ProGen: Progressive Zero-shot Dataset Generation via In-context Feedback

Jiacheng Ye, Jiahui Gao, Zhiyong Wu, Jiangtao Feng, Tao Yu, Lingpeng Kong


Abstract
Recently, dataset-generation-based zero-shot learning has shown promising results by training a task-specific model with a dataset synthesized from large pre-trained language models (PLMs). The final task-specific model often achieves compatible or even better performance than PLMs under the zero-shot setting, with orders of magnitude fewer parameters.However, synthetic datasets have their drawbacks. They have long being suffering from the low-quality issue (e.g., low informativeness, redundancy). This explains why the massive synthetic data does not lead to better performance – a scenario we would expect in the human-labeled data. To improve the quality in dataset synthesis, we propose a progressive zero-shot dataset generation framework, ProGen, which leverages the feedback from the task-specific model to guide the generation of new training data via in-context examples.Extensive experiments on five text classification datasets demonstrate the effectiveness of the proposed approach. We also show ProGen achieves on-par or superior performance with only 1% synthetic dataset size, when comparing to baseline methods without in-context feedback.
Anthology ID:
2022.findings-emnlp.269
Volume:
Findings of the Association for Computational Linguistics: EMNLP 2022
Month:
December
Year:
2022
Address:
Abu Dhabi, United Arab Emirates
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
3671–3683
Language:
URL:
https://aclanthology.org/2022.findings-emnlp.269
DOI:
Bibkey:
Cite (ACL):
Jiacheng Ye, Jiahui Gao, Zhiyong Wu, Jiangtao Feng, Tao Yu, and Lingpeng Kong. 2022. ProGen: Progressive Zero-shot Dataset Generation via In-context Feedback. In Findings of the Association for Computational Linguistics: EMNLP 2022, pages 3671–3683, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics.
Cite (Informal):
ProGen: Progressive Zero-shot Dataset Generation via In-context Feedback (Ye et al., Findings 2022)
Copy Citation:
PDF:
https://preview.aclanthology.org/ingestion-script-update/2022.findings-emnlp.269.pdf