a verbalizer is used to bridge the predicted token and label prediction. Due to the limited training data, prompt initialization is crucial for prompt tuning. Recently, MetaPrompting (Hou et al., 2022) uses meta-learning to learn a shared initialization for all task-specific prompts. However, a single initialization is insufficient to obtain good prompts for all tasks and samples when the tasks are complex. Moreover, MetaPrompting requires tuning the whole MLM, causing a heavy burden on computation and memory as the MLM is usually large. To address these issues, we use a prompt pool to extract more task knowledge and construct instance-dependent prompts via attention. We further propose a novel soft verbalizer (RepVerb) which constructs label embedding from feature embeddings directly. Combining meta-learning the prompt pool and RepVerb, we propose MetaPrompter for effective structured prompting. MetaPrompter is parameter-efficient as only the pool is required to be tuned. Experimental results demonstrate that MetaPrompter performs better than the recent state-of-the-arts and RepVerb outperforms existing soft verbalizers.