GuardReasoner-VL: Safeguarding VLMs via Reinforced 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_ | 1866908367142780928 |
|---|---|
| author | Liu, Yue Zhai, Shengfang Du, Mingzhe Chen, Yulin Cao, Tri Gao, Hongcheng Wang, Cheng Li, Xinfeng Wang, Kun Fang, Junfeng Zhang, Jiaheng Hooi, Bryan |
| author_facet | Liu, Yue Zhai, Shengfang Du, Mingzhe Chen, Yulin Cao, Tri Gao, Hongcheng Wang, Cheng Li, Xinfeng Wang, Kun Fang, Junfeng Zhang, Jiaheng Hooi, Bryan |
| contents | To enhance the safety of VLMs, this paper introduces a novel reasoning-based VLM guard model dubbed GuardReasoner-VL. The core idea is to incentivize the guard model to deliberatively reason before making moderation decisions via online RL. First, we construct GuardReasoner-VLTrain, a reasoning corpus with 123K samples and 631K reasoning steps, spanning text, image, and text-image inputs. Then, based on it, we cold-start our model's reasoning ability via SFT. In addition, we further enhance reasoning regarding moderation through online RL. Concretely, to enhance diversity and difficulty of samples, we conduct rejection sampling followed by data augmentation via the proposed safety-aware data concatenation. Besides, we use a dynamic clipping parameter to encourage exploration in early stages and exploitation in later stages. To balance performance and token efficiency, we design a length-aware safety reward that integrates accuracy, format, and token cost. Extensive experiments demonstrate the superiority of our model. Remarkably, it surpasses the runner-up by 19.27% F1 score on average. We release data, code, and models (3B/7B) of GuardReasoner-VL at https://github.com/yueliu1999/GuardReasoner-VL/ |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_11049 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | GuardReasoner-VL: Safeguarding VLMs via Reinforced Reasoning Liu, Yue Zhai, Shengfang Du, Mingzhe Chen, Yulin Cao, Tri Gao, Hongcheng Wang, Cheng Li, Xinfeng Wang, Kun Fang, Junfeng Zhang, Jiaheng Hooi, Bryan Artificial Intelligence Cryptography and Security To enhance the safety of VLMs, this paper introduces a novel reasoning-based VLM guard model dubbed GuardReasoner-VL. The core idea is to incentivize the guard model to deliberatively reason before making moderation decisions via online RL. First, we construct GuardReasoner-VLTrain, a reasoning corpus with 123K samples and 631K reasoning steps, spanning text, image, and text-image inputs. Then, based on it, we cold-start our model's reasoning ability via SFT. In addition, we further enhance reasoning regarding moderation through online RL. Concretely, to enhance diversity and difficulty of samples, we conduct rejection sampling followed by data augmentation via the proposed safety-aware data concatenation. Besides, we use a dynamic clipping parameter to encourage exploration in early stages and exploitation in later stages. To balance performance and token efficiency, we design a length-aware safety reward that integrates accuracy, format, and token cost. Extensive experiments demonstrate the superiority of our model. Remarkably, it surpasses the runner-up by 19.27% F1 score on average. We release data, code, and models (3B/7B) of GuardReasoner-VL at https://github.com/yueliu1999/GuardReasoner-VL/ |
| title | GuardReasoner-VL: Safeguarding VLMs via Reinforced Reasoning |
| topic | Artificial Intelligence Cryptography and Security |
| url | https://arxiv.org/abs/2505.11049 |