Oyster-I: Beyond Refusal -- Constructive Safety Alignment for Responsible Language Models
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909843206438912 |
|---|---|
| author | Duan, Ranjie Liu, Jiexi Jia, Xiaojun Zhao, Shiji Cheng, Ruoxi Wang, Fengxiang Wei, Cheng Xie, Yong Liu, Chang Li, Defeng Dong, Yinpeng Zhang, Yichi Chen, Yuefeng Wang, Chongwen Ma, Xingjun Wei, Xingxing Liu, Yang Su, Hang Zhu, Jun Li, Xinfeng Sun, Yitong Zhang, Jie Hu, Jinzhao Xu, Sha Yang, Wenchao Yang, Yitong Zhang, Xingyao Tan, Yingshui Tao, Jialing Xue, Hui |
| author_facet | Duan, Ranjie Liu, Jiexi Jia, Xiaojun Zhao, Shiji Cheng, Ruoxi Wang, Fengxiang Wei, Cheng Xie, Yong Liu, Chang Li, Defeng Dong, Yinpeng Zhang, Yichi Chen, Yuefeng Wang, Chongwen Ma, Xingjun Wei, Xingxing Liu, Yang Su, Hang Zhu, Jun Li, Xinfeng Sun, Yitong Zhang, Jie Hu, Jinzhao Xu, Sha Yang, Wenchao Yang, Yitong Zhang, Xingyao Tan, Yingshui Tao, Jialing Xue, Hui |
| contents | Large language models (LLMs) typically deploy safety mechanisms to prevent harmful content generation. Most current approaches focus narrowly on risks posed by malicious actors, often framing risks as adversarial events and relying on defensive refusals. However, in real-world settings, risks also come from non-malicious users seeking help while under psychological distress (e.g., self-harm intentions). In such cases, the model's response can strongly influence the user's next actions. Simple refusals may lead them to repeat, escalate, or move to unsafe platforms, creating worse outcomes. We introduce Constructive Safety Alignment (CSA), a human-centric paradigm that protects against malicious misuse while actively guiding vulnerable users toward safe and helpful results. Implemented in Oyster-I (Oy1), CSA combines game-theoretic anticipation of user reactions, fine-grained risk boundary discovery, and interpretable reasoning control, turning safety into a trust-building process. Oy1 achieves state-of-the-art safety among open models while retaining high general capabilities. On our Constructive Benchmark, it shows strong constructive engagement, close to GPT-5, and unmatched robustness on the Strata-Sword jailbreak dataset, nearing GPT-o1 levels. By shifting from refusal-first to guidance-first safety, CSA redefines the model-user relationship, aiming for systems that are not just safe, but meaningfully helpful. We release Oy1, code, and the benchmark to support responsible, user-centered AI. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_01909 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Oyster-I: Beyond Refusal -- Constructive Safety Alignment for Responsible Language Models Duan, Ranjie Liu, Jiexi Jia, Xiaojun Zhao, Shiji Cheng, Ruoxi Wang, Fengxiang Wei, Cheng Xie, Yong Liu, Chang Li, Defeng Dong, Yinpeng Zhang, Yichi Chen, Yuefeng Wang, Chongwen Ma, Xingjun Wei, Xingxing Liu, Yang Su, Hang Zhu, Jun Li, Xinfeng Sun, Yitong Zhang, Jie Hu, Jinzhao Xu, Sha Yang, Wenchao Yang, Yitong Zhang, Xingyao Tan, Yingshui Tao, Jialing Xue, Hui Artificial Intelligence Computation and Language Computers and Society Human-Computer Interaction Symbolic Computation Large language models (LLMs) typically deploy safety mechanisms to prevent harmful content generation. Most current approaches focus narrowly on risks posed by malicious actors, often framing risks as adversarial events and relying on defensive refusals. However, in real-world settings, risks also come from non-malicious users seeking help while under psychological distress (e.g., self-harm intentions). In such cases, the model's response can strongly influence the user's next actions. Simple refusals may lead them to repeat, escalate, or move to unsafe platforms, creating worse outcomes. We introduce Constructive Safety Alignment (CSA), a human-centric paradigm that protects against malicious misuse while actively guiding vulnerable users toward safe and helpful results. Implemented in Oyster-I (Oy1), CSA combines game-theoretic anticipation of user reactions, fine-grained risk boundary discovery, and interpretable reasoning control, turning safety into a trust-building process. Oy1 achieves state-of-the-art safety among open models while retaining high general capabilities. On our Constructive Benchmark, it shows strong constructive engagement, close to GPT-5, and unmatched robustness on the Strata-Sword jailbreak dataset, nearing GPT-o1 levels. By shifting from refusal-first to guidance-first safety, CSA redefines the model-user relationship, aiming for systems that are not just safe, but meaningfully helpful. We release Oy1, code, and the benchmark to support responsible, user-centered AI. |
| title | Oyster-I: Beyond Refusal -- Constructive Safety Alignment for Responsible Language Models |
| topic | Artificial Intelligence Computation and Language Computers and Society Human-Computer Interaction Symbolic Computation |
| url | https://arxiv.org/abs/2509.01909 |