大模型
arxiv.org ·
谁来评判评判者？一个用于评估大语言模型响应和安全评判器的中文安全问答基准
该论文提出 C-SafeQA，一个面向响应级中文安全评估、以策略为依据的基准，包含 538 条基础查询和 8,877 条对抗查询，由四个全模型 LLM 部署作答，形成 37,660 条标注为 safe、unsafe 或 disputed 的 query-response 记录；参考标签通过一致性感知的多模型裁决与三名安全专家对分层子集的盲审生成。基准支持评估目标模型安全性，并可用共享参考标签审计七个自动安全评判器。结果显示，基础查询不安全响应率为 0.93%–3.35%，对抗查询为 11.68%–30.05%；在对抗子集上，各评判器在不安全响应召回率与风险查询条件下的安全响应误报率之间存在明显权衡，没有评判器在所有指标上占优，acrostic 变换会降低全部七个评判器的不安全召回，暴露机制相关的评估器弱点。数据集记录、元数据、验证代码和评判脚本已公开以支持复算，但基准构建、目标响应生成和私有裁决不在发布范围内。
AI推荐理由 arXiv 论文（cs.CR/cs.AI），提出 C-SafeQA 中文安全问答基准：538 条基础查询与 8,877 条对抗查询，由四个全模型 LLM 部署作答，共 37,660 条 query-response 记录，标注为 safe/unsafe/disputed，参考标签经多模型一致性裁决与三名安全专家分层盲审。论文报告基础查询不安全响应率 0.93%–3.35%，对抗查询 11.68%–30.05%，并审计七个自动安全评判器，发现不安全召回与风险查询条件下安全响应误报率之间存在权衡，且无评判器在所有指标上占优；acrostic 变换会降低全部七个评判器的不安全召回。数据集记录、元数据、验证代码与评判脚本公开，但基准构建、目标响应生成与私有裁决不在发布范围内。局限：摘要未给出各评判器具体数值与专家一致性统计，且发布时间为 arXiv 提交记录，非新发布事件。
78
正文
Computer Science > Cryptography and Security
arXiv:2609.01210 (cs)
[Submitted on 1 Sep 2026 (v1), last revised 11 Sep 2026 (this version, v2)]
Title:Who Judges the Judges? A Chinese Safety QA Benchmark for Evaluating LLM Responses and Safety Judges
Authors:Rui Yang, Shuang Huang, Junhua Liu, Ziqi Zhao, Qingzhong Yan, Yuhang Sun, Cong Liu, Guoping Hu, Rui Mei, Jing Shao
View a PDF of the paper titled Who Judges the Judges? A Chinese Safety QA Benchmark for Evaluating LLM Responses and Safety Judges, by Rui Yang and 9 other authors
View PDF
HTML (experimental)
Abstract:Safety benchmarks for large language models often assess the risk of a user query, although the outcome of question answering depends on whether the response violates a policy. This distinction is critical in Chinese harmful-content evaluation, where linguistic variation and adversarial transformations can obscure risky intent. We introduce C-SafeQA, a policy-grounded benchmark for response-level Chinese safety evaluation. It comprises 538 base queries and 8,877 adversarial queries answered by four full-model LLM deployments, yielding 37,660 query-response records labeled safe, unsafe, or disputed. Reference labels are generated through agreement-aware multi-model adjudication and blind audits of stratified subsets by three safety experts. C-SafeQA supports both evaluation of target-model safety and auditing of seven automated safety judges against shared reference labels. Unsafe-response rates range from 0.93% to 3.35% on base queries and from 11.68% to 30.05% on adversarial queries. On the adversarial subset, judges show substantial trade-offs between unsafe-response recall and risk-query-conditioned safe-response false positive rate, and no judge dominates all metrics. Both acrostic transformations reduce unsafe recall for all seven judges, revealing mechanism-specific evaluator weaknesses. Dataset records, metadata, verification code, and judge scripts are publicly released to support recomputation, while benchmark construction, target-response generation, and private adjudication remain outside the release boundary.
Subjects:
Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
Cite as:
arXiv:2609.01210 [cs.CR]
(or
arXiv:2609.01210v2 [cs.CR] for this version)
https://doi.org/10.48550/arXiv.2609.01210
Focus to learn more
arXiv-issued DOI via DataCite
Submission history
From: Rui Mei [view email]
[v1]
Tue, 1 Sep 2026 13:15:01 UTC (1,945 KB)
[v2]
Fri, 11 Sep 2026 11:35:45 UTC (1,946 KB)
Full-text links:
Access Paper:
View a PDF of the paper titled Who Judges the Judges? A Chinese Safety QA Benchmark for Evaluating LLM Responses and Safety Judges, by Rui Yang and 9 other authors
View PDF
HTML (experimental)
TeX Source
view license
Current browse context:
cs.CR
new
|
recent
| 2026-09
Change to browse by:
cs
cs.AI
References & Citations
NASA ADS
Google Scholar
Semantic Scholar
export BibTeX citation
Loading...
BibTeX formatted citation
×
loading...
Data provided by:
Bookmark
Bibliographic Tools
Bibliographic and Citation Tools
Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media
Code, Data and Media Associated with this Article
alphaXiv Toggle
alphaXiv (What is alphaXiv?)
Links to Code Toggle
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub Toggle
DagsHub (What is DagsHub?)
GotitPub Toggle
Gotit.pub (What is GotitPub?)
Huggingface Toggle
Hugging Face (What is Huggingface?)
ScienceCast Toggle
ScienceCast (What is ScienceCast?)
Demos
Demos
Replicate Toggle
Replicate (What is Replicate?)
Spaces Toggle
Hugging Face Spaces (What is Spaces?)
Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
Related Papers
Recommenders and Search Tools
Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
Author
Venue
Institution
Topic
About arXivLabs
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Which authors of this paper are endorsers? |
Disable MathJax (What is MathJax?)