[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84308-en":3,"doc-seo-84308-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},84308,1374391974585,"Genevieve","https://ap-avatar.wpscdn.com/davatar_276721f389ce27ea32af1340a28f341c",8,"Research & Report","COBART Controlled Optimized Bidirectional and Auto-Regressive Transformer for Ad Headline Generation","Online ads rely on strong headline creative, yet changing ad formats and requirements make it difficult to generate headlines that are both optimized and customizable. COBART proposes a controllable approach using prefix control tokens with BART fine-tuning to generate ad headlines with high click-through-rate (CTR). The method supports user control over headline length for different formats and devices and adapts to varied optimization criteria. Experiments show a 25.82% increase in ROUGE-L and a 5.82% gain in estimated CTR versus strong baselines.","COBART: Controlled, Optimized, Bidirectional and Auto-Regressive Transformer for Ad Headline Generation∗  \nYashal Shakti Kanungo† Amazon United States of America  \nGyanendra Das∗‡ Amazon India  \narXiv :2607 .0807 1v 1 [ cs .CL] 9 Jul 2026  \nPooja A  \nAmazon  \nIndia  \nAbstract  \nOnline ads are essential to all businesses and ad headlines are one of their core creative component. Existing methods can generate headlines automatically and also optimize their click-through-rate (CTR) and quality. However, evolving ad formats and changing creative requirements make it difficult to generate optimized & customized headlines. We propose a novel method that uses prefix control tokens along with BART [16] fine-tuning. It yields the highest CTRand also allows users to control the length of generated headlines for use across different ad formats. The method is also flexible and can easily be adapted to other architectures, creative requirements and optimization criteria. Our experiments demonstrate a 25.82% increment in Rouge-L and a 5.82% increment in estimated CTR over previously published strong ad headline generation baseline.  \nCCS Concepts  \n• Computing methodologies → Natural language generation;  \n• Information systems → Online advertising.  \nKeywords  \nad generation; controlled generation; sponsored advertising; ad optimization; headline generation; E-commerce  \n1 Introduction  \nOnline advertising is an evolving domain that is now integral to all businesses. Organizations across the world use ads to improve discovery and reach of their businesses. Sellers on E-commerce platforms use online ads to build brand awareness and connect shoppers to their products.  \nOne of the key component of ads is their ‘headline’. Headlines are a part of all ad formats including text, image, and video ads as illustrated in Figure 1 . They summarize the key properties of the underlying products and promote the customers to engage. They may also be the first preview of the product/brand for shoppers. Thus having the right style and appeal is important to attractshopper attention. Style and content are not the only factors for generating headlines. Length is equally important for ads across different formats and target device screen sizes. Long headlines  \n∗ © Kanungo, Das, Pooja A, and Negi 2022 . This is the author’s version of the work. It is posted here for your personal use. Not for redistribution. The definitive Version of Record was published in Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD’22), August 14–18, 2022, Washington, DC, USA, [https://doi.org/10.1145/3534678.3539069](https://doi.org/10.1145/3534678.3539069) .  \n†Equal contribution.  \n‡Work done during internship at Amazon.  \nSumit Negi  \nAmazon India  \n\n| Input (Single ad with multiple products; two shown) |  |  |  |\n| --- | --- | --- | --- |\n| Product 1 | Boyd’s Cosmetics NYC-No Lines Temporary Wrinkle Remover with our Brush It Away Medium   |  |  |\n| Product 2 | No Lines Temporary Wrinkle Remover . . . |  |  |\n| Generated Headlines |  |  |  |\n| UniLM\u003Cbr>BART | Boyd’s New York City Cosmetics New & improved anti-aging with our brush |  |  |\n| COBART CTR ↓ | Short | Length →\u003Cbr>Medium | Long |\n| Low\u003Cbr>(Bucket\u003Cbr>1/15) | Boyd’s\u003Cbr>Cosmetics\u003Cbr>NYC-No\u003Cbr>Lines | Save Now on\u003Cbr>New York City\u003Cbr>Collections | No Lines\u003Cbr>Temporary\u003Cbr>Wrinkle\u003Cbr>Remover with\u003Cbr>Brush It Away |\n| Medium\u003Cbr>(Bucket\u003Cbr>7/15) | Shop Boyd’s\u003Cbr>NYC, No\u003Cbr>Lines\u003Cbr>Remover | Tackle Fine\u003Cbr>Lines &\u003Cbr>Wrinkles with\u003Cbr>Boyd’s NYC | Boyd’s\u003Cbr>Cosmetics NYC\u003Cbr>No Lines\u003Cbr>Temporary\u003Cbr>Wrinkle\u003Cbr>Remover |\n| High\u003Cbr>(Bucket\u003Cbr>15/15) | Shop New\u003Cbr>York City’s\u003Cbr>\\#1 Makeup\u003Cbr>Brand | New &\u003Cbr>improved\u003Cbr>anti-aging\u003Cbr>with our brush\u003Cbr>-Boyd’s NYC | A New\u003Cbr>Generation of\u003Cbr>Makeup That\u003Cbr>Targets Fine\u003Cbr>Lines with a\u003Cbr>New Look |\n\nTable 1: Ad headline generation for a multi-product ad. Our method allows to control and optimize the length and clickthrough-rate (CTR) of headli","cbCaioPjwxRwVzuc","https://ap.wps.com/l/cbCaioPjwxRwVzuc","pdf",3950585,1,10,"English","en",105,"# Introduction\n## Ad headlines and their importance\n## Challenges in large-scale headline generation\n## Metrics and requirements (CTR, length, style)\n# Proposed approach\n## COBART with prefix control tokens and BART","[{\"question\":\"What problem does COBART address in ad headline generation?\",\"answer\":\"COBART targets the difficulty of producing ad headlines that are both optimized (especially for CTR) and customizable when ad formats and creative requirements change over time.\"},{\"question\":\"How does COBART control the generated headline output?\",\"answer\":\"COBART uses prefix control tokens together with BART fine-tuning, enabling length control during generation and supporting different optimization criteria.\"},{\"question\":\"What improvements does the paper report compared with prior baselines?\",\"answer\":\"Experiments report a 25.82% increment in ROUGE-L and a 5.82% increment in estimated CTR over previously published strong ad headline generation baselines.\"}]",1784194734,25,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"cobart-controlled-optimized-bidirectional-and-auto-regressive-transformer-for-ad-headline-generation","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/cobart-controlled-optimized-bidirectional-and-auto-regressive-transformer-for-ad-headline-generation/84308/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-28","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does COBART address in ad headline generation?","Question",{"text":75,"@type":76},"COBART targets the difficulty of producing ad headlines that are both optimized (especially for CTR) and customizable when ad formats and creative requirements change over time.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does COBART control the generated headline output?",{"text":80,"@type":76},"COBART uses prefix control tokens together with BART fine-tuning, enabling length control during generation and supporting different optimization criteria.",{"name":82,"@type":73,"acceptedAnswer":83},"What improvements does the paper report compared with prior baselines?",{"text":84,"@type":76},"Experiments report a 25.82% increment in ROUGE-L and a 5.82% increment in estimated CTR over previously published strong ad headline generation baselines.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,134],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":21,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":21,"slug":133},"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]