{"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/control-prefixes-for-text-generation","title":"Control Prefixes for Parameter-Efficient Text Generation","arxiv_id":"2110.08329","date":"2021-10-15","proceeding":null,"authors":["Jordan Clive","Kris Cao","Marek Rei"],"abstract":"Prefix-tuning is a powerful lightweight technique for adapting a large pre-trained language model to a downstream application. However, it uses the same dataset-level tuned prompt for all examples in the dataset. We extend this idea and propose a dynamic method, Control Prefixes, which allows for the inclusion of conditional input-dependent information, combining the benefits of prompt tuning and controlled generation. The method incorporates attribute-level learnable representations into different layers of a pre-trained transformer, allowing for the generated text to be guided in a particular direction. We provide a systematic evaluation of the technique and apply it to five datasets from the GEM benchmark for natural language generation (NLG). Although the aim is to develop a parameter-efficient model, we show Control Prefixes can even outperform full fine-tuning methods. We present state-of-the-art results on several data-to-text datasets, including WebNLG.","url_abs":"https://arxiv.org/abs/2110.08329v2","url_pdf":"https://arxiv.org/pdf/2110.08329v2.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":"control-prefixes-for-text-generation","repo_url":"https://github.com/Yale-LILY/dart","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"control-prefixes-for-text-generation","repo_url":"https://github.com/jordiclive/ControlPrefixes","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"abstractive-text-summarization","task_name":"Abstractive Text Summarization"},{"task_slug":"attribute","task_name":"Attribute"},{"task_slug":"data-to-text-generation","task_name":"Data-to-Text Generation"},{"task_slug":"language-modeling","task_name":"Language Modeling"},{"task_slug":"language-modelling","task_name":"Language Modelling"},{"task_slug":"text-generation","task_name":"Text Generation"},{"task_slug":"text-simplification","task_name":"Text Simplification"},{"task_slug":"text-summarization","task_name":"Text Summarization"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/data-to-text-generation-on-cleaned-e2e-nlg-1","task":"Data-to-Text Generation","dataset":"Cleaned E2E NLG Challenge","model":"Control Prefixes (T5-large)","rank_in_archive_order":1,"of":7,"metrics":{"BLEU (Test set)":"44.15"},"uses_additional_data":true},{"leaderboard":"/sota/data-to-text-generation-on-webnlg","task":"Data-to-Text Generation","dataset":"WebNLG","model":"Control Prefixes (A1, T5-large)","rank_in_archive_order":1,"of":20,"metrics":{"BLEU":"67.32"},"uses_additional_data":false},{"leaderboard":"/sota/data-to-text-generation-on-webnlg","task":"Data-to-Text Generation","dataset":"WebNLG","model":"Control Prefixes (A1, A2, T5-large)","rank_in_archive_order":2,"of":20,"metrics":{"BLEU":"67.15"},"uses_additional_data":true},{"leaderboard":"/sota/data-to-text-generation-on-webnlg-full-1","task":"Data-to-Text Generation","dataset":"WebNLG Full","model":"Control Prefixes (A1, A2, T5-large)","rank_in_archive_order":1,"of":8,"metrics":{"BLEU":"62.27 "},"uses_additional_data":true},{"leaderboard":"/sota/data-to-text-generation-on-webnlg-full-1","task":"Data-to-Text Generation","dataset":"WebNLG Full","model":"Control Prefixes (A1, T5-large)","rank_in_archive_order":2,"of":8,"metrics":{"BLEU":"61.94"},"uses_additional_data":false},{"leaderboard":"/sota/text-generation-on-dart","task":"Text Generation","dataset":"DART","model":"Control Prefixes (T5-large)","rank_in_archive_order":5,"of":7,"metrics":{"METEOR":"0.411"},"uses_additional_data":false},{"leaderboard":"/sota/text-simplification-on-asset","task":"Text Simplification","dataset":"ASSET","model":"Control Prefixes (BART)","rank_in_archive_order":3,"of":12,"metrics":{"FKGL":"5.97","QuestEval (Reference-less, BERTScore)":"0.64","SARI (EASSE>=0.2.1)":"43.58"},"uses_additional_data":false},{"leaderboard":"/sota/text-simplification-on-turkcorpus","task":"Text Simplification","dataset":"TurkCorpus","model":"Control Prefixes (BART)","rank_in_archive_order":3,"of":25,"metrics":{"FKGL":"7.74","QuestEval (Reference-less, BERTScore)":"0.66","SARI (EASSE>=0.2.1)":"42.32"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2110.08329","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}