R^3: Advertisement Compliance Rectification via Group-Relative Experience Extractor and Curriculum Reinforcement
Quick Answer
This paper shows that R^3 introduces a novel framework for rectifying textual violations in video ads, integrating a group-relative experience extractor and a curriculum reinforcement learning strategy.
Quick Take
R^3 introduces a novel framework for rectifying textual violations in video ads, integrating a group-relative experience extractor and a curriculum reinforcement learning strategy. Extensive experiments show R^3 significantly outperforms existing methods, achieving a balance between compliance and semantic intent preservation in industrial applications.
Key Points
- R^3 targets textual violations in video ads, including speech transcripts and on-screen text.
- The framework utilizes a group-relative compliance experience extractor for high-quality supervision.
- Curriculum reinforcement learning maximizes semantic consistency while enforcing compliance.
- R^3 integrates text recognition, rewriting, and re-rendering for seamless industrial deployment.
- Extensive A/B testing shows R^3 outperforms state-of-the-art baselines in violation rectification.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Rigorous 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 R^3, 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 a 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 re-rendering for industrial deployment. Extensive experiments on industrial datasets and online A/B testing demonstrate that R^3 significantly outperforms state-of-the-art baselines, achieving an optimal trade-off between violation rectification and intent preservation.
| Comments: | ACL 2026 (Poster, Industry Track) |
| Subjects: | Computation and Language (cs.CL); Machine Learning (cs.LG) |
| Cite as: | arXiv:2607.07318 [cs.CL] |
| (or arXiv:2607.07318v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2607.07318 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Yuan Chen [view email]
[v1]
Wed, 8 Jul 2026 12:05:41 UTC (760 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →Quantifying Prior Dominance in Systems
The study introduces the Normalized Context Utilization (NCU) metric to evaluate Retrieval-Augmented Generation (RAG) systems, revealing that Small Language Models (SLMs) outperform larger models in factual extraction. The findings indicate that traditional scaling laws yield diminishing returns, with a commercial API frequently failing against adversarial evidence due to systemic confidence collapse.