English(EN) HiveTraceGuard-Pro: A Compact Generative Guardrail for Prompt Injection, Jailbreaks, and Adversarial Obfuscation

新型护栏模型HiveTraceGuard-Pro 针对提示注入

作者 PulseAugur 编辑部 · [1 个来源] · 2026-09-02 04:00

研究人员开发了HiveTraceGuard-Pro，这是一种新颖的生成式护栏模型，旨在保护大型语言模型免受提示注入和对抗性攻击。该模型是Qwen3-0.6B的一个0.6B参数LoRA微调版本，使用俄语和英语进行训练，并采用二元评分系统将响应分类为安全或不安全。虽然在一个基准测试中，其总体性能略低于得分最高的模型，但HiveTraceGuard-Pro在俄语任务中表现出强大的鲁棒性，取得了最高的俄语干净鲁棒性组合F1分数和近乎完美的提示注入召回率。与同类模型相比，该模型的平均延迟也最低，并且其权重在Hugging Face上公开可用。 AI

影响这项研究引入了一个专门的护栏模型，可以提高LLM在多语言环境中对抗性攻击的安全性和鲁棒性。

排序理由该集群包含一篇详细介绍新模型及其评估的研究论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 29 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇详细介绍新模型及其评估的研究论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, model release; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.AI TIER_1 English(EN) · Nikita Oblakov, Sabrina Sadiekh, Evgeniy Kokuykin · 2026-09-02 04:00

HiveTraceGuard-Pro：用于提示注入、越狱和对抗性混淆的紧凑型生成式防护栏

arXiv:2609.01046v1 Announce Type: cross Abstract: Production LLMs must handle inputs that attempt to override system instructions, bypass safety policies or elicit harmful responses. A common mitigation is a separate guardrail model. Existing reports, however, provide little evid…

本文如何被排名

报道来源 [1]

HiveTraceGuard-Pro：用于提示注入、越狱和对抗性混淆的紧凑型生成式防护栏

相关实体

相关话题