arXiv cs.CL· Viet-Thanh Pham, Thuy-Trang Vu, Junae Kim, Rollin Omari, Dinh Phung, Gholamreza Haffari, Lizhen Qu·· 13 小时前AI 评分38
UniGuard：面向跨接口安全泛化的多视图护栏学习框架
UniGuard: Multi-View Guardrail Learning for Cross-Interface Safety Generalization
AI 导读
UniGuard 提出多视图护栏训练框架，从同一安全案例的语义对齐多视图学习接口不变的安全表征，并构建 UniGuard-FT 数据集，将 1,000 个公开安全案例转换为 prompt、response、多轮对话、agent-trace 和多智能体 trace 五种视图。
正文 · 原文
Abstract:Guardrail models are increasingly used as safety layers for LLMs and agentic systems, but most are trained on fixed prompt or response interfaces and become unreliable when the same safety case appears as a conversation, agent trajectory, multi-agent trace, or policy-conditioned input. We propose UniGuard, a training framework that learns interface-invariant safety representations from multiple semantically aligned views of the same safety case. We also construct the UniGuard-FT dataset by transforming 1,000 public safety cases into prompt, response, multi-turn conversation, agent-trace, and multi-agent-trace views, and train UniGuard-0.6B with classification, supervised contrastive learning, and asymmetric confidence calibration losses. Across unseen-interface benchmarks ATBench, PolicyGuardBench, and DynaBench, UniGuard-0.6B improves over its pretrained guardrail by 45.12 F1 points on average and reduces expected calibration error by 35.41 points, while remaining competitive with much larger guardrails such as Claude-Sonnet-4.5 and GPT-5.2 - despite being a 0.6B model and trained only with synthetic data.
| Comments: | 31 pages |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.30742 [cs.CL] |
| (or arXiv:2610.30742v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.30742 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Viet Thanh Pham [view email]
[v1]
Thu, 22 Oct 2026 05:14:30 UTC (1,582 KB)
来源：arXiv cs.CL · arxiv.org