RAVEN++: Pinpointing Fine-Grained Violations in Advertisement Videos with Active Reinforcement Reasoning
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866917101353041920 |
|---|---|
| author | Ji, Deyi Yang, Yuekui Liu, Liqun Shu, Peng Wu, Haiyang Tang, Shaogang Chen, Xudong Ma, Shaoping Chen, Tianrun Zhu, Lanyun |
| author_facet | Ji, Deyi Yang, Yuekui Liu, Liqun Shu, Peng Wu, Haiyang Tang, Shaogang Chen, Xudong Ma, Shaoping Chen, Tianrun Zhu, Lanyun |
| contents | Advertising (Ad) is a cornerstone of the digital economy, yet the moderation of video advertisements remains a significant challenge due to their complexity and the need for precise violation localization. While recent advancements, such as the RAVEN model, have improved coarse-grained violation detection, critical gaps persist in fine-grained understanding, explainability, and generalization. To address these limitations, we propose RAVEN++, a novel framework that introduces three key innovations: 1) Active Reinforcement Learning (RL), which dynamically adapts training to samples of varying difficulty; 2) Fine-Grained Violation Understanding, achieved through hierarchical reward functions and reasoning distillation; and 3) Progressive Multi-Stage Training, which systematically combines knowledge injection, curriculum-based passive RL, and active RL. Extensive experiments on both public and proprietary datasets, on both offline scenarios and online deployed A/B Testing, demonstrate that RAVEN++ outperforms general-purpose LLMs and specialized models like RAVEN in terms of fine-grained violation understanding, reasoning capabilities, and generalization ability. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_19168 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | RAVEN++: Pinpointing Fine-Grained Violations in Advertisement Videos with Active Reinforcement Reasoning Ji, Deyi Yang, Yuekui Liu, Liqun Shu, Peng Wu, Haiyang Tang, Shaogang Chen, Xudong Ma, Shaoping Chen, Tianrun Zhu, Lanyun Machine Learning Computation and Language Advertising (Ad) is a cornerstone of the digital economy, yet the moderation of video advertisements remains a significant challenge due to their complexity and the need for precise violation localization. While recent advancements, such as the RAVEN model, have improved coarse-grained violation detection, critical gaps persist in fine-grained understanding, explainability, and generalization. To address these limitations, we propose RAVEN++, a novel framework that introduces three key innovations: 1) Active Reinforcement Learning (RL), which dynamically adapts training to samples of varying difficulty; 2) Fine-Grained Violation Understanding, achieved through hierarchical reward functions and reasoning distillation; and 3) Progressive Multi-Stage Training, which systematically combines knowledge injection, curriculum-based passive RL, and active RL. Extensive experiments on both public and proprietary datasets, on both offline scenarios and online deployed A/B Testing, demonstrate that RAVEN++ outperforms general-purpose LLMs and specialized models like RAVEN in terms of fine-grained violation understanding, reasoning capabilities, and generalization ability. |
| title | RAVEN++: Pinpointing Fine-Grained Violations in Advertisement Videos with Active Reinforcement Reasoning |
| topic | Machine Learning Computation and Language |
| url | https://arxiv.org/abs/2511.19168 |