Qwen3Guard Technical Report
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866912651699814400 |
|---|---|
| author | Zhao, Haiquan Yuan, Chenhan Huang, Fei Hu, Xiaomeng Zhang, Yichang Yang, An Yu, Bowen Liu, Dayiheng Zhou, Jingren Lin, Junyang Yang, Baosong Cheng, Chen Tang, Jialong Jiang, Jiandong Zhang, Jianwei Xu, Jijie Yan, Ming Sun, Minmin Zhang, Pei Xie, Pengjun Tang, Qiaoyu Zhu, Qin Zhang, Rong Wu, Shibin Zhang, Shuo He, Tao Tang, Tianyi Xia, Tingyu Liao, Wei Shen, Weizhou Yin, Wenbiao Zhou, Wenmeng Yu, Wenyuan Wang, Xiaobin Deng, Xiaodong Xu, Xiaodong Zhang, Xinyu Liu, Yang Li, Yeqiu Zhang, Yi Jiang, Yong Wan, Yu Zhou, Yuxin |
| author_facet | Zhao, Haiquan Yuan, Chenhan Huang, Fei Hu, Xiaomeng Zhang, Yichang Yang, An Yu, Bowen Liu, Dayiheng Zhou, Jingren Lin, Junyang Yang, Baosong Cheng, Chen Tang, Jialong Jiang, Jiandong Zhang, Jianwei Xu, Jijie Yan, Ming Sun, Minmin Zhang, Pei Xie, Pengjun Tang, Qiaoyu Zhu, Qin Zhang, Rong Wu, Shibin Zhang, Shuo He, Tao Tang, Tianyi Xia, Tingyu Liao, Wei Shen, Weizhou Yin, Wenbiao Zhou, Wenmeng Yu, Wenyuan Wang, Xiaobin Deng, Xiaodong Xu, Xiaodong Zhang, Xinyu Liu, Yang Li, Yeqiu Zhang, Yi Jiang, Yong Wan, Yu Zhou, Yuxin |
| contents | As large language models (LLMs) become more capable and widely used, ensuring the safety of their outputs is increasingly critical. Existing guardrail models, though useful in static evaluation settings, face two major limitations in real-world applications: (1) they typically output only binary "safe/unsafe" labels, which can be interpreted inconsistently across diverse safety policies, rendering them incapable of accommodating varying safety tolerances across domains; and (2) they require complete model outputs before performing safety checks, making them fundamentally incompatible with streaming LLM inference, thereby preventing timely intervention during generation and increasing exposure to harmful partial outputs. To address these challenges, we present Qwen3Guard, a series of multilingual safety guardrail models with two specialized variants: Generative Qwen3Guard, which casts safety classification as an instruction-following task to enable fine-grained tri-class judgments (safe, controversial, unsafe); and Stream Qwen3Guard, which introduces a token-level classification head for real-time safety monitoring during incremental text generation. Both variants are available in three sizes (0.6B, 4B, and 8B parameters) and support up to 119 languages and dialects, providing comprehensive, scalable, and low-latency safety moderation for global LLM deployments. Evaluated across English, Chinese, and multilingual benchmarks, Qwen3Guard achieves state-of-the-art performance in both prompt and response safety classification. All models are released under the Apache 2.0 license for public use. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_14276 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Qwen3Guard Technical Report Zhao, Haiquan Yuan, Chenhan Huang, Fei Hu, Xiaomeng Zhang, Yichang Yang, An Yu, Bowen Liu, Dayiheng Zhou, Jingren Lin, Junyang Yang, Baosong Cheng, Chen Tang, Jialong Jiang, Jiandong Zhang, Jianwei Xu, Jijie Yan, Ming Sun, Minmin Zhang, Pei Xie, Pengjun Tang, Qiaoyu Zhu, Qin Zhang, Rong Wu, Shibin Zhang, Shuo He, Tao Tang, Tianyi Xia, Tingyu Liao, Wei Shen, Weizhou Yin, Wenbiao Zhou, Wenmeng Yu, Wenyuan Wang, Xiaobin Deng, Xiaodong Xu, Xiaodong Zhang, Xinyu Liu, Yang Li, Yeqiu Zhang, Yi Jiang, Yong Wan, Yu Zhou, Yuxin Computation and Language As large language models (LLMs) become more capable and widely used, ensuring the safety of their outputs is increasingly critical. Existing guardrail models, though useful in static evaluation settings, face two major limitations in real-world applications: (1) they typically output only binary "safe/unsafe" labels, which can be interpreted inconsistently across diverse safety policies, rendering them incapable of accommodating varying safety tolerances across domains; and (2) they require complete model outputs before performing safety checks, making them fundamentally incompatible with streaming LLM inference, thereby preventing timely intervention during generation and increasing exposure to harmful partial outputs. To address these challenges, we present Qwen3Guard, a series of multilingual safety guardrail models with two specialized variants: Generative Qwen3Guard, which casts safety classification as an instruction-following task to enable fine-grained tri-class judgments (safe, controversial, unsafe); and Stream Qwen3Guard, which introduces a token-level classification head for real-time safety monitoring during incremental text generation. Both variants are available in three sizes (0.6B, 4B, and 8B parameters) and support up to 119 languages and dialects, providing comprehensive, scalable, and low-latency safety moderation for global LLM deployments. Evaluated across English, Chinese, and multilingual benchmarks, Qwen3Guard achieves state-of-the-art performance in both prompt and response safety classification. All models are released under the Apache 2.0 license for public use. |
| title | Qwen3Guard Technical Report |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2510.14276 |