{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/progen-progressive-zero-shot-dataset","title":"ProGen: Progressive Zero-shot Dataset Generation via In-context Feedback","arxiv_id":"2210.12329","date":"2022-10-22","proceeding":null,"authors":["Jiacheng Ye","Jiahui Gao","Jiangtao Feng","Zhiyong Wu","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 been suffering from low-quality issues (e.g., low informativeness and 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 of 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 compared to baseline methods without in-context feedback.","url_abs":"https://arxiv.org/abs/2210.12329v1","url_pdf":"https://arxiv.org/pdf/2210.12329v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"progen-progressive-zero-shot-dataset","repo_url":"https://github.com/hkunlp/progen","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"progen-progressive-zero-shot-dataset","repo_url":"https://github.com/hkunlp/symgen","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":"data-free-knowledge-distillation","task_name":"Data-free Knowledge Distillation"},{"task_slug":"dataset-generation","task_name":"Dataset Generation"},{"task_slug":"informativeness","task_name":"Informativeness"},{"task_slug":"text-classification","task_name":"Text Classification"},{"task_slug":"zero-shot-learning","task_name":"Zero-Shot Learning"},{"task_slug":"text-classification-1","task_name":"text-classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/data-free-knowledge-distillation-on-qnli","task":"Data-free Knowledge Distillation","dataset":"QNLI","model":"ProGen (T5-base)","rank_in_archive_order":3,"of":4,"metrics":{"Accuracy":"85.9"},"uses_additional_data":false},{"leaderboard":"/sota/data-free-knowledge-distillation-on-squad","task":"Data-free Knowledge Distillation","dataset":"SQuAD","model":"ProGen (T5-base)","rank_in_archive_order":4,"of":4,"metrics":{"Exact Match":"68.1"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2210.12329","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}