头条
5 家媒体·5 篇报道

OpenAI的“妖精”问题为何重要——以及如何自行释放这些妖精

OpenAI的Codex系统提示中明确指示“绝不谈论妖精”。从GPT-5.1起,模型开始在隐喻中提及妖精、妖怪等类似生物,促使OpenAI加入缓解指令以抑制此行为。公司承认了这一问题,并指出其与模型开发的相关性。

本文综合以下全部报道写成,而非依据任何单一来源。

本页由机器翻译生成,原始报道为其原文语言。

各家媒体如何报道

  1. Ars Technica·
    OpenAI Codex system prompt includes explicit directive to "never talk about goblins"
  2. TechMeme·
    OpenAI says its models, starting with GPT-5.1, "increasingly mentioned goblins, gremlins, and other creatures", leading to prompt instructions to mitigate it (OpenAI)
  3. The Verge·
    OpenAI talks about not talking about goblins
  4. Engadget·
    ChatGPT developed a goblin obsession after OpenAI tried to make it nerdy
  5. VentureBeat·
    Why OpenAI's 'goblin' problem matters — and how you can release the goblins on your own