Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Generating Synthetic Datasets for Few-shot Prompt Tuning

Xu Guo, Zilin Du, Boyang Li, Chunyan Miao

TL;DR

Results on QQP, MRPC, and SICK datasets are even comparable to the performance of transfer learning from large real-world datasets, showing the promise of synthetic data as an alternative for enhancing soft prompt tuning.

Abstract

A major limitation of prompt tuning is its dependence on large labeled training datasets. Under few-shot learning settings, prompt tuning lags far behind full-model fine-tuning, limiting its scope of application. In this paper, we leverage the powerful LLMs to synthesize task-specific labeled data for training the soft prompts. We first introduce a distribution-aligned weighted generator tuning (DawGen) method to encourage generating in-distribution data that aligns with the few-shot real data. Then, we train soft prompts on both synthetic and real datasets using a gradient surgery approach, which eliminates the conflicting gradients from different data sources. Experiments on seven sentence-pair classification datasets demonstrate the effectiveness of our proposed method for boosting prompt tuning in few-shot learning settings. Results on QQP, MRPC, and SICK datasets are even comparable to the performance of transfer learning from large real-world datasets, showing the promise of synthetic data as an alternative for enhancing soft prompt tuning.

Generating Synthetic Datasets for Few-shot Prompt Tuning

TL;DR

Results on QQP, MRPC, and SICK datasets are even comparable to the performance of transfer learning from large real-world datasets, showing the promise of synthetic data as an alternative for enhancing soft prompt tuning.

Abstract

A major limitation of prompt tuning is its dependence on large labeled training datasets. Under few-shot learning settings, prompt tuning lags far behind full-model fine-tuning, limiting its scope of application. In this paper, we leverage the powerful LLMs to synthesize task-specific labeled data for training the soft prompts. We first introduce a distribution-aligned weighted generator tuning (DawGen) method to encourage generating in-distribution data that aligns with the few-shot real data. Then, we train soft prompts on both synthetic and real datasets using a gradient surgery approach, which eliminates the conflicting gradients from different data sources. Experiments on seven sentence-pair classification datasets demonstrate the effectiveness of our proposed method for boosting prompt tuning in few-shot learning settings. Results on QQP, MRPC, and SICK datasets are even comparable to the performance of transfer learning from large real-world datasets, showing the promise of synthetic data as an alternative for enhancing soft prompt tuning.

Paper Structure

This paper contains 25 sections, 10 equations, 1 figure, 8 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Few-shot Learning with Pre-trained Language Models
  4. LLMs as Task-specific Training Data Generators
  5. Learning with Synthetic Training Data
  6. Synthesizing Training Data for Prompt Tuning
  7. Preliminaries
  8. Distribution-Aligned Weighted Generator Tuning
  9. Training Soft Prompts with Synthetic Data Augmentation
  10. Experiments
  11. Datasets, Metrics, and Settings
  12. Baselines
  13. Few-shot learning performance
  14. The order in which synthetic and real data appears does matter
  15. Ablation study
  16. ...and 10 more sections

Figures (1)

  • Figure 1: A schematic overview of the framework.