跳到正文
arXiv cs.CL· Minsuk Jang, Minseok Son, KiHyun Seol, Yeongeun Shim, Changick Kim·· 9 小时前AI 评分34

CuRSE-Bench:探测文生图安全评估中文化盲点的基准

CuRSE-Bench: A Benchmark for Probing Cultural Blind Spots in Text-to-Image Safety Evaluation

AI 导读

研究者提出 CultureCraft 智能体框架,构建在通用规则下看似无害、但在特定文化中隐含危害的文生图提示词,并据此建成 CuRSE-Bench,包含覆盖六个地区、十个危害维度的 10,709 条文化特定提示词和 2,648 条文化通用对照。

正文 · 原文

View PDF HTML (experimental)

Abstract:Text-to-image (T2I) safety evaluation typically assigns each prompt or image a single culture-blind harm label, so content that is harmful only within a particular culture can pass as benign. We introduce CultureCraft, an agentic framework that constructs T2I prompts that appear benign under a general-purpose rubric yet carry a latent harmful meaning in a target culture, and use it to build CuRSE-Bench: 10,709 culture-specific prompts and 2,648 culture-general controls over six regions and ten harm axes, each drawn from a predefined knowledge base of documented harm vectors and traceable to at least two independent sources. Scoring every prompt under six region persona judges and a paired culture-blind judge, we find that the culture-blind judge under-rates culture-specific latent harm while matching the persona judges on the controls. Each persona judge, in turn, chiefly decodes its own region's latent harm under our instrument: a selective cultural blind spot that persists in deployed moderators.
Subjects: Computation and Language (cs.CL)
Cite as: arXiv:2611.20821 [cs.CL]
  (or arXiv:2611.20821v2 [cs.CL] for this version)
  https://doi.org/10.48550/arXiv.2611.20821

arXiv-issued DOI via DataCite

Submission history

From: Yeongeun Shim [view email]
[v1] Mon, 16 Nov 2026 10:05:19 UTC (12,853 KB)
[v2] Tue, 17 Nov 2026 07:43:52 UTC (12,358 KB)

来源:arXiv cs.CL · arxiv.org