arXiv cs.CL· Zheng Chen, Fei Yu, Haohao Huang, Yang Li, Anlong Chen, Lei Chen·· 3 天前AI 评分32
SecJev：为 System One 决策模型注入安全领域专业知识
SecJev: Bringing Security Expertise to System One Decision Models
AI 导读
面向安全场景的 Jev 风格决策模型家族 SecJev 发布，参数规模覆盖 0.8B 至 9B，可从文本、遥测数据和观测历史中学习布尔、选择与有序决策。其中 SecJev-0.8B 在任务宏平均准确率上比通用 Kev-9B 高出 20.51 个百分点，0.8B 与 4B 版本在准确率和延迟接近仅答案生成式微调的同时，峰值 GPU 显存占用更低。
正文 · 原文
Abstract:Security workflows need models that turn complex observations and explicit policies into decisions. System One models introduced by Jev return typed predictions and probabilities; security specialization supplies the domain expertise behind those predictions. We introduce SecJev, to our knowledge the first family of Jev-style decision models specialized for security, spanning 0.8B to 9B parameters. Built on Kev's single-pass candidate scorer, SecJev learns Boolean, choice, and ordered decisions from text, telemetry, and observation histories. We develop SecJev-Corpus to unify source-label prediction and explicit-policy evaluation across 14 tasks and eight sources. It covers tool outputs, traffic, federated updates, consensus, authentication, and vehicle messages. Scene-weighted training adapts the models across these domains while preserving a shared typed decision interface. Security specialization improves every model in the family; SecJev-0.8B outperforms general Kev-9B by 20.51 percentage points in task-macro accuracy. At 0.8B and 4B, SecJev achieves accuracy and latency close to answer-only generative fine-tuning with lower peak allocated GPU memory. SecJev-4B improves on Kev-4B and security-trained generation in prompt-injection and traffic transfer tests. We release adapters, decision heads, SecJev-Corpus, and training and inference code.
| Comments: | 16 pages, 1 figure; updated experiments and analysis, revised presentation, and streamlined appendix |
| Subjects: | Cryptography and Security (cs.CR); Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.03073 [cs.CR] |
| (or arXiv:2610.03073v2 [cs.CR] for this version) | |
| https://doi.org/10.48550/arXiv.2610.03073 arXiv-issued DOI via DataCite |
Submission history
From: Zheng Chen [view email]
[v1]
Fri, 2 Oct 2026 09:57:05 UTC (286 KB)
[v2]
Wed, 14 Oct 2026 00:59:10 UTC (278 KB)
来源：arXiv cs.CL · arxiv.org