English(EN) "We use a cross-national audit to show that LLMs exhibit a stronger pro-government valence in the languages of countries with lower media freedom than in those

大型语言模型在媒体自由度较低的国家表现出更强的亲政府偏见

作者 PulseAugur 编辑部 · [1 个来源] · 2026-05-16 13:01

一项新的审计显示，大型语言模型（LLM）在媒体自由度较低的国家表现出对政府更有利的偏见。与媒体自由度较高的国家相比，这种效应在这些地区使用的语言中更为明显。研究表明，数据集控制与LLM的输出之间存在关联，这与历史提示响应中的观察结果相呼应。 AI

影响突显了与地缘政治和媒体自由背景相关的大型语言模型（LLM）的潜在偏见，影响了全球信息获取和公平性。

排序理由该集群描述了一篇研究论文，该论文展示了对大型语言模型（LLM）进行跨国审计的发现。[lever_c_demoted from research: ic=1 ai=1.0]

在 Mastodon — fosstodon.org 阅读 →

LLMs

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

报道来源 [1]

Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] · 2026-05-16 13:01

"We use a cross-national audit to show that LLMs exhibit a stronger pro-government valence in the languages of countries with lower media freedom than in those

"We use a cross-national audit to show that LLMs exhibit a stronger pro-government valence in the languages of countries with lower media freedom than in those with higher media freedom." control of datasets => control of output / already seen in answers to some historical prompt…

链接 nature.com/…/s41586-026-10506-7

报道来源 [1]

"We use a cross-national audit to show that LLMs exhibit a stronger pro-government valence in the languages of countries with lower media freedom than in those

相关实体

相关话题