[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84978-en":3,"doc-seo-84978-105":29,"detail-sidebar-cat-0-en-105":83},{"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},84978,7971461740909,"Levi","https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d",8,"Research & Report","R3: Advertisement Compliance Rectification via Group-Relative Experience Extractor and Curriculum Reinforcement","Rigorous content moderation is crucial for online advertising but causes millions of daily rejections, making manual rectification infeasible for video ads. Existing safety-oriented approaches often over-edit, sacrificing the advertiser’s original semantic intent just to meet compliance. R3 targets textual violations in video advertisements, covering speech transcripts and on-screen text. The framework combines experience-driven group-relative supervision synthesis, curriculum reinforcement learning with hierarchical rewards for compliance and intent consistency, and an industrial video rectification pipeline integrating recognition, rewriting, and rerendering.","R3 : Advertisement Compliance Rectification via Group-Relative Experience Extractor and Curriculum Reinforcement  \nYuan Chen* Zhenyu Hu* Mengge Xue† Te Cao Liqun Liu‡  \nPeng Shu Huan Yu Jie Jiang  \nTencent  \n{izayoiychen, mapleshu, berryxue, rosaliecao, [liqunliu}@tencent.com](liqunliu}@tencent.com)[ ](liqunliu}@tencent.com){archershu, huanyu, [zeus}@tencent.com](zeus}@tencent.com)  \narXiv :2607 .073 18v 1 [ cs .CL] 8 Jul 2026  \nAbstract  \nRigorous content moderation is crucial for online advertising but leads to millions of daily rejections. This scale renders manual rectification infeasible, particularly for video advertisements. However, existing safety-driven methods often suffer from aggressive over-editing, which compromises the advertiser’s original semantic intent merely to satisfy compliance. In this work, we target the rectification of textual violations in video ads, covering both speech transcripts and on-screen text. We propose R3 , a novel framework designed to harmonize compliance with original semantic intent preservation. Our approach integrates three key innovations: (1) an experience-driven data synthesis framework that bootstraps high-quality supervision via group-Relative compliance experience extractor; (2) a curriculum Reinforcement learning strategy with hierarchical rewards designed to enforce compliance while maximizing semantic consistency; and (3) a comprehensive video Rectification framework seamlessly integrating text recognition, rewriting, and rerendering for industrial deployment. Extensive experiments on industrial datasets and online A/B testing demonstrate that R3 significantly outperforms state-of-the-art baselines, achieving an optimal trade-off between violation rectification and intent preservation.  \n1 Introduction  \nAdvertising serves as a cornerstone of the digital economy, acting as the primary engine for revenue and growth across online platforms (Rathee and Milfeld, 2024 ; Campbell et al., 2025) . In pursuit of strict regulatory compliance and user safety, these platforms impose rigorous content moderation policies (Ji et al., 2025b,a; Madio and Quinn, 2025) . However, advertisers often struggle to navigate the  \n*Equal Contribution.†Project Leader.‡Corresponding Author.  \ncomplexity of these moderation rules, resulting in millions of advertisements being rejected daily. Consequently, it is imperative for online platforms to assist advertisers in automatically rectifying ad material, thereby unlocking ad supply and enhancing the overall advertiser experience (Xia et al., 2025) . With the advent of the 5G era, video has emerged as the predominant medium for information consumption, a trend particularly evident in online advertising. Consequently, violations within video advertisements constitute a significant proportion of overall content compliance issues. While these violations span both visual and textual modalities, this work focus specifically on the rectification of textual violations within video ads.  \nWhile recent advancements in Large Language Models (LLMs) (Touvron et al., 2023 ; Yang et al., 2025 ; OpenAI, 2023) have bolstered capabilities in content moderation and compliance rectification (Pi et al., 2024 ; Laugier et al., 2021), directly applying them to video ad rectification remains non-trivial. Specifically, deploying generalpurpose or naively fine-tuned models for this task faces significant challenges: 1) Inadequate compliance: Ad moderation policies are voluminous and highly context-dependent, thus that generalpurpose models often fail to grasp the nuanced boundaries of moderation rules via direct prompting, leading to frequent hallucinations or missed detections of subtle violations. This necessitates domain-specific alignment, yet standard Supervised Fine-Tuning (SFT) is hindered by data availability; 2) Prohibitive Annotation Costs: The prerequisite data annotation phase for supervised fine-tuning is severely constrained by the complexity and rapid evolution o","cbCaij1mk9pCo7b8","https://ap.wps.com/l/cbCaij1mk9pCo7b8","pdf",1035130,1,13,"English","en",105,"# Introduction\n## Problem motivation and challenges\n## Proposed solution overview\n# Approach\n## Experience-driven data synthesis\n## Curriculum RL with hierarchical rewards\n## Video rectification pipeline (recognition, rewriting, rerendering)","[{\"question\":\"How does R3 preserve semantic intent while ensuring compliance?\",\"answer\":\"R3 integrates group-relative compliance experience extraction for better supervision, curriculum reinforcement learning with hierarchical rewards, and a full rectification pipeline that performs text recognition, rewriting, and rerendering for coherent output.\"}]",1784199953,33,{"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":78,"head_meta":80,"extra_data":82,"updated_unix":27},"r3-advertisement-compliance-rectification-via-group-relative-experience-extractor-and-curriculum-reinforcement","",{"@graph":35,"@context":77},[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/r3-advertisement-compliance-rectification-via-group-relative-experience-extractor-and-curriculum-reinforcement/84978/",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","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71],{"name":72,"@type":73,"acceptedAnswer":74},"How does R3 preserve semantic intent while ensuring compliance?","Question",{"text":75,"@type":76},"R3 integrates group-relative compliance experience extraction for better supervision, curriculum reinforcement learning with hierarchical rewards, and a full rectification pipeline that performs text recognition, rewriting, and rerendering for coherent output.","Answer","https://schema.org",{"og:url":51,"og:type":79,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":81,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":84},[85,89,93,97,102,107,112,115,120,123,127],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":86,"show_sort_weight":87,"slug":88},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":90,"show_sort_weight":91,"slug":92},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Exam",70,"exam",{"id":98,"doc_module":4,"doc_module_name":45,"category_name":99,"show_sort_weight":100,"slug":101},5,"Comic",60,"comic",{"id":103,"doc_module":4,"doc_module_name":45,"category_name":104,"show_sort_weight":105,"slug":106},6,"Technology",50,"technology",{"id":108,"doc_module":4,"doc_module_name":45,"category_name":109,"show_sort_weight":110,"slug":111},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":113,"slug":114},30,"research-report",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},9,"Religion & Spirituality",20,"religion-spirituality",{"id":118,"doc_module":4,"doc_module_name":45,"category_name":121,"show_sort_weight":118,"slug":122},"World Cup","world-cup",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":124,"slug":126},10,"Lifestyle","lifestyle",{"id":128,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":98,"slug":130},19,"General","general"]